Author: Guosheng Communication Team

Abstract

Standing at the current point in time, we re-evaluate the development trend and development trend of AGI investor expectations. The market starts with computing power, extends to GPU, optical modules, switches, storage and other tracks, and leverages overseas mapping to eagerly look forward to AI applications, but ignores the pull on upstream infrastructure when computing power increases. If applications are the most explosive direction, then the infrastructure will take a long time to develop. Not only liquid cooling, but also the demand for energy is fundamental. This is also the starting point of this article.

Marginal changes: One of the biggest differences between AIDC and traditional data centers is that the level of electricity consumption has increased significantly. AIDC has the characteristics of large data volume, complex algorithms and 24/7 instant response, so compared with traditional data centers, AIDC consumes a lot of power. With the rapid development of AI, it is expected that AI software integrating large language models will develop rapidly, and training needs and inference needs will resonate. In the future, the power consumption of data centers will increase significantly. AIDC will become a new generation of "electric tigers", and data center consumption will The proportion of electricity will further increase. SemiAnalysis predicts that global data center critical IT power demand will surge from 49GW in 2023 to 96GW in 2026, of which AI will consume approximately 40GW. Vertiv predicts that data center power consumption will increase by 100GW over the next five years, with global data center power demand rising to 140GW by 2029.

Dilemma: The U.S. power grid is unable to support the development of AI computing power. Compared with the construction speed of data centers, the current construction speed of the U.S. power grid is relatively slow and the power generation capacity is limited. Therefore, in the short term, the United States will face a power demand dilemma due to the development of AI. Currently, U.S. power supply faces obstacles such as long infrastructure construction cycles, shortage of infrastructure facilities, labor shortages, lack of experience among practitioners, and the need to coordinate multiple stakeholders when building a power grid. The rapid development of AI has caused power supply shortages in some areas. North American utility Dominion Energy said it may not be able to meet the power demand in Virginia, leading to multi-year delays in the construction of the world's fastest-growing data center hub.

Solutions: short term - natural gas, medium term - SMR nuclear power, long term - controlled nuclear fusion. The rise of AI is leading resource competition to computing power + energy. In the AI-driven digital world, computing power is the basis for iteration and innovation, and energy is the key to supporting the operation of these computing powers.key. In the short term, natural gas combined with fuel cells will provide flexible and efficient power generation solutions for data centers to meet the current rapid expansion needs. In the medium term, small modular reactors (SMRs) will become a key path to address power bottlenecks in data centers due to their stability and adaptability to distributed deployment. In the long term, controllable nuclear fusion is expected to completely break through energy supply constraints and provide unlimited and clean power support for the future computing power ecosystem. In this process, from the continuous innovation of energy technology to the efficient collaboration of computing power ecology, it not only promotes the leap of AI technology, but also reshapes the future pattern of deep integration of energy and computing.

We believe that we are still in a battle for computing power, but looking forward to the next five years, the battle for energy infrastructure may become mainstream. In the short term, the capital expenditures of CSP giants in the third quarter of this year have all reached new highs, and they tend to be on the computing power side. In the next 5-10 years, combined with the continued increase in investment in AI computing power and the current power supply situation in the United States, we believe that the current power supply in the United States will be flat. The era of 2020 is coming to an end, and the battle for computing power will gradually transform into a battle for energy. The investment plans of computing giants such as Amazon, Microsoft, and Google in nuclear power projects such as SMR have initially proved this. The addition of IT giants will significantly introduce new technologies and accelerate iteration, and investment opportunities in related energy infrastructure will gradually emerge.

Investment advice: To sum up, energy is the next battle in technological competition. Just like the process of liquid cooling from optional to mandatory, AI upstream The infrastructure track is also moving from traditional industries to core technology supporting facilities, and seizing the opportunity for layout is the key to winning in the future. It is recommended to pay attention to the core targets of U.S. stocks such as ETN, EMR, SMR, OKLO, NNE, BE, etc. For A-shares in the nuclear power, natural gas and infrastructure supply chains, it is recommended to pay attention to Guangdong Nuclear Power, Nuclear Power, New Natural Gas, CGN Mining, Jinpan Technology, Invic, and Mai Gemet, Nengke Technology, Kehua Data, Euroland, One Stone, etc.

Risk reminder: Technical and regulatory risks, high capital requirements and financing pressure, market demand and competition risks

Investment Requirements

OpenAI founder Sam Altman once said in an interview: The two important resources in the future will be computing power and energy. AI's pursuit of performance has gradually become more intense in the field of computing power, and the core factors of competition in the next stage will initially appear in energy infrastructure.

[From computing power to energy: the next battle in technological competition]

PeopleThe rise of artificial intelligence has more directly led resource competition to computing power and energy. In the AI-driven digital world, computing power is the basis for iteration and innovation, and energy is the key to supporting the operation of these computing power. "The two most important resources in the future are computing power and energy." This trend will run through every stage of AI technology development, from algorithm optimization to hardware breakthroughs to the current demand for efficient energy systems.

[Acceleration requirements for computing power and hardware limits]

The demand for AI computing power is Exponential growth. Taking the NVIDIA H100 GPU as an example, the computing power of 60 TFLOPS is promoting large-scale training of large models, and the surge in computing power has brought huge energy consumption challenges. Vertiv predicts that total installed power demand for global data centers is expected to soar from 40GW to 140GW by 2029, while the value of data centers per MW will increase from US$2.5-3 million to US$3-3.5 million. The power consumption of more than 1MW in a single cabinet of NVIDIA's next-generation product Rubin ultra also shows that the increase in AI computing power is exerting unprecedented pressure on power infrastructure. How fast the calculation can be done depends largely on the power.

[The emergence of energy bottlenecks and infrastructure challenges]

The expansion of data centers is exposed the vulnerability of the power supply system. Elon Musk once pointed out that the production capacity of key electrical equipment such as transformers cannot meet the current demand for AI, and this shortage of power infrastructure will further amplify the load fluctuation of the power grid, especially during the peak period of AI training. Power demand may instantly exceed the average load by several times, and peak and valley power consumption patterns pose a huge threat to the stability of the energy system. This bottleneck was not obvious in the early stages of AI development, but will become more obvious as the cluster scale expands and AI applications increase in volume. This dilemma can be seen in the implementation process of Sora.

[Energy technology innovation and computing power ecological synergy]

With the rapid growth of computing power demand In the context of the epidemic, energy bottlenecks are becoming the core obstacles limiting the development of AI. Nuclear energy, especially small modular reactors (SMRs), has gradually emerged as one of the best solutions for AIDC. Emerging nuclear energy companies represented by OKLO\Nuscale are developing microreactor technology, and cloud services such as Google and MicrosoftThe provider has launched the SMR project layout, with the goal of powering future data centers through distributed small nuclear power plants to provide continuous and stable computing power support. Solutions such as natural gas + fuel cells / clean energy / energy storage are also being actively promoted as one of the options for rapid implementation. Start-ups represented by Bloom Energy are also rapidly rising with the help of industry trends.

From an investment perspective, the market has already recognized the importance of computing power, and is eagerly looking forward to the implementation of applications, constantly looking for mapping, and ignoring The importance of AI infrastructure is not just an opportunity for liquid cooling and computer rooms. From a larger perspective, the next stage of competition is gradually gaining momentum in various energy (natural gas, nuclear power, etc.) fields.

1. “Electricity Tiger” AIDC and Weak Grid 1.1 Electricity Consumption: AIDC’s Next Short Board

1.1.1 Supply and Demand of Electricity in the United States

< p style="text-align: left;">Demand side: Data centers are already "big consumers of electricity," accounting for 4% of the nation's electricity consumption. The total power of U.S. data centers in 2023 is about 19GW. Based on this estimate, the annual electricity consumption is about 166TWh (terawatt hours), accounting for 4% of the national electricity consumption.

The data center consumes 166 TWh of electricity, which is more than the annual electricity consumption of New York City and equivalent to the annual electricity consumption of 15.38 million household users. In terms of regions, New York's annual electricity consumption in 2022 was 143.2TWh, Texas' annual electricity consumption was 475.4TWh, California's 251.9TWh, Florida's 248.8TWh, and Washington's 90.9TWh. The U.S. data center's annual electricity consumption exceeded New York City's annual electricity consumption. power consumption. The average annual electricity consumption per residential user in 2022 is 10,791kWh. Based on this estimate, 166TWh is equivalent to the annual electricity consumption of approximately 15.38 million household users.

*1 TWh = 1000 GWh = 10^6 MWh = 10^9KWh

p>

Supply side: The annual power generation in the United States is relatively fixed, and thermal power is still the main source. New energy power generation is growing rapidly, and the proportion of nuclear energy has further increased. The annual power generation capacity in the United States is approximately 4,000-4,300 terawatt hours (TWh), of which thermal power (coal, natural gas, fossilOil) accounts for about 60% and is the main energy source; new energy power generation (wind energy, solar energy, etc.) has grown rapidly in recent years and accounts for 21%; nuclear energy accounts for about 19%, and its proportion has further increased.

Electricity prices: The United States has one of the lowest electricity prices in the world, and some states have energy advantages. There are lower electricity prices. The U.S. electricity consumption structure is mainly divided into four areas: residential, commercial, industrial and transportation. In September 2024, the electricity price for residential users is US$0.17/kWh (approximately 1.24 yuan/kWh, the exchange rate is as of December 13), and the electricity price for commercial users is US$0.135/kWh (approximately 0.98 yuan/kWh); industrial The electricity price is US$0.09/kWh, and the transportation electricity price is US$0.13/kWh. kWh, wholesale electricity prices in 2023 are $0.036/kWh. Some states have lower electricity prices due to their energy advantages. As of April 2024, the electricity price in Texas (rich in natural gas and renewable energy) is approximately US$0.147/kWh, and in Louisiana (rich in energy resources), the electricity price is approximately US$0.147/kWh. It is US$0.115/10 million hours, and Tennessee (rich in hydropower resources) is about US$0.125/kWh. Some large-scale power-consuming infrastructure, such as data centers, are often built in provinces with low electricity prices. The above-mentioned state capitals have also become the concentration of today's computing power industry.

Estimation of annual electricity cost for data centers: Based on the wholesale price of US$0.036/kWh, the U.S. data center (when AI has not yet been applied on a large scale) for one year It consumes 166TWh of electricity and is estimated to require approximately US$6 billion.

1.1.2 Marginal changes: AI’s challenge to the power grid

[Challenge 1: Total electricity consumption increases significantly]

Compared with traditional data centers, AI data centers It consumes a lot of power. The main reasons are the massive growth in data volumes, complex algorithms, and the need for instant response 24/7. For example, a Google traditional search request consumes about 0.3Wh, while a ChatGPT request consumes 2.9Wh, which is ten times the former; a paper published in "Joule" stated that if Google uses AIGC for every search, its usage Electricity will rise to29 billion KWh, which will exceed the total electricity consumption of Kenya, Croatia and many other countries; according to the New Yorker Magazine, ChatGPT consumes more than 500,000 KWh every day.

[Challenge 2: Using electricity to intensify the voltage]

Phenomena: The current demand of AI data centers (whether training or inference) is highly transient, with huge swings occurring within a few seconds. As the task load of the neural network model increases or decreases, the current demand will fluctuate wildly, even up to 2000A per microsecond.

Principle: 1) Peak load fluctuation: The training and inference of AI models require huge computing power, but they do not run continuously. Peak loads will occur when model training starts. , while basic operation is maintained during low periods, causing power consumption to fluctuate; 2) Dynamic resource scheduling: AI tasks are cyclical. For example, large-scale training requires centralized resources, while the inference phase is relatively dispersed, which makes the power consumption curve more unstable; 3) Respond to needs in real time: Generative AI and large model applications require low latency and high throughput, driving real-time expansion of infrastructure and further amplifying power consumption fluctuations.

Result: Affects the stability of the power grid. The design of the power grid is not suitable for excessive swing voltage. The power grid is basically designed for the power load. We hope to see a relatively stable, regular and slowly changing load. For example, an electrical device with a power load of 100GW may change after being connected to the power grid. There are two 200GW transmission lines for power supply, and operation can be guaranteed if one of the two transmission lines is normal. The AI ​​power consumption characteristics will have huge swings within a few seconds, and this violent fluctuation may affect the stability of the power grid.

[Challenge 3: The subsequent demand for electricity will be greater]

Inference in the AI ​​data center consumes more energy than training due to the large number of requests from users. Currently, Google has announced in the first half of this year that it will add new AI features to improve the search experience and will launch Gemini-based AI Overviews, which is already available for trial for some users; Microsoft has launched a personal AI assistant called Microsoft Copilot and has already ChatGPT is integrated into Bing. At present, the number of visits to Google's search engine has reached 82 billion times per month, and the number of Office business products has reached 82 billion.The number of paying users has exceeded 400 million. The huge user base means that if the trained large model is integrated into the company's products, the number of user requests will increase significantly, and the number of AI instant responses will surge, causing the model inference energy consumption to exceed the training energy consumption. U.S. data center power loads could account for 30% to 40% of all new demand until 2030, according to McKinsey estimates.

Conclusion: With the rapid development of AI, AI software that integrates large language models is expected to It will develop rapidly, and training needs and inference needs will resonate. In the future, the power consumption of data centers will increase significantly. AIDC will become a new generation of "electric tigers", and the proportion of data center power consumption will further increase.

1.2 Realistic dilemma: The power grid is difficult to support

The economic development structure determines that the power grid infrastructure in North America is relatively weak. Over the past 20 years, the decoupling of U.S. electricity demand from economic growth has accelerated dramatically. Since 2010, the U.S. economy has grown by a cumulative 24%, while electricity demand has remained almost unchanged, and in 2023, U.S. electricity consumption even fell by 2% from 2022. Its essence is that it is different from the fact that the economy is mainly driven by industry and service industries. The economic growth of the United States does not mainly rely on electricity or energy consumption, but mainly relies on high-tech industries, with low energy consumption. And efficiency gains, primarily the replacement of incandescent lights with fluorescents and LEDs, have offset demand for electricity from population and economic growth, leaving utilities and regulators without expanding grids or generating capacity.

Current situation: lack of time, lack of people, lack of infrastructure, lack of experience, and many obstacles.

Lack of time: It takes about two years to build a data center, but construction of the power grid is much slower, and it may take three to five years to build a power station , and it will take 8 or even 10 years to build a long-distance, high-capacity transmission line. According to MISO, the U.S. regional transmission organization, the 18 new transmission projects it is planning could take seven to nine years, compared with 10 to 12 years for similar projects historically. It can be deduced from this that the construction speed of the power grid is likely to be unable to catch up with the growth rate of AI.

Lack of infrastructure: According to the power investment trend in the United States, capital expenditures of U.S. utilities will increase significantly from 2016 to 2023, especially for power generation, distribution and Transmission field, grid investmentThe acceleration started in 2018, mainly due to the reshoring of manufacturing industry to promote power demand. Against this background, the United States still has not expanded the power grid on a large scale. According to a survey report issued by Grid Strategy, from 2010 to 2014, the United States installed an average of 1,700 miles of power grids per year. of new high-voltage transmission miles, but fell to only 645 miles per year in 2015-2019.

Shortage of workers: A tight workforce is also a constraint, especially a shortage of professional electrical workers necessary to implement new grid projects. According to McKinsey estimates, the U.S. could see a shortage of 400,000 specialized workers based on projected construction of data centers and similar assets requiring similar skills.

Lack of experience: For the United States, practitioners in the entire power industry have not seen large-scale growth in power demand in the past 20 years, and these 20 years Years will likely mean an entire group of engineers and staff without experience in building a new power grid on a large scale.

Many resistances: The construction of the power grid requires infrastructure such as power stations and transmission lines, and these may require the joint efforts of countless stakeholders to reach a compromise on the route of the lines and the cost. .

Conclusion: Compared with the construction speed of data centers, the current speed of power grid construction in the United States is relatively slow. It is slow and has limited power generation capacity, so in the short term the United States will face a power demand dilemma under the development of AI. North American utility Dominion Energy, for example, said it may not be able to meet Virginia's power demand, delaying construction of one of the world's fastest-growing data center hubs for years. And in the power industry, new infrastructure planning takes five to 10 years, according to Wood Mackenzie. Additionally, most state public utility commissions have little regulatory experience in a growth environment. It can be inferred that electric energy may become one of the biggest constraints on the development of AI in the next few years. Although the market is paying attention to innovative solutions such as controllable nuclear fusion, water from afar cannot quench the thirst for nearness, and it is inevitable to form short-, medium-, and long-term comprehensive solutions.

1.3 Multi-angle calculation: How much power does AIDC consume?

* Total power (GWh) = Total power (GW) × Time (h)

* Total power (GW)=IT equipment power (GW)×PUE (energy efficiency ratio)

1.3.1 Calculation angle one (conservative): AI chips

Calculation logic: Calculation angle one is from the number of chips From this perspective, extrapolate to 2030, and then use the number of chips * chip power consumption to predict the total power consumption, without considering that the overall power consumption of the server will be greater than that of a single chip * The quantity does not take into account the possible increase in single-chip power consumption after future chip upgrade iterations. Therefore, we believe that the calculation angle 1 is a "conservative" calculation. The calculation data is the smaller of several methods. The AIDC power demand in 2030 is 57GW.

Number of GPUs and TPUs in use: According to DCD reports, the total shipments of GPUs in the three enterprise data centers of Nvidia, AMD and Intel in 2023 are estimated to be 3.85 million units , the number of TPUs produced for Google in 2023 is expected to be 930,000. Further tracing the supply chain, TSMC predicts that the year-on-year growth rate of demand for AI server manufacturing from 2024 to 2029 will be approximately 50%. Based on this calculation, GPU shipments in 2030 will be approximately 65.78 million, and TPU shipments will be approximately 15.89 million. According to NVIDIA's official statement, the average service life of most H100 and A100 is 5 years. Therefore, we assume that the number of chips in use in 2030 is the sum of chip shipments in 26-30 years. Therefore, the number of GPUs and TPUs in use in 2030 is approximately were 171.36 million and 41.39 million.

GPU, TPU power consumption: The maximum power of H100 NVL can reach 800W. Then there are expected to be 171.36 million GPUs in 2030. Assume that GPU and TPU energy consumption account for 90% of the total energy consumption of IT equipment. Assume that the United States accounts for 34%, the utilization rate is 80%, and the PUE is 1.3. Calculate, in 2030, the US AIDC GPU power demand is approximately 54GW (Number of GPUs*GPU power consumption*U.S. share*PUE* Utilization rate ÷ chip ratio = 171.36 million * 0.8kW * 34% * 1.3 * 80% ÷ 90% = 54GW);

According to Google’s official statement , the average power of TPU v4 chips is 200W. Combined with the above estimate that the number of TPUs in use in 2030 is about 41.39 million, we predict that the number of TPUs in use in 2030 Total power consumption is approximately 3.3GW (other metrics are assumed to be the same as for the GPU).

Angle 1 Conclusion: The total AIDC electricity consumption in the United States in 2030 is 57GW. The chip inventory in 23-26 years only takes into account the chip shipments after 23 years. Other calculation methods are the same as the above methods. , the calculation method is the same as above from 27 to 30 years. Finally, adding up the power consumption of GPU and TPU, we can get that the power capacity required by the US AIDC will reach 24 to 30 years respectively. 3/6/10/17/25/38/57GW.

Hypothesis 1: The chip growth rate is 50% per year (refer to TSMC's statement)

Assumption 2: Assume that the average chip life is 5 years (refer to the GPU life given by NVIDIA).

Assumption 3: The average power utilization rate of IT equipment is 90% (considering the power consumption of NVSwitches, NVLink, NIC, retimers, network transceivers, etc. in IT equipment, assuming that GPU and TPU energy consumption account for 90%, and others IT equipment energy consumption accounts for 10%).

Assumption 4: Considering that IT cannot be operated at full capacity and cannot be operated 24 hours a day, refer to Semi analysis, set the possible utilization rate to 80%

Assumption 5: PUE is 1.3 (PUE is the total power consumption of the data center divided by the power used by IT equipment). .

Hypothesis 6: The United States’ computing power demand accounts for 34% of the world’s computing power demand (according to the calculation of the Institute of Information and Communications Technology, the United States’ share of global computing power is 34%. ).

1.3.2 Calculation angle two (optimistic): data center

Calculation logic: The second calculation angle is from the data center From a construction perspective, refer to the global data center construction progress predicted by a third party (compound growth rate of 25%). At the same time, since the forecast data ends in 2026, we assume that it will still maintain 25% from 2027 to 2030. The compound growth rate of global data center power demand is predicted, and the power consumption and proportion of AIDC are assumed. Therefore, we believe that the data obtained from this forecast perspective is relatively "optimistic", and the final forecast is that by 2030, the United States willAIDC power demand tops out at 91GW.

Research firm SemiAnalysis used analysis and construction forecasts of more than 5,000 data centers and combined these data with global data and satellite image analysis to predict the next few years. Annual data center power capacity growth will accelerate to a compound annual growth rate of 25%, and the proportion of AIDC will further increase. In terms of data centers, according to forecast data, global data center critical IT power demand will surge from 49GW in 2023 to 26 years. of 96GW. We assume that data centers will continue to maintain a compound growth rate of 25% from 27 to 30 (refer to the growth rate from 2023 to 2026, which is 25%). Then, by 29 and 30, the key IT power demand of global data centers will increase to 188 and 188 GW respectively. 234GW; referring to Semi Analysis data, combined with the booming development of AI computing power and the explosion of downstream applications, we believe that the future of AI The proportion in data centers is expected to continue to accelerate, so we assume that the global AIDC proportion will reach 12%/16%/30%/44%/56%/68%/78%/88% respectively in 23-30, so as to calculate In 2029 and 2030, the global power demand for AIDC IT equipment was 65GW and 91GW respectively.

Conclusion from perspective two: Calculated based on the US share of 34% and PUE of 1.3, US AIDC power demand will reach 91GW by 2030.

Hypothesis 1: Combined with the background of the booming development of AI computing power and the explosion of downstream applications, we believe that the proportion of AI in data centers is expected to continue to accelerate in the future. Therefore, we assume that the proportion of global AIDC in 23-30 will reach 12%/16%/30%/44%/56%/68%/78%/88%.

Assumption 2: PUE is 1.3 (PUE is the total power consumption of the data center divided by the power used by IT equipment).

Hypothesis 3: The United States’ computing power demand accounts for 34% of the world’s computing power demand (as measured by the Institute of Information and Communications Technology, the United States’ share of global computing power is 34%) .

1.3.3 Summary 1: AIDC’s proportion of total U.S. electricity consumption has increased

(1) AI’s proportion of U.S. electricity consumption has increased. The proportion is expected to exceed 10%

According to Statista forecast data, in 2022, the electricity consumption in the United States will be approximately 4,085 terawatt hours. Electricity usage will continue to rise to It will reach 4315 terawatt hours (corresponding to 493GW) in 2030, and will reach 5178 terawatt hours in 2050. According to our previous "Calculation Angle 1", if the total power consumption of AIDC in 2030 is the highest at 57GW, it will account for 50% of the US electricity consumption. The proportion will increase to 12% (57GW/493GW), a significant increase from 4% in 2023.

1.3.3 Summary 2: AIDC power consumption is expected to be comparable to Bitcoin mining< /p>

In our report "AI "The east wind has arrived, and Bitcoin mines have started their second growth curve", we have made assumptions and predictions about the electricity consumption of Bitcoin mines. In this report, we predict that Texas will have The loads of Bitcoin mines are 4.7/6.5/8.3/10.1/11.9GW respectively (assuming that the annual new load of Bitcoin mines in Texas is 1.8GW), regarding the share of Texas Bitcoin mine load in the United States, we assume that it remains unchanged at 28.5%. Therefore, according to our forecast, the annual load of Bitcoin mines in the United States is 17/23/29/36/42GW respectively.

For the convenience of comparison, we forecast the data to 2030, assuming: 1) The annual new load of the Texas Bitcoin mine is 1.8GW, 2) Assumption Texas mine share remains unchanged at 28.5% in 2029 and 2030. Therefore, it is concluded that in 2024/2025/2026/2027/2028/2029/2030, the annual power consumption of U.S. Bitcoin mining farms is 17GW/23GW/29GW/26GW/42GW/48GW/54GW respectively.

Conclusion: Under conservative forecasts, the power consumption of AIDC in the United States will surpass Bitcoin mining power demand in 2030; under optimistic forecasts, AIDC power demand in the United States will exceed Bitcoin mining in 2029.

< p style="text-align:center"> 2. What is the solution to the dilemma: "Natural gas +" is the mainstream in the short term 2.1 The fastest implementation plan in the short term is natural gas

2.1.1 Substations have become traditional power bottlenecks

[Current situation of data center power supply]

Purchasing electricity and Substation: Data centers typically purchase electricity through a contract with a power company, which means that the data center's electricity supply is generated from the power station and transported to the data center through the transmission network. However, after power is transported over long distances, the voltage often needs to be adjusted through substations to ensure that the power meets the voltage needs of the data center.

Necessity of substations: Substations convert high-voltage electricity into low-voltage suitable for local use. Most power systems require voltage conversion and distribution through substations. Without a local substation, power cannot be used directly in the data center.

The construction of substations is difficult, takes a long time, and costs high: The construction of substations usually requires a large amount of capital investment, involving land, infrastructure construction, and equipment. Procurement and manpower reserves, etc. In addition, substation construction takes a long time and needs to meet strict environmental and safety standards.

Conclusion: Under the current electricity purchase method, substations have become a bottleneck restricting AIDC's power consumption. As data center power demands continue to grow, building new substations or expanding existing substations takes a long time and requires significant approval and construction time, which may not keep up with data center demand quickly.

[Natural gas does not require a substation and is the first choice for distributed power supply]

Natural gas power generation does not rely on substations. Natural gas power generation generates electricity by burning natural gas. Natural gas power stations are usually connected to data centers through dedicated pipelines. The natural gas is directly transported to the power generation facility for combustion and power generation. The generated electricity is then supplied to the data center through the local power grid or dedicated lines. It can usually be Completed at a power generation facility near the data center, unlike traditional power transmission methods, natural gas power generation does not need to pass through the high-voltage power transmission network, and therefore does not rely on remote substations and power transmission facilities. Natural gas power generation can build small natural gas power stations (such as distributed power generation systems) near data centers, reducing dependence on external power grids and shortening the response time of power supply.

2.1.2 There is a time lag between the rapid development of AI and the implementation of SMR nuclear power

< p style="text-align: left;">Although nuclear power has advantages in many aspects, the most important demand in the North American computing power market is "rapid implementation". Quickly light up the GPU to obtain computing power, and natural gas has become the current first choice.

Although in February 2023, the U.S. Nuclear Regulatory Commission approved the design of the first SMR (Small Modular Reactors) by nuclear power company Nuscale Power, and China, Russia and other world Countries are racing to put SMR technology into practice, but the commercialization of SMR will still take some time, and the safety approval process is complex and time-consuming. It can already be seen that SMR has sparked global interest in nuclear energy. The U.S. nuclear fission industry has received a boost from the Inflation Reduction Act, which includes a number of tax credits and incentives while providing $700 million in funding for the Office of Nuclear Energy to support the development of high-purity low-enriched uranium (SMR). fuel) supply; there are more than 70 commercial SMR designs under development around the world, and there are currently two SMR projects operating in China and Russia. But according to the U.S. Energy Regulator, nuclear reactors are extremely complex systems that must meet strict safety requirements and account for a wide variety of accident scenarios, and the licensing process is cumbersome and varied. This means that SMRs require some standardization before they can enter the commercial market, so other solutions need to be found to address short-term energy shortages.

2.2 The combination of "natural gas + multiple energy sources" is more stable

The combination of natural gas + other multiple energy sources The solution is currently the fastest implementation solution that can meet the power needs of AI. Compared with SMR nuclear power, an independent solution with high energy density but long deployment period, natural gas power generation can be used as a basic energy source to quickly respond to load demand due to its high efficiency and flexibility. At the same time, it can be combined with renewable energy, fuel cells, and energy storage systems. Used in conjunction, they can effectively compensate for intermittency and lack of stability. This multi-energy combination can not only meet the needs of AI data centers for stable power supply, but alsoProviding a balance between carbon emissions and costs has become an important choice in current data center energy strategies.

Collaboration is not necessary, but for large-scale AI data centers that need to comprehensively balance stability, environmental protection and cost, collaborative use of multi-energy solutions is more flexible and long-term choice, when there are clear goals (such as low cost, ultra-fast deployment), a single solution can also satisfy:

[Use only natural gas to generate electricity (single solution) )】

Advantages: Natural gas power generation itself can be used as an independent power supply solution, suitable for scenarios that require stable power demand and rapid deployment, especially AI data centers that require high dispatchability;

Limitations: Although deployment is fast, carbon emissions are higher in the long run.

[The necessity of multi-energy synergy]

More stable and secure: AI data center for The power continuity requirements are extremely high (short-term power outages are not allowed), and natural gas + energy storage system or fuel cells can be used as backup support;

More environmentally friendly: natural gas + Combination of low-carbon energy sources such as wind energy and solar energy.

2.3 Natural gas solution: taking xAI as an example

Natural gas power generation technology has mature technology paths, complete supporting equipment, and high cost performance. In the short term, it is the fastest choice to solve the AI ​​power shortage problem. Tesla xAI uses natural gas solutions as emergency power supply. A natural gas generator is a generator that uses natural gas instead of gasoline or diesel. Compared with diesel, the purchase cost of natural gas is lower and there is no "wet accumulation" problem. Therefore, from the perspective of short-term energy solutions, natural gas generators have the advantages of cost-effectiveness, high operating efficiency, and more environmental protection than other generators using fossil fuels such as oil. According to DCD reports, Tesla CEO Musk has purchased 14 mobile natural gas generators from Voltagrid, each of which can provide 2.5 MW of power, to alleviate the power shortage problem in the data center of his startup xAI.

* Additional details 1: Musk xAI mainly uses NVIDIA H-series servers, and the cluster heat dissipation uses liquid cooling solutions. Each liquid-cooled rack in the xAI data center contains 8 NVIDIA H100 GPU servers, with a total of 64 GPUs. The dense layout requires each computing node to dissipate heat efficiently. The traditional air cooling method is difficult to adapt to, so xAI chose AMD's Liquid cooling solution.

* Additional details 2: The xAI data center also uses the Megapack energy storage system. xAI said that when building the computing cluster, its team found that the AI ​​server did not run at 100% power all day long, but had many peaks and valleys in power consumption, so Tesla's battery storage product Megapack was added in the middle to buffer the fluctuations. Thereby improving the reliability of the overall system and reducing power loss.

2.4 Fuel cells: Take Bloom Energy as an example

Company profile: Bloom Energy focuses on development Efficient, low-emission energy technology, committed to promoting global energy transformation through innovative solid oxide fuel cell (SOFC) and solid oxide electrolyzer (SOEC) technologies. As a leading clean energy company, the company is committed to providing sustainable, reliable energy solutions for high-demand areas such as industry, commerce and data centers through its advanced hydrogen and fuel cell technologies. Founded in 2001, the company is headquartered in California, USA, and has expanded operations globally.

Core technology: The company's core technologies include solid oxide fuel cells (SOFC) and solid oxide electrolyzers (SOEC). The SOFC system uses 100% hydrogen Provides efficient power output with electrical efficiency as high as 65%, far exceeding traditional energy systems. Bloom Energy’s fuel cell systems can also integrate combined heat and power (CHP) technology, resulting in total energy efficiency of up to 90%, effectively reducing energy consumption and carbon emissions. In addition, SOEC technology can be used for efficient hydrogen production and is one of the key technologies in the clean energy transition.

Product application: The company's products are widely used in many fields, including industrial power supply, Commercial energy management and data center energy solutions. Particularly in the data center space, as demand for energy efficiency and carbon neutrality goals continues to increase, the high efficiency and low emissions characteristics of Bloom Energy's fuel cell technology are even more prominent.Its hydrogen solutions can not only meet large-scale energy needs, but also provide enterprises with reliable backup power to ensure the continuity and stability of operations. Bloom Energy's market currently covers many regions such as North America, Asia and Europe, especially in South Korea's cooperation with SK Ecoplant. Bloom Energy's hydrogen fuel cell project is expected to come online in 2025. Additionally, the company has announced a gigawatt fuel cell purchase agreement with AEP to power AI data centers.

3. Medium-term plan: SMR nuclear power stands out 3.1 Why nuclear power: more suitable for AI

3.1.1 Characteristics of AIDC: Distributed and High Density

Compared with the traditional IDC data center, there are two most significant differences between the AIDC computing power center and the AIDC computing power center. important characteristics.

[AIDC Feature 1: Distributed Deployment]

AI application scenarios and task requirements etc. determined that AIDC needs to adopt a distributed deployment method. There are significant differences between AIDC and traditional IDC in terms of computing requirements, application scenarios, resource consumption, etc. AIDC tasks are usually computationally intensive, especially large-scale deep learning, machine learning, data analysis and other tasks in the AI ​​field. A single calculation Nodes cannot carry all tasks. Therefore, AIDC needs to split computing tasks into multiple small tasks and distribute the tasks to multiple nodes for parallel computing through a distributed computing framework. This requires data centers or computing nodes in multiple geographical locations. Work collaboratively.

[AIDC Feature 2: 24-hour high-density computing]

Continuity of AI computing tasks The performance and high load determine that AIDC must operate at high load 24 hours a day, which requires higher power resources and cooling support. AI model training is often a long-term process that requires continuous computing power support, so AIDC usually performs long-term and continuous computing tasks; the load of traditional IDC generally fluctuates according to business needs, and many applications do not require such long-term , uninterrupted computing support. Therefore, AIDC's high-power computing hardware requires round-the-clock strong power supply and cooling support.

AIDC's distributed deployment + high-density computing characteristics determine that other energy sources are difficult to meet the adaptation needs, and small nuclear power SMR best meets the power supply needs.

Thrust - other energy sources are not suitable for AI needs, and stability and geographical location are difficult to meet AIDC:

Hydropower has obvious seasonality, which makes it difficult to meet stable and large power supply needs. At the same time, the geographical location with rich water resources is fixed, and it is difficult to meet the distributed deployment needs of AIDC. At the same time, hydropower requires a distribution network to transmit power. , the overall cost is higher, and the new construction cost and time are higher;

Thermal power has high fuel costs and strict carbon emission restrictions. Even if you purchase carbon emission indicators, the overall cost of thermal power will be higher. Therefore, it is not suitable for AIDC that requires a lot of power. At the same time, thermal power also faces power distribution The problem of higher costs caused by the network;

Although other new energy sources (such as solar energy, wind energy, etc.) are clean, their power generation capacity is affected by weather conditions and geographical restrictions. Large, intermittent and unstable make them unable to ensure AIDC under high loads Stable operation around the clock. In addition, the conversion efficiency of some new energy sources such as photovoltaics is still low, and the later operation and maintenance costs are high. From a cost-effective perspective, it is not suitable for AIDC.

Attractiveness - SMR nuclear power has stronger comparative advantages, its modular design is suitable for distributed deployment, and it also adapts to the environmental protection requirements of carbon emission reduction. The modular characteristics of SMR technology enable it to be flexibly applied in distributed deployment scenarios. Modules can be flexibly added or reduced according to the needs of different regions, ensuring that the power supply of AIDC distributed data centers is not affected by geographical location, weather, and energy price fluctuations. Moreover, as a clean energy source, nuclear power is in line with the global trend of reducing carbon emissions and is suitable for AIDC's demand for green energy. Therefore, nuclear power SMR is suitable as the main power supply source for AIDC.

3.1.2 Nuclear power SMR is the fastest to be implemented

What is an SMR - a modular, smaller, more deployable nuclear reactor. SMR (Small Modular Reactor) is a new development of nuclear energy technology. SMR is a type of nuclear power plant.type, but are significantly different from traditional nuclear power plants. An SMR is a small, modular nuclear reactor that is designed to provide smaller-scale power output and is built using modular components to facilitate factory production and transportation. Generally, the output power of an SMR is lower than that of a traditional large nuclear reactor. smaller. Before the advent of AIDC SMRs were commonly used in remote areas off the grid, on small islands, on military bases, or as a supplemental source of power for industrial use.

Compared with traditional nuclear power plants, SMR has the advantages of small scale, short construction time, (construction and maintenance), lower cost, higher safety, cleaner and greener, longer life, etc.:

Small module output power: SMR The output power is smaller than that of traditional nuclear power plants, usually between tens to hundreds of megawatts, while the scale of traditional nuclear power plants is usually more than 1,000 megawatts. For example, NuScale's SMR module can provide 77MW of power individually, and can be assembled into up to 12 modules. Can provide 924 MW of electricity;

Shorter construction time: Because SMR adopts a modular design, it allows factory prefabrication and rapid assembly, such as NuScale's SMR nuclear power plant takes only 36 months (3 years), compared to the typically longer construction period for conventional nuclear power plants, which can take more than five to ten years.

Small footprint: Traditional nuclear power plants occupy a large area, usually larger than 1 square mile (approximately 2.6 square kilometers), while modular SMRs occupy Areas are generally smaller, with NuScale predicting that the SMR plant would cover an area of ​​0.06 square miles, close to the size of a small park.

Lower costs: The construction costs of traditional nuclear power plants are usually higher and are affected by economies of scale, but the construction costs of SMRs are relatively low, partly due to the use of standardization , Modular design enables each module to be mass-produced and reduces the construction and maintenance costs of a single reactor.

Higher safety: SMR designs often have higher passive safety features and disaster resistance, and can automatically shut down the reactor in the event of a fault without human intervention. Moreover, SMR reactors are smaller and therefore have higher safety and reliability.

Cleaner: SMR uses advanced reactor design,It can use fuel more efficiently and reduce the generation of nuclear waste, which is more in line with the requirements of clean energy;

Longer life: SMR is designed to have a service life of several There is no need to change the fuel for ten years, and the lifespan is far longer than that of traditional power generation modes. For example, Nuscale's SMR has a design life of up to 60 years.

The principle of SMR - basically the same as that of large nuclear reactors, it is still produced through nuclear fission reactions The thermal energy forms steam, which drives a generator to generate electricity. (1) Nuclear fission reaction: Like traditional nuclear power plants, the core of SMR is a nuclear reactor, which generates heat through nuclear fission reaction. Uranium-235 in the reactor When fissile materials (such as uranium or plutonium) absorb neutrons and undergo fission, the fission process will release a large amount of heat energy and neutrons; (2) Heat exchange and steam generation: The heat generated by the fission reaction in the reactor can be used for heating Coolant, the coolant flows inside the nuclear reactor, taking away heat and transferring it to the steam generator The generator or directly transfers heat to water through a heat exchanger to form steam; (3) Steam-driven generator: The generated steam is introduced into the turbine, and the generator is driven by the rotation of the turbine. The generator then converts mechanical energy into electrical energy to supply Grid or user; (4) Cooling system and safety mechanism: SMR Natural circulation cooling systems or passive safety systems are often used, using natural physical processes (such as thermal convection) to keep the reactor cool, thereby reducing reliance on external power and equipment. These systems can automatically shut down the reactor and cool down in the event of a failure.

The composition of SMR - usually includes multiple modules, using standardized components, which can Quickly assemble and deploy. (1) Reactor core: contains nuclear fuel, nuclear fission occurs, and generates a large amount of heat energy; (2) Cooling system: heat is taken away from the reactor core by circulating coolant, which can be liquid metal (such as sodium), gas (such as carbon dioxide) or helium), or water, some SMR The design adopts natural convection or passive safety system, which does not rely on external power to maintain cooling, which enhances the safety of the system; (3) Steam generator: transfers the heat-exchanged coolant to water to generate steam, which is introduced into the turbine to drive Generate electricity; (4) Turbines and generators: convert mechanical energy into electrical energy; (5) Control system: SMR uses a digital control system, and some also introduce AI Technology; (6) Safety system: Use a passive safety system, that is, the system can automatically cool the reactor without external power supply or operator intervention. Common designs include natural convection cooling, thermal storage devices, etc. These designs can emergencyWhen operating, the safety of the reactor is maintained through physical principles (such as thermal convection or gravity); (7) Nuclear waste treatment system: stores or processes nuclear waste and radioactive materials.

At present, there are several different technical routes for small modular reactor SMR, the most mainstream The most popular one is the light water reactor (LWR-SMR) because the technology foundation is mature and it is easy to obtain regulatory approval. As of 2021, countries around the world have proposed more than 70 different SMR nuclear power solutions, including pressurized water reactor solutions, helium gas-cooled reactor solutions (HTGR), high-temperature gas-cooled practical reactor solutions, and sodium-cooled fast neutron reactor solutions (SFR). About half of the plans are light water reactor reactions, which evolved from second-generation nuclear power technology. The technology is highly adaptable and can be commercialized quickly. However, due to the Fukushima Nuclear Power Plant issue in 2011, the technology tree selection for nuclear power has become more complicated, and safety concerns about light water reactors have become more prominent. Safer non-light water reactor solutions have become more popular, and high-temperature gas-cooled reactor solutions have gradually become popular. ：

Light water reactor (LWR-SMR): based on mature light water cooling technology, such as NuScale's design, the most mainstream and close to commercialization;

High-temperature gas-cooled reactor (HTGR): cooled by inert gas (such as helium), suitable for high-temperature process heat requirements, such as Huaneng high-temperature gas-cooled reactor;

Liquid metal-cooled reactor (such as sodium-cooled reactor): such as the Natrium reactor developed by TerraPower, which has efficient heat dissipation capabilities;

Molten Salt Reactor (MSR): uses high-temperature lava as cooling Fast Neutron Reactor (FNR): uses fast neutrons for high-efficiency fission fuel, such as the Russian BREST reactor type.

3.3 SMR Nuclear Power Current Situation and Industrial Chain

3.3.1 Cloud giants are vigorously deploying nuclear power

There is a shortage of power, and various cloud giants are deploying SMR nuclear power. On the one hand, the data center’s demand for power The demand is huge, SMR Providing long-term stable clean energy can reduce dependence on traditional power grids. On the other hand, in the long run, SMR can reduce the risk of electricity price fluctuations, optimize long-term operating costs, andAnd help the company achieve its carbon neutrality commitment:

Amazon: As early as March this year, it began to look for nuclear power support solutions and acquired Susquehanna, Pennsylvania, for $650 million. The Talen Energy data center park next to the Steam Electric Station nuclear power plant; and in October this year announced three major nuclear power investment agreements, cooperating with Energy Northwest and Dominion Energy to build 960MW and 300MW in Washington and Virginia respectively. SMR; led the US$50 billion C-1 round of financing obtained by nuclear energy startup X-energy;

Microsoft: Support for nuclear power is also significant, Bill Gates said in June this year that he would continue to invest billions of dollars in the "next generation" nuclear power plant in Wyoming, the United States, through the start-up company he founded, TerraPower LLC. The first commercial reactor is expected to be completed in 2030; in September In March, it reached a strategic agreement with Constellation Energy to restart the Three Mile Island nuclear power plant, which will provide approximately 835 megawatts of power for Microsoft's data centers.

Google: In October it said it had agreed to buy nuclear power from a small modular reactor being developed by a startup called Kairos Power to develop more than 500MW of electricity and The first reactor is expected to be operational in 2030;

Oracle: Founder Larry Ellison said in September that Oracle planned to build a three-SMR reactor Supported 1GW Data center campus;

Meta: is actively soliciting proposals from nuclear power developers, aiming to promote the development of its artificial intelligence technology and achieve environmental protection by increasing nuclear power generation capacity Target, plans to add 1 to 4 gigawatts of US nuclear power generation capacity by the early 2030s.

The huge power gap caused by AI data centers and the urgent power requirements faced by CSPs have made the trend of SMR nuclear power industry more and more obvious. It is expected that more SMR layouts will be announced in the future. .

3.3.2 SMR Nuclear Power Upstream and Downstream

The SMR nuclear power industry chain covers all aspects from upstream fuel uranium mines, midstream R&D and construction, downstream operations and waste processing. Relatively speaking, upstream design and manufacturing have higher thresholds for professionalism and technical barriers, so upstream manufacturers have higher bargaining power. Due to the long and stable operation cycle of the downstream operation and maintenance links, they can bring long-term cash flow and are also relatively profitable. The profit margin of midstream project construction is subject to factors such as construction cost, project cycle and engineering risks, and the profit margin is relatively less stable than that of upstream or downstream.

[Upstream: raw materials and processing]

The upstream industrial chain mainly involves the requirements for nuclear energy development The supply of basic raw materials, key equipment and nuclear fuel mainly includes uranium mining and uranium enrichment.

(1) Uranium mining and uranium processing

Uranium mining: uranium around the world The supply market is highly concentrated. U.S. uranium mines mainly rely on imports. Global uranium mines are mainly dominated by Kazakhstan, Canada and Australia.

Main uranium mining and typical companies that mine locally: Kazatomprom in Kazakhstan, Cameco in Canada and Orano (formerly Areva, a French company but mining uranium globally ) and Denison Mines, Australia's BHP (BHP Billiton) and Rio Tinto (Rio Tinto Group), Russia's Rosatom, etc. In addition, there are also some uranium mining companies in the United States, such as Energy Fuels (NYSE: UUUU), Uranium Energy (NYSE: UEC), etc.

Uranium processing: Uranium enrichment technology has very high safety, cost and technical requirements. Therefore it is mainly dominated by a few multinational companies. Natural uranium is mainly composed of uranium-235 and uranium-238. When a neutron collides with uranium-235, it will release huge energy through a fission reaction. The fissionability of uranium-238 is smaller than that of uranium-235. Natural uranium only contains About 0.7% uranium-235, so isotope separation (uranium enrichment) is required to increase its content to 3% to 5% for use as fuel for light water reactors. Concentration methods include gas diffusion, laser concentration and centrifugation.

*Principle of centrifugation: The gaseous uranium compound uranium hexafluoride is fed into the rapidly rotating rotor of the centrifuge to separate U-235 and U-238. The heavier isotope U-238 is pushed outward, while the lighter isotope U -235 is concentrated in the center of the rotor. The gas with a higher concentration of U-235 is extracted and fed into another centrifuge, where the process is repeated several times to produce uranium with a higher concentration of U-235.

Main uranium enrichment companies: Centrus Energy (NYSE: LEU, United States, dominates the global market), Orano (France, both mining and processing), Rosatom (Russia) ), Urenco (Europe).

（ 2) Nuclear fuel assembly manufacturing

The fuel used in SMR reactors includes uranium fuel rods, fuel elements and control rods, etc. The components must meet specific standards to ensure the safety of the reactor and efficient operation.

Participants: Such as Westinghouse, Orano, etc., providing nuclear fuel components and technical support.

(3) Reactor component manufacturing

Reactor components are an important part of SMR, including Reactor pressure vessels, cooling systems, control systems, reactor cores and other related facilities require a high degree of radiation resistance, high temperature resistance and reliability. Due to the modular design of SMRs, reactor components are typically mass-manufactured in factories and transported to site for rapid assembly, reducing on-site construction time.

Participants: such as NuScale Power, Rolls-Royce, etc.

[Midstream: Design, R&D and Construction]

(1) SMR design and R&D

Design and R&D: The design company is responsible for the technical development and design standardization of SMR reactors. The R&D of SMR usually includes the structural design, cooling system design, and control system of the nuclear reactor. integration, etc., assumingDesign R&D companies work closely with departments and regulatory agencies to ensure that designs comply with nuclear safety standards.

Participants: SMR design and R&D companies such as NuScale Power, OKLO, TerraPower, Rolls-Royce, etc.; institutions such as the U.S. Department of Energy (DOE) provide financial support And supervise and verify the design of SMR.

(2) Reactor construction and installation

The modular design of the SMR allows most components Prefabricated in the factory and then transported to site for quick installation. The construction phase is simpler than that of traditional nuclear power plants because SMRs are smaller in scale and highly modular, and can be put into operation without large-scale construction. For example, the construction company is responsible for assembling the various modules of the SMR reactor into a complete For nuclear power plants, complete on-site installation and factory-prefabricated components will greatly shorten the on-site construction cycle.

Participants: Construction companies such as Bechtel, Fluor, etc., are responsible for the construction and assembly of SMR power plants.

[Downstream: operations, sales and waste processing]

(1) SMR nuclear power plant operation

The operator is responsible for the long-term management, maintenance, monitoring of the reactor operation of the power plant, and ensuring that the reactor in a safe state. The operation and management complexity of SMR nuclear power plants is lower than that of traditional nuclear power plants. In addition, operators are also responsible for regular maintenance of the SMR system, including fuel replacement, equipment inspections and technical upgrades.

Participants: such as American Electric Power Company (AEP), British Electric Power Company (EDF), Southern Company, Exelon Corporation, Duke Energy (NYSE: DUK), Entergy Corporation (NYSE: ETR), PSEG (Public Service Enterprise Group, NYSE: PEG), Dominion Energy, etc. Some operators may purchase SMR power stations and operate them; management and monitoring companies will provide intelligent monitoring, data analysis and system optimization.Serve.

(2) Electricity sales and power grid connection

The electricity produced by SMR power stations is sold to grid companies or industrial users through power purchase agreements (PPA). SMR is suitable for small grids and is particularly suitable for specific markets such as remote areas, remote cities or industrial projects.

* Power Purchase Agreement (PPA): Operators sign long-term contracts with power purchasers (such as power grid companies, large industrial users, etc.) to ensure stable cash flow and profit model.

Participants: Power purchasers such as local power grid companies, large industrial enterprises, institutions, etc.

(3) Waste and nuclear power decommissioning treatment

SMR reactors require waste management after the life cycle. The long-term storage and processing of nuclear waste is an important part of the nuclear power industry. Waste management companies are responsible for the safe handling, transportation and storage of waste to ensure compliance with nuclear safety. standard.

Players: Waste disposal companies such as Waste Control Specialists, which specialize in the disposal of nuclear waste.

4. Long-term outlook: Controllable nuclear fusion

Nuclear fusion is achieved through two light The process by which atomic nuclei combine to form a heavier nucleus and release large amounts of energy. Controlled nuclear fusion reactions release about 4 million times more energy than burning coal, oil or natural gas and four times more than nuclear fission, providing unlimited clean and affordable energy if the process could be replicated on an industrial scale. Currently, more than 50 countries are conducting nuclear fusion research. However, due to the strict conditions for nuclear fusion to occur, breakthroughs in new materials and new technologies are still needed to achieve controllable nuclear fusion. How long it will take to achieve controllable nuclear fusion will depend on the technology development progress of the industry. At the same time, it is necessary to develop the necessary infrastructure and formulate management requirements and standards for the technology. According to space reports, British company Tokamak Energy has heated hydrogen plasma to 27 million degrees Fahrenheit for the first time in a new reactor, which is hotter than the core of the sun. The company said that using nuclear polymerizationChanging the production of commercial electricity could be achieved by 2030.

5. Business models and participants in the energy war 5.1 SMR Nuclear Power US Stocks

5.1. 1 SMR (NuScale, R&D manufacturer)

Company profile: NuScale Power is the first SMR nuclear power manufacturer to go public. The company originated from the SMR research project jointly carried out by the Idaho Laboratory and Oregon State University in 2002, and received support from the U.S. Department of Energy (DOE). NuScale Power. LLC was established in 2007 and became the first to obtain NRC ( The SMR has been designed and approved by the U.S. Nuclear Regulatory Commission and will become the first SMR technology provider to go on the market in 2022.

Core products: The company's core product SMR power module. The NuScale Power Module is the smallest light water SMR, measuring 76 feet tall and 15 feet in diameter, and can generate 77 MW of power from a single module. The modules, including seals, are completely manufactured in the factory and transported to the factory site by truck, rail or barge, eliminating the need for on-site fabrication or construction, reducing the schedule and cost risks associated with on-site construction.

Competitive advantage: The company has its own nuclear power plant - VOYGR Plant Models. VOYGR Plant Models is a standardized nuclear power plant designed by NuScale for its small modular reactor SMR. It has flexible power output and higher operating efficiency and can meet power needs of different sizes. It is the first and only design approved by the U.S. Nuclear Regulatory Commission (NRC) Approved small modular reactor.

VOYGR Plant Models different parameter modules:

VOYGR-4: composed of 4 NuScale SMRs Composed of modules, it provides approximately 308 MW of power output, suitable for providing power to small and medium-sized communities and industrial applications;

VOYGR-6: Contains 6 modules, providing approximately 462 Megawatt power, suitable for applications with medium-sized power needs, such as small cities or larger industrial facilities;

VOYGR-12: Composed of 12 modules with a total power output of approximately 924 MW, this is NuScale’s largest capacity VOYGR layout and is suitable for meeting large-scale For urban and industrial applications where power is needed, and even as baseload power for grid-level grids, the VOYGR-12 can supply 154 MW of power for 12 years without the use of new fuel, even in the event of a catastrophic loss.

Business layout: The company provides downstream customers with services ranging from license application, construction and commissioning to operation and maintenance full service support. The services provided by the company can be divided into two categories: pre-commercial application (COD) and post-commercial application:

Pre-commercial application: startup and testing, ITAAC management (inspection, Testing, Analysis and Acceptance Criteria), COLA Management (Co-License Application for VOYGR™ Power Plant);

Post-Commercial Application: Design Engineering Management, O&M Engineering project management, requalification training and simulator support, procurement and spare parts management, nuclear fuel and fuel outages, system verification and validation.

Project progress: We have cooperated with many customers around the world on SMR nuclear power projects. To date, the company has partnered with RoPower Nuclear S.A. (Romania), KGHM Polska Miedź S.A. (Poland), Kozloduy Power Plant (Bulgaria), Standard Power (Ohio and Pennsylvania), Prodigy Marine Power Plant (Canada), Indonesia Power (Indonesia) ) and GS Energy (South Korea) have project cooperation.

"Soft power": Focus on scientific research and cultivating talents, and have opened E2 nuclear energy exploration center laboratories in many universities around the world. In addition, the company has also set up an E2 Center (Energy Exploration Center) to provide users with practical opportunities to apply nuclear science and engineering principles through simulated real nuclear power plant operating scenarios. E2 has center points in multiple universities and regions around the world. Such as Texas College Station, Bucharest Polytechnic University (Romania), Seoul National University in South Korea, Oregon State University, etc.

Financial analysis: The company's financial situation is currently in a volatile stage, with abundant cash flow and no debt, and excellent results in cost reduction and efficiency increase. The company's latest third quarter report shows that the third quarter of 2024:

Revenue: The company’s operating income was US$500,000, compared with US$7 million in the same period last year. The decrease in revenue was mainly due to the termination of the contract with CFPP (on November 8, 2023, UAMPS and NuScale announced that both parties have agreed to terminate the carbon-free power project CFPP);

Net profit: The company's net loss was US$45.5 million (of which US$7.2 million was the fair value of outstanding warrants) Value-related non-cash charges), the company's net loss in the same period last year was US$58.3 million, and the net loss further narrowed;

Expenses: Operating expenses were US$41.2 million, while The same period last year was 93.9 million US dollars, operating expenses decreased by 52.7 million US dollars year-on-year, and the company further improved its ability to reduce costs and increase efficiency;

Cash: As of the third quarter report of 2024, cash and cash equivalents and short-term investments of $160 million (of which $5.1 million is restricted) and no debt

Capital background:

Fluor Corporation: A world-renowned engineering and construction company, it is the major shareholder and owns a large number of shares. Its investment in NuScale began in 2011 , to help companies obtain support in technology research and development and commercialization;

U.S. Department of Energy (DOE): The United States has provided a large amount of funding for NuScale's research and development through the Department of Energy. Support (over $300 million), support SMR Technology development and deployment;

Japanese trading company JGC Group;

Public and private equity: In 2021, NuScale announced that it would go public through a merger with Spring Valley Acquisition Corp.. Through this merger with a SPAC (Special Purpose Acquisition Company), NuScale entered the public capital market and providedNuScale brings about $235 million in funding;

South Korean company Doosan Heavy Industries: the world's leading heavy industry company, not only participates in investment, but also plans to build reactors for NuScale Provide some parts and manufacturing support;

5.1.2 OKLO (R&D manufacturer)

Company Profile: The company is owned by Jacob DeWitte and Caroline Cochran (both founders have a background in nuclear energy engineering) was officially established in 2013, focusing on the development of small modular reactors (SMRs), headquartered in California. In 2014, OKLO entered the well-known start-up accelerator Y Combinator and received start-up funds, 24 In September, OKLO received site authorization for a mini-reactor in Idaho and plans to deploy it in 2027. The company's Aurora microreactors are metal-fueled (unlike other nuclear reactors that use uranium fuel) and currently provide 24/7 clean energy to data centers, factories, industrial sites, communities and defense installations.

Core product: The company's core product "Aurora Microreactor", with a single module power of 1.5 MW, the Aurora module is refueled every ten years (so major expected shutdowns time is maintenance of the power conversion system), Aurora power plants offer power from 15 MW to 50 MW The plant covers only a few acres, has low operating and maintenance costs, and can be located where customers need power, avoiding expensive and long power lines.

Competitive advantage (fuel is different from others):

Microreactors are more suitable for distributed needs : OKLO's Aurora microreactor is a medium-scale SMR. The power plant is usually around 50MW and has a competitive advantage in meeting medium-sized distributed, remote and independent power needs. In comparison, NuScale's power is larger and more adaptable to the grid. Level energy solutions;

Fuel and cooling technology are cleaner, more environmentally friendly, and cheaper: Oklo's Aurora reactor uses metal fuel instead of traditional light water reactor fuel, and its cooling system is also different from common water cooling. Use liquid sodium ascoolant. On the one hand, such a fuel and cooling design can improve the thermal conductivity and efficiency of the reactor, and on the other hand, it can reduce the output of nuclear waste, thereby reducing the cost and environmental impact of nuclear waste treatment. In contrast, systems like NuScale, which uses traditional light water as the cooling medium and uranium as the fuel, are more suitable for existing nuclear power plant technology and supply chains.

Financial analysis: The company continues to expand investment in preparation for initial commercialization. The company's current cash flow is relatively abundant. The company's latest third quarter report shows that the third quarter of 2024:< /p>

Expenses: Operating expenses were US$12.28 million, compared with US$4.66 million in the same period last year. The company continues to expand investment;

Net profit: company's net loss US$9.96 million, compared with US$8.67 million in the same period last year. The expansion of net profit loss was mainly caused by continued investment;

Adequate cash: As of the third quarter report of 2024: Cash and cash Total marketable securities were $290 million, including $91.8 million in cash and cash equivalents and $197 million in marketable securities.

Capital background:

Sam Altman (Founder of Open AI): Oklo’s main funder In 2014, Altman included Oklo into the Y Combinator incubator. In 2024, Altman further helped Oklo successfully list on the New York Stock Exchange through a merger with his special purpose acquisition company (SPAC) AltC Acquisition Corp., raising approximately $306 million in funding to support its nuclear energy projects. Commercialization and future development;

Y Combinator: OKLO is a start-up company incubated by Y Combinator. Early financing mainly comes from YC The incubation project received start-up capital support. After Oklo merged with AltC Acquisition Corp, Oklo went public with a pre-investment valuation of approximately US$850 million. Early backer Y Combinator may retain an indirect shareholding in Oklo, but has not yet announced a target. Additional investment in post-IPO stages;

DCVC(Data Collective): A well-known venture capital company that focuses on investments in technology and deep technology fields, providing financial support to OKLO to help its technology development and market expansion;

U.S. Department of Energy (DOE): DOE has provided funding for OKLO's research and development for the commercialization of advanced fuel cycles and manufacturing technologies. DOE's funded projects have played a key role in promoting the maturity and verification of OKLO's technology.

5.1.3 NNE (NANO, R&D and manufacturing + fuel processing)

Company profile: Nano Nuclear Energy's main business covers 4 SMR-related contents, including manufacturing, fuel, transportation and other links, aiming to create a diversified vertically integrated industrial chain. NNE is an American start-up company. Its founder and chairman, Jay Jiang Yu, was an analyst at the investment banking department of Deutsche Bank. James Walker, CEO and chief R&D nuclear physicist, was responsible for the project of the new Rolls-Royce nuclear chemical plant. Principal, the company is focused on developing small modular reactors and is committed to becoming a commercially focused, diversified and vertically integrated company spanning four business lines:

Micro nuclear reactor technology development: NANO Nuclear Its main products include solid core battery reactor "ZEUS" and low-pressure coolant reactor "ODIN";

Nuclear fuel manufacturing: Establishment of nuclear fuel subsidiary HALEU Energy Fuel Inc. ( HEF), providing HALEU nuclear fuel (an advanced nuclear fuel containing 5%-20% uranium 235), which can be used by itself or supplied externally;

Nuclear fuel transportation: Establish a transportation subsidiary, Advanced Fuel Transportation Inc. (AFT), to provide HALEU nuclear fuel for small modular reactors, microreactor companies, laboratories, the military, U.S. Department of Energy projects, etc.;

Nuclear energy industry consulting services;

Other subsidiaries: NNE also established a space business subsidiary NANO Nuclear Space Inc. ( NNS), explore the application of NNE micronuclear reactor technology inPotential commercial applications in space.

Core products (manufacturing side): NNE microreactors can provide 1-20MW of thermal energy, among which the Zeus nuclear microreactor has a completely sealed core and relies on high conductivity It uses a thermal moderator matrix to dissipate fission energy, and then uses proprietary technology to take away heat from the outside of the container, avoiding the possible risks caused by coolant loss. In addition, Zeus The reactor core and power conversion system are housed in a standard container for easy transport and are designed to operate for 10 years. The ODIN nuclear reactor is the second Advanced Nuclear Reactor (ANR) being developed by NNE. It uses a low-pressure coolant. The reactor will operate at a higher temperature than traditional water-cooled reactors, maximizing the natural convection of the coolant.

Core products (fuel side): subsidiary HALEU Energy focuses on its reactors and Other SMR and microreactor companies develop and manufacture high-purity low-enriched uranium HALEU and have been selected as official founding members of the U.S. Department of Energy’s new High-Purity Low-Enriched Uranium Alliance (HALEU Alliance established on December 7, 2022). HALEU is enriched uranium with a concentration of the fissile isotope U-235 ranging from 5% to 19.9% ​​of the fuel mass. Compared with traditional uranium fuel, HALEU has many advantages - the reactor does not need to be refueled frequently, reduces the amount of waste, can be used as the next generation fuel of existing reactors, and is more economical and safer. According to NNE data, nearly 600 metric tons of HALEU will be needed by 2030 to bring new reactors to market.

Company finance: The company is currently in the project development stage, with sufficient cash flow. In 2024 Second quarter of the year:

Expenses: Operating expenses were US$4.32 million, compared with US$2.7 million in the same period last year, a significant expansion of investment

Net profit: The company’s net loss is 467 million, compared with $2.7 million in the same period last year. The expansion of net profit losses was caused by continued investment

Cash: As of the end of the second quarter of 2024, cash and cash equivalents were US$13.79 million, mainly from May 2024 Listed financing.

Capital background:

Citizens Financial Group Inc: Citizens Financial Group Inc holds shares through open market investments There are shares in NNE.

BlackRock: One of the world's largest asset managers, BlackRock holds shares in NNE through public market investments.

5.1.4 LEU (Centrus Energy, fuel processing)

Company Profile: Centrus Energy Positioned as a nuclear fuel and service provider (located in the middle of the industrial chain), it focuses on providing high-purity low-enriched uranium and efficient nuclear fuel solutions for the global nuclear energy market. Headquartered in the United States, the company's business covers the research and development of uranium enrichment services and related technologies. It is particularly at the forefront of the market in the field of advanced fuels (such as HALEU) and supports the commercialization of small modular reactors (SMRs) and next-generation nuclear energy projects. In the field of traditional nuclear energy, Centrus designs, manufactures and successfully operates gas centrifugal enrichment technology - American centrifuges, which have passed the test of the US Department of Energy. The company is currently gradually expanding from traditional nuclear fuel business to more advanced fuel business.

Main business: The company's main business has three categories, 1) Nuclear fuel supply: providing low-enriched uranium (LEU) and highly-enriched uranium (HALEU) to serve the nuclear energy and new reactor markets; 2) Advanced manufacturing: utilizing high-precision engineering Technology manufactures complex components. Specific products include high-efficiency equipment for the nuclear fuel cycle, ultra-high-precision mechanical components, and complex modules for nuclear energy and safety systems, etc., providing support for the energy, defense, and aerospace industries; 3) National Defense: for the United States Provide nuclear fuel technology and related services to ensure the safety of nuclear energy infrastructure.

Company Finance: The company's revenue maintains a growth trend, mainly driven by the HALEU operation contract signed with the Department of Energy (DOE), but the gross profit increases with the sales volume of SWU The decrease occurred to a greater extent. The company's latest third quarter report shows that in the third quarter of 2024:

Revenue: The company achieved revenue of US$57.7 million in the third quarter, compared with US$51.3 million in the same period last year. Stable and rising, mainly due to the company and the Department of Energy (DOE) The signed HALEU operation contract is transitioning from the first phase to the second phase at the end of 2023 to bring revenue expansion;

Gross profit: The company's gross profit is as of $8.9 million and $11.3 million for the three months ended September 30, 2024, 2023, respectively, with the decrease in the three months ended September 30, 2024 primarily attributable to LEU The decrease in segment gross profit was primarily due to an increase in SWU unit costs due to a decrease in the number of SWUs sold;

Expenses: Operating expenses were $16.5 million, compared with $1,420 in the same period last year million US dollars, operating expenses increased by US$2.3 million year-on-year, mainly due to the increase in sales and administrative expenses;

Net profit: The company's net loss was US$5 million, a net loss of US$5 million in the same period last year income $8.2 million, primarily due to lower gross profit.

Capital background:

U.S. Department of Energy (DOE): Centrus Energy acquired U.S. energy and strong support from the Ministry, especially in the production of highly enriched low-enriched uranium (HALEU), to promote the development and deployment of advanced nuclear fuel.

BlackRock: One of the world's largest asset managers, BlackRock holds a stake in Centrus through public market investments.

TerraPower: The nuclear energy company founded by Bill Gates partners with Centrus to develop advanced nuclear fuel for small modular reactors (SMRs).

X-Energy: An American nuclear energy company that cooperates with Centrus to develop fuel technology for high-temperature gas-cooled reactors.

5.1.5 UUUU (Energy Fuels, raw material mining)

Company Profile: Energy Fuels It is a mining and energy company headquartered in the United States (located upstream in the industrial chain and directly reserves fuel resources). It was formally established in 2006 and listed on the New York Stock Exchange in 2013. It focuses on the production of natural uranium and thorium and is a key player in nuclear energy and advanced fuels (such as HALEU) technology, it is also involved in the separation and refining of rare earth elements (REE). The company has multiple production facilities in the United States and is an important player in the North American nuclear fuel supply chain.

Main business: The company's main business includes the mining and processing of natural uranium and thorium, providing key fuels for the nuclear energy industry, and refining rare earth oxides through its facilities to use For clean energy and high-tech applications such as wind power generation and electronic equipment. The company's business model revolves around the extraction, processing and sale of mineral resources and establishing long-term supply relationships with customers in the energy and technology sectors.

Company Finance: The company continues to promote the mining of various rare earth elements. Recently, output has declined due to transportation problems. Lower than expected. The company successfully completed the acquisition of Base Resources, which included the advanced, world-class Catulia titanium and zirconium project in Madagascar, ensuring the company's leading position in the titanium and zirconium mineral industry. The company's latest third quarter report shows that in the third quarter of 2024:

Revenue: The company achieved revenue of US$4.04 million in the third quarter, compared with US$11 million in the same period last year, mainly due to Ore shipments from the Pinyon Plain mine to the White Mesa Mill are experiencing delays, with the issue expected to be resolved in the fourth quarter of 2024;

Net profit: The company's net loss attributable to the company in the third quarter was US$12.06 million, compared with net income of US$10.56 million in the same period last year, mainly due to transaction and integration costs related to advancing the Donald project, costs related to the acquisition of BaseResources and recurring costs. Operating expenses;

Gross profit margin: The company's uranium mining business gross profit in the third quarter was US$2.15 million, with a gross profit margin of 54%;

Expenses: The company's operating expenses in the third quarter were US$14.11 million, compared with US$12.38 million in the same period last year. The increase was due to the increase in new project integration costs and M&A transaction costs.

Consortium background:

BlackRock: a world-renowned asset management company that holds shares of Energy Fuels Publicly traded shares;

Vanguard Group: Another large asset management company that invests in Energy Fuels through the public market;

State Street Corporation: As a large financial services institution, State Street holds Energy Partial stake in Fuels.

5.1.6 Others

The small and micro nuclear power industry chain is huge. In addition to focusing on SMR technology In addition to innovative companies, traditional nuclear power companies and power operators are also involved:

UEC (Uranium Energy): focuses on the exploration, development and production of uranium resources in North America companies mainly use in-situ leaching (In-Situ Recovery (ISR) technology, the mining method is lower cost and more environmentally friendly. The company currently operates two major ISR platforms: one is located in Texas, supported by the Hobson factory; the other is located in Wyoming, relying on Irigaray and Christensen Ranch (formerly Willow Creek Project), these platforms manage multiple uranium projects and have a high degree of production readiness. In addition, the company owns high-grade traditional uranium projects in Canada such as Henday Lake and Carswell.

CCJ (Cameco): Canadian uranium mining and supplier, focusing on upstream uranium mining and processing, is one of the largest uranium suppliers in the world , providing raw materials for the nuclear fuel market.

BWXT (BWX Technologies): Focuses on nuclear reactor component manufacturing and nuclear energy technology. The biggest difference from SMR/OKLO is that BWXT is a large equipment supplier and technology Service providers mainly provide nuclear reactor components, nuclear fuel, and defense-related nuclear technologies to the commercial and commercial sectors. Customers include the United States (such as providing nuclear reactors for Navy nuclear submarines).

DUK (Duke Energy), CEG (Constellation Energy Group), EXC (Exelon Corporation), ETR (Entergy Corporation): large integrated power companies in the United States, operating Traditional nuclear power plants and provide power services, the core business includes power generation, transmission andDistribution services, etc., DUK focuses on the southeastern region, with a relatively balanced power generation portfolio, including natural gas, coal and renewable energy; CEG focuses on clean energy and operates the largest carbon-free nuclear power plant group in the United States, focusing on carbon emission reduction; EXC focuses on Nuclear Power Generation is the largest nuclear power operator in the United States, covering multiple states; ETR serves the southern United States, focusing on nuclear power and natural gas power generation, focusing on high-reliability power supply.

5.2 Competitive landscape and advantages

From the perspective of industry competition landscape, since SMR nuclear power is still in the early stages of development, the competitive landscape has not yet stabilized, and the market share gap between each company is not large. , the main differences are reflected in technical paths, business models and market layout:

1) Differences in technical routes

Pressurized water reactor (PWR) dominance: At present, the pressurized water reactor technology promoted by NuScale Power and other companies occupies the mainstream market because of its high technological maturity, clear regulatory approval path, and easier to gain the trust of investors;

The rise of innovative technologies: such as X-energy’s High Temperature Gas-cooled Reactor (HTGR) and Terrestrial Energy The Molten Salt Reactor (MSR) represents the next generation of innovative technology in nuclear energy, providing higher efficiency and flexibility, but faces challenges such as long R&D cycles and complex supervision;

Key differences: The robustness of traditional technologies and the potential breakthroughs of advanced technologies form a differentiated competitive landscape.

2) Differences in business models

Modularity and scalability: NuScale Power et al. The company focuses on modular design, making the reactor easier to produce, transport and assemble, thereby reducing construction and operating costs;

Specific market positioning: Nano Nuclear Energy Targeting the small and efficient reactor market in remote areas and military bases to provide more flexible power solutions.

3) Regional differences in market layout

United States: Benefit from support (such as "Inflation Reduction Act") and technological advantages, US companies (NuScale, X-energy) have first-mover advantages in technological leadership and access to capital;

Russia: Rosatom has achieved commercialization with RITM-200 and is the leader in floating nuclear power plants and polar markets;

: The Shidao Bay high-temperature gas-cooled reactor was successfully connected to the grid, providing support for energy security and showing potential in the export market.

Looking to the future, technology maturity, cost competitiveness, support, and market positioning are several important factors that determine the success or failure of SMR participants:

1) Technology maturity and safety

The core of the nuclear power industry is the safety and maturity of technology. It is the primary threshold for entering the market. NuScale has obtained design certification from the U.S. Nuclear Regulatory Commission (NRC) and is the first SMR company in the world to obtain this certification; important development direction of nuclear fuel.

2) Cost advantage

SMR needs to prove its cost in the whole life cycle (construction, Operation, decommissioning) is superior to traditional nuclear power plants and other forms of energy. Modular design is the key to reducing costs. NuScale, OKLO, etc. all reduce per-stack costs through standardized manufacturing and mass production.

3) and financial support

Support and initial capital investment are the key factors that determine the success of SMR projects. an important factor in whether it can be implemented. The United States supports the renaissance of nuclear energy through various incentives, including direct grants and tax incentives. In addition, companies can also raise funds through international cooperation, such as NuScale’s agreements with Romania and Poland to promote global deployment.

4) Market positioning and application scenarios

Diversified applications are an important competitive advantage of SMR , covering power generation, industrial heat supply, seawater desalination, hydrogen production, etc. For example, NuScale’s target market focuses on public power supply and industrial power markets, while Nano Nuclear targets remote areas and special-purpose markets, emphasizing miniaturization, mobility and rapid deployment capabilities.

5) Internationalization and customer resources

International market competition will become the key to the future, and companies need to prove that their technologies can compete in different applicability under regulatory, geological, and economic conditions. At present, NuScale has signed supply agreements with multiple companies to seize the global market. With the support of the British local market, Rolls-Royce plans to expand to Europe.

Horizontal comparison of financial data shows that nuclear fuel mining companies located in the upper reaches of the industry chain, such as Energy Fuels and Centrus Energy, are progressing rapidly in commercialization, with annual revenue and profits in 2023. is positive, while micro-reactor manufacturing companies located in the middle reaches of the industry chain, such as NuScale Power, OKLO, and NANO Nuclear Energy, are still in the business model verification stage as a whole. High R&D and production expenses have resulted in negative net profits. Among them, NuScale Power Progress is rapid. Through SMR projects in cooperation with many countries, revenue of US$22.81 million will be achieved in 2023. Specifically, the competitive advantages of each company include:

NuScale Power: 1. The company has its own nuclear power plant-VOYGR Plant Models, which is the first and only A small modular reactor that has received design approval from the U.S. Nuclear Regulatory Commission (NRC); 2. The company is currently cooperating with multiple customers around the world on SMR nuclear power projects;

OKLO ：1. OKLO’s Aurora Microreactors are medium-scale SMRs and have competitive advantages in meeting medium-sized distributed, remote and independent power needs; 2. Oklo’s Aurora reactor uses metal fuel instead of traditional light water reactor fuel, which is cleaner, more environmentally friendly and cost-effective Lower;

NANO Nuclear Energy: Broad business coverage, the company’s main business covers 4 SMRs Relevant content covers manufacturing, fuel, transportation and other aspects, aiming to create a diversified vertically integrated industrial chain;

Energy Fuels: Large production capacity, the company is The United States has multiple production facilities and is an important player in the North American nuclear fuel supply chain. It also recently successfully completed the acquisition of Base Resources, including the advanced and world-class Catulia titanium and zirconium project in Madagascar.The purpose is to ensure the company's leading position in the titanium and zirconium mineral industry;

Centrus Energy: 1. Focus on providing high-purity low-enriched uranium and high-efficiency uranium to the global nuclear energy market Nuclear fuel solutions, especially in the field of advanced fuels (such as HALEU), are at the forefront of the market; 2. The company has received strong support from the U.S. Department of Energy, especially in the production of highly enriched low-enriched uranium (HALEU), to promote the research and development of advanced nuclear fuels and deploy.

6. Investment advice

[From computing power to energy: why the energy infrastructure track is recommended now]

Existence of expectation gap: The core logic of currently recommending energy tracks stems from the fact that the technology industry chain driven by AI is extending from computing power ecology to energy IT infrastructure, and the market currently has no confidence in the medium and long term of this key link. The value is not yet fully understood. At the same time, the market believes that China's power infrastructure is complete, AI accounts for a small proportion, and it is difficult to be flexible. However, we believe that the increase in global computing power has become a trend, and the increase in computing power consumption is inevitable. Our advantages in the field of IT infrastructure can better take advantage of this east wind to achieve overseas layout.

1. From computing power to energy: the inevitable path driven by the industry chain

AI industry Under accelerated development, various subdivisions such as GPU and CPU, storage, communications, and copper cables have become hot topics in the current market. However, behind these computing power ecosystems, there is a strong reliance on the continuous supply of energy and infrastructure. From downstream AI application scenarios (whether it is games, finance, or medical care, etc.) to upstream basic supporting facilities (including cooling, IDC, energy, etc.), every link is interconnected. However, the current market is facing the backdrop of fierce competition in computing power. At present, power and IT infrastructure have become the bottleneck of the North American computing power market. There is no doubt that computing power software and hardware has become a hot spot in the market. Looking forward to the next 3-5 years, to explore more computing power-related opportunities, it is even more necessary to study the upstream infrastructure links in advance, especially the energy link.

2. The market’s neglect of mid- and long-term infrastructure planning has created an investment window for the energy track

1) The contradiction between supply and demand is becoming increasingly prominent

The power supply in North America is in a tight balance, and the demand for AI computing power is increasing rapidly. It is expected thatGlobal data center installed power demand will grow from 40GW to 140GW in 2029. This exponential increase in energy consumption has exposed the shortcomings of current infrastructure planning. The expansion of power infrastructure lags behind, and the production cycle of key equipment such as transformers restricts energy supply capabilities.

2) Long-term hidden dangers caused by short-term market behavior

The current capital market has a negative impact on computing power Relevant tracks (such as GPU, storage, communications) have attracted a lot of attention, but insufficient attention has been paid to the mid- to long-term layout of infrastructure such as cooling, IDC, and energy. These infrastructure links are the core of promoting the sustainable development of computing power ecology. Take liquid cooling as an example. When we released an in-depth report on the liquid cooling industry in early 2024, industry changes were not taken seriously in the capital market. Its market popularity gradually emerged after the AI ​​power consumption demand exploded. Similar logic is the same for the energy racetrack, and it's just getting started.

3. Why is the energy track worth laying out in advance?

Energy is the core element of competition in the next stage of the technology industry, and the market’s planning and understanding of energy are still in its infancy:

Strategic scarce resources: Sam Altman mentioned that the most important resources in the future are computing power and energy. The energy track is not only the support of computing power, but also the basis for achieving sustainable technological development.

Initial investment window: The current AI energy track is in its infancy, the investment valuation is relatively reasonable, and there is plenty of room for upward revisions in future demand.

Driven by collaboration with technology: Take SMR nuclear power as an example. It has the characteristics of low carbon, environmental protection, and efficient power supply, and is highly consistent with the global carbon neutrality goal. Natural gas, as a transitional energy source, will also benefit from data center expansion needs in the short term.

To sum up, the logic of the energy track is consistent with our recommendation for the liquid cooling industry last year The logic is similar: in the early stages of the industry's outbreak, market valuations were not low, but investments were made for long-term growth rather than short-term cheap prices. Energy is the next battle in technological competition. Just as liquid cooling has evolved from optional to mandatory, the AI ​​upstream infrastructure track is also moving from traditional industries to core technology supporting equipment. Seizing the opportunity for deployment is the key to winning in the future.

6.1 SMR Nuclear Power US Stocks

SMR nuclear power can meet power supply needs as a single solution, so we have sorted out the upstream, midstream and downstream players in the industry chain:

< /p>

6.2 Natural Gas + Multi-Energy US Stocks

Except SMR Nuclear power also has many ways to address energy challenges, including natural gas power generation, renewable energy (such as solar energy, wind energy), energy storage systems, and the use of innovative technologies such as fuel cells, usually using a composite solution of natural gas + other energy sources, so we The main players in each link are sorted out:

Natural gas power generation: NextEra Energy (NYSE: NEE), Dominion Energy (NYSE: D), Cheniere Energy (NYSEAMERICAN: LNG ) wait.

Renewable energy (solar and wind): First Solar (NASDAQ: FSLR), Enphase Energy (NASDAQ: ENPH), Brookfield Renewable Partners (NYSE: BEP) wait.

Energy storage technology (to balance the intermittency of renewable energy): Tesla (NASDAQ: TSLA), Fluence Energy (NASDAQ: FLNC), etc.

Fuel cells and distributed generation (fuel cells run on natural gas or hydrogen): Bloom Energy Corporation (NYSE: BE), Plug Power (NASDAQ: PLUG) wait.

Core energy efficiency technologies (data center cooling): Vertiv (NYSE: VRT), Schneider Electric, etc.

6.3 A-share related targets

7. Risk warning

1. Technology and regulatory risks.

SMR technology is still in the research and development and early deployment stages, and many designs have not yet received full regulatory approval. The development cycle is long and there are technical uncertainties, such as safety testing, materials Performance verification, etc., any technical failures or regulatory delays may significantly increase costs and affect the commercialization process.

2. High capital requirements and financing pressure.

The development and deployment of SMR requires huge capital investment, including design, construction and approval costs. Many start-ups rely on external financing to maintain operations. Once the capital chain is broken, they may causing the project to be terminated. In addition, the investment return cycle is long and it takes decades to recoup the initial investment.

3. Market demand and competition risks.

Market acceptance and demand may be affected by energy sources, public attitudes and technology substitution (such as energy storage technology and green hydrogen), if market demand is insufficient or support decreases , SMR companies may face profitability difficulties.

Abstract

Standing at the current point in time, we re-evaluate the development trend and investment of AGI or expectations. The market starts with computing power, extends to GPU, optical modules, switches, storage and other tracks, and leverages overseas mapping to eagerly look forward to AI applications, but ignores the pull on upstream infrastructure when computing power increases. If applications are the most explosive direction, then the infrastructure will take a long time to develop. Not only liquid cooling, but also the demand for energy is fundamental. This is also the starting point of this article.

Marginal changes: One of the biggest differences between AIDC and traditional data centers is that the level of electricity consumption has increased significantly. AIDC has the characteristics of large data volume, complex algorithms and 24/7 instant response. Therefore, compared with traditional data centers, AIDC consumes a lot of power. With the rapid development of AI, it is expected that AI software integrating large language models will develop rapidly. Training needs and inference needs will resonate. In the future, the power consumption of data centers will increase significantly. AIDC will become a new generation of "electric tigers", and data center consumption will increase. The proportion of electricity will further increase. SemiAnalysis predicts that global data center critical IT power demand will surge from 49GW in 2023 to 96GW in 2026, of which AI will consume approximately 40GW. Vertiv predicts that data center power consumption will increase by 100GW in the next five years, and global data center power demand will reach 2029increased to 140GW.

Dilemma: The US power grid is difficult to support the development of AI computing power. Compared with the construction speed of data centers, the current construction speed of the U.S. power grid is relatively slow and the power generation capacity is limited. Therefore, in the short term, the United States will face a power demand dilemma due to the development of AI. Currently, U.S. power supply faces obstacles such as long infrastructure construction cycles, shortage of infrastructure facilities, labor shortages, lack of experience among practitioners, and the need to coordinate multiple stakeholders when building a power grid. The rapid development of AI has led to power supply shortages in some areas. North American utility company Dominion Energy said it may not be able to meet Virginia's power needs, leading to years of delays in the construction of the world's fastest-growing data center hub.

Solution: short-term - natural gas, medium-term - SMR nuclear power, long-term - controllable nuclear fusion. The rise of AI is leading resource competition to computing power + energy. In the AI-driven digital world, computing power is the basis for iteration and innovation, and energy is the key to supporting the operation of these computing power. In the short term, natural gas combined with fuel cells will provide flexible and efficient power generation solutions for data centers to meet the current rapid expansion needs. In the medium term, small modular reactors (SMRs) will become a key path to address power bottlenecks in data centers due to their stability and adaptability to distributed deployment. In the long term, controllable nuclear fusion is expected to completely break through energy supply constraints and provide unlimited and clean power support for the future computing power ecosystem. In this process, from the continuous innovation of energy technology to the efficient collaboration of computing power ecology, it not only promotes the leap of AI technology, but also reshapes the future pattern of deep integration of energy and computing.

We believe that we are still in a battle for computing power, but looking forward to the next five years, the battle for energy infrastructure may become mainstream. In the short term, the capital expenditures of CSP giants in the third quarter of this year all reached new highs, and tended to be on the computing power side. In the next 5-10 years, combined with the continued increase in investment in AI computing power and the current power supply situation in the United States, we believe that the current power supply in the United States will be flat. The era of 2020 is coming to an end, and the battle for computing power will gradually transform into a battle for energy. The investment plans of computing power giants such as Amazon, Microsoft, and Google in nuclear power projects such as SMR have initially proved this. The participation of IT giants will significantly introduce new technologies and accelerate iteration, and investment opportunities in related energy infrastructure will gradually emerge.

Investment advice: To sum up, energy is the next battle in technological competition. Just like the process of liquid cooling from optional to mandatory, AI upstream The infrastructure track is also moving from traditional industries to core technology supporting facilities, and seizing the opportunity for layout is the key to winning in the future. It is recommended to pay attention to the core targets of U.S. stocks such as ETN, EMR, SMR, OKLO, NNE, BE, etc. A shares are in nuclear power, spaceFor the natural gas and infrastructure supply chain, it is recommended to pay attention to Guangdong Nuclear Power, Nuclear Power, New Natural Gas, CGN Mining, Jinpan Technology, Invic, Megmeet, Nengke Technology, Kehua Data, Eurofins, One Stone, etc.

Risk reminder: technology and regulatory risks, high capital requirements and financing pressure, market demand and competition risks

Investment requirements

OpenAI founder Sam Altman once said in an interview: The two important resources in the future will be computing power and energy. AI’s pursuit of performance has gradually become more intense in the field of computing power, and the core factors of competition in the next stage will initially appear in energy infrastructure.

[From computing power to energy: the next battle in technological competition]

Artificial intelligence The rise of intelligence has more directly led resource competition to computing power and energy. In the AI-driven digital world, computing power is the basis for iteration and innovation, and energy is the key to supporting the operation of these computing power. "The two most important resources in the future are computing power and energy." This trend will run through every stage of AI technology development, from algorithm optimization to hardware breakthroughs to the current demand for efficient energy systems.

[Acceleration requirements for computing power and hardware limits]

The demand for AI computing power is Exponential growth. Taking NVIDIA H100 GPU as an example, the computing power of 60 TFLOPS is promoting large-scale training of large models, and the surge in computing power has brought huge energy consumption challenges. Vertiv predicts that by 2029, the total installed power demand of global data centers is expected to soar from 40GW to 140GW, while the value of data centers per MW will increase from US$2.5-3 million to US$3-3.5 million. The power consumption of Nvidia's next-generation product, Rubin ultra, in a single cabinet is over 1MW. It also shows that the increase in AI computing power is exerting unprecedented pressure on power infrastructure. How fast the calculation is depends largely on the power.

[The emergence of energy bottlenecks andInfrastructure Challenges]

The expansion of data centers has exposed the vulnerability of the power supply system. Elon Musk once pointed out that the production capacity of key electrical equipment such as transformers is difficult to meet the current demand for AI, and this shortage of power infrastructure will further amplify the load fluctuation effect of the power grid, especially during the peak period of AI training. Power demand may instantly exceed the average load by several times, and peak and valley power consumption patterns pose a huge threat to the stability of the energy system. This bottleneck was not obvious in the early stages of AI development, but will become more obvious as the cluster scale expands and AI applications increase in volume. This dilemma can be seen in the implementation of Sora.

[Energy technology innovation and computing power ecological synergy]

With the rapid growth of computing power demand In the context of the epidemic, energy bottlenecks are becoming the core obstacles limiting the development of AI. Nuclear energy, especially small modular reactors (SMR), has gradually emerged and become one of the best solutions to adapt to AIDC. Emerging nuclear energy companies represented by OKLO\Nuscale are developing micro-reactor technology, and cloud service providers such as Google and Microsoft have launched SMR project layouts. The goal is to power future data centers through distributed small nuclear power plants and provide continuous and stable computing. Strong support. Natural gas + fuel cell/clean energy/energy storage and other solutions are also being actively promoted as one of the options for rapid implementation. Start-up companies represented by Bloom Energy are also rapidly rising with the help of industry trends.

From an investment perspective, the market has already recognized the importance of computing power, and is eagerly looking forward to the implementation of applications, constantly looking for mapping, and ignoring The importance of AI infrastructure is not just an opportunity for liquid cooling and computer rooms. From a larger perspective, the next stage of competition is gradually gaining momentum in various energy (natural gas, nuclear power, etc.) fields.

1. "Electric Tiger" AIDC and weak power grid

1.1 Electricity consumption: AIDC's next shortcoming

1.1.1 Supply and demand of electricity in the United States

Demand side: Data centers are already “big consumers of electricity”, accounting for the largest share of electricity consumption in the United States 4%. In 2023, the total power of U.S. data centers will be about 19GW. According to this estimate, the annual electricity consumption will be about 166TWh (terawatt hours), accounting for 4% of the national electricity consumption.

The data center consumes 166 TWh of electricity, which exceeds the annual electricity consumption of New York City and is equivalent to the annual electricity consumption of 15.38 million household users. From a regional perspective, 2In 2 years, New York's annual electricity consumption was 143.2TWh, Texas' annual electricity consumption was 475.4TWh, California's 251.9TWh, Florida's 248.8TWh, and Washington's 90.9TWh. The annual electricity consumption of U.S. data centers exceeded the annual electricity consumption of New York City. The average annual electricity consumption per residential user in 2022 is 10,791kWh. Based on this estimate, 166TWh is equivalent to the annual electricity consumption of approximately 15.38 million household users.

*1 TWh = 1000 GWh = 10^6 MWh = 10^9KWh

< p>

Supply side: The annual power generation in the United States is relatively fixed. Currently, thermal power is still the main source, and new energy power generation is growing rapidly. , the proportion of nuclear energy has further increased. The annual power generation in the United States is approximately 4,000-4,300 terawatt hours (TWh), of which thermal power (coal, natural gas, oil) accounts for approximately 60% in 2023 and is the main energy source; new energy power generation (wind energy, solar energy) etc.) has grown rapidly in recent years, accounting for 21%; nuclear energy accounts for approximately 19%, and its proportion has further increased.

Electricity prices: The United States has the lowest electricity prices in the world One, and some states have lower electricity prices due to their energy advantages. The U.S. electricity consumption structure is mainly divided into four areas: residential, commercial, industrial and transportation. In September 2024, the electricity price for residential users is US$0.17/kWh (approximately 1.24 yuan/kWh, the exchange rate is as of December 13), and the electricity price for commercial users is US$0.135/kWh ( Approximately 0.98 yuan/kWh); the industrial electricity price is 0.09 US dollars/kWh, the transportation electricity price is 0.13 US dollars/kWh, and the wholesale electricity price in 2023 is 0.036 US dollars/kWh. Some states have lower electricity prices due to their energy advantages. As of April 2024, the electricity price in Texas (rich in natural gas and renewable energy) is approximately US$0.147/kWh, and in Louisiana (rich in energy resources), the electricity price is approximately US$0.147/kWh. It is US$0.115/10 million hours, and Tennessee (rich in hydropower resources) is about US$0.125/kWh. Some large-scale power-consuming infrastructure, such as data centers, are often built in provinces with low electricity prices. The above-mentioned state capitals have also become the concentration of today's computing power industry.

Estimation of annual electricity cost for data centers: Based on the wholesale price of US$0.036/kWh, U.S. data centers (when AI has not yet been applied on a large scale) consume 166TWh of electricity per year. , estimated to require approximately US$6 billion.

1.1.2 Marginal changes: AI vs. Challenges of the power grid

[Challenges1: The total power consumption has increased significantly]

Compared with traditional data centers, AI data centers consume a lot of power. The main reasons are the huge growth in data volume, complex algorithms and the need for 24/7 instant response. For example, a Google traditional search request consumes about 0.3Wh, while a ChatGPT request consumes 2.9Wh, which is ten times the former; a paper published in "Joule" stated that if Google uses AIGC for every search, its usage Electricity will rise to 29 billion KWh per year, which will exceed the total electricity consumption of Kenya, Croatia and many others; according to the New Yorker Magazine, ChatGPT consumes more than 500,000 KWh every day.

[Challenge 2: Increasing voltage swing by using electricity]

< p>

Phenomena: The current demand of the AI ​​data center (whether for training or inference) is highly transient, with huge swings occurring within a few seconds. As the task load of the neural network model increases or decreases, the current demand will fluctuate wildly, even up to 2000A per microsecond.

Principle: 1) Peak load fluctuations: The training and inference of AI models require huge computing power, but they do not run continuously. There will be peak loads when model training starts, and troughs. When operating, the basic operation is maintained, causing power consumption to fluctuate; 2) Dynamic resource scheduling: AI Tasks are cyclical. For example, large-scale training requires centralized resources, while the inference stage is relatively dispersed, which makes the power consumption curve more unstable; 3) Real-time response requirements: Generative AI and large model applications require low latency and high throughput, driving the foundation Facilities expand in real time, further amplifying power consumption fluctuations.

Result: Affects the stability of the power grid. The design of the power grid is not suitable for excessive swing voltage. The power grid is basically designed for the power load. We hope to see a relatively stable, regular and slowly changing load. For example, an electrical device with a power load of 100GW may change after being connected to the power grid. There are two 200GW transmission lines for power supply. If one of the two transmission lines is normal, operation can be guaranteed. The AI ​​power consumption characteristics will have huge swings within a few seconds, and this violent fluctuation may affect the stability of the power grid.

[Challenge 3: Subsequent electricity demand Larger]

Inference in the AI ​​data center consumes more energy than training due to the large number of requests from users. Currently, Google has announced in the first half of this year that it will add new AI features to improve the search experience and will launch Gemini-based AI Overviews, which is already available for trial for some users; Microsoft has launched a personal AI assistant called Microsoft Copilot and has already ChatGPT is integrated into Bing. At present, the number of visits to Google search engine has reached 82 billion times per month. OfThe number of paying users of fice's commercial products has exceeded 400 million. The huge user base means that if the trained large model is integrated into the company's products, the number of user requests will increase significantly, and the number of AI instant responses will increase sharply, causing model inference to consume more energy than training. Energy consumption. According to McKinsey estimates, U.S. data center power loads may account for 30% to 40% of all new demand by 2030.

Conclusion: With the rapid development of AI, it is expected that the integration of large language models AI software will develop rapidly, and training needs and inference needs will resonate. In the future, the power consumption of data centers will increase significantly. AIDC will become a new generation of "electric tigers", and the proportion of power consumption in data centers will further increase.

1.2 Realistic dilemma: The power grid is difficult to support

The economic development structure determines that the power grid infrastructure in North America is relatively weak. In the past 20 years, the decoupling of electricity demand and economic growth in the United States has accelerated sharply. Since 2010, the U.S. economy has grown by a cumulative 24%, while electricity demand has remained almost unchanged. In 2023, U.S. electricity consumption even dropped by 2% from 2022. Its essence is that it is different from the fact that the economy is mainly driven by industry and service industries. The economic growth of the United States does not mainly rely on electricity or energy consumption, but mainly relies on high-tech industries, with low energy consumption. And efficiency gains, primarily the replacement of incandescent lights with fluorescents and LEDs, have offset demand for electricity from population and economic growth, leaving utilities and regulators without expanding grids or generating capacity.

Situation: Lack of time, lack of people, Lack of infrastructure, lack of experience, and lots of resistance.

Lack of time: It takes about two years to build a data center, but the construction of the power grid is much slower. It may take three to five years to build a power station, and it may take three to five years to build a long-distance high-capacity power grid. Transmission lines will take 8 or even 10 years. According to MISO, the U.S. regional transmission organization, the 18 new transmission projects it is planning could take seven to nine years, compared with 10 to 12 years for similar projects historically. It can be deduced from this that the construction speed of the power grid is likely to be unable to catch up with the growth rate of AI.

Lack of infrastructure: According to the power investment trend in the United States, capital expenditures by U.S. utilities increased significantly from 2016 to 2023, especially in the fields of power generation, distribution and transmission, and grid investment began to accelerate in 2018. , mainly due to the reshoring of manufacturing industry to promote power demand. Against this background, the United States still has not expanded its power grid on a large scale. According to Grid According to a survey report issued by Strategy, the United States installed an average of 1,700 miles of new high-voltage transmission miles per year from 2010 to 2014, but this dropped to only 645 miles per year from 2015 to 2019.

Short of people: LaoPower constraints are also a constraint, particularly a shortage of specialized electrical workers necessary to implement new grid projects. According to McKinsey estimates, the U.S. could see a shortage of 400,000 professional workers based on projected data center construction and similar assets requiring similar skills.

Lack of experience: For the United States, the entire power industry has not seen a large-scale increase in power demand in the past 20 years, and these 20 years are likely to mean a whole group of engineers , the staff have no experience in large-scale construction of new power grids.

Many resistances: The construction of the power grid requires infrastructure such as power stations and transmission lines, which may require countless stakeholders to work together to reach compromises on line directions and costs.

Conclusion: Compared with data center Construction speed. The current construction speed of the US power grid is relatively slow and the power generation capacity is limited. Therefore, in the short term, the United States will face a power demand dilemma under the development of AI. For example, North American utility Dominion Energy said it may not be able to meet Virginia's power demand, leading to years of delays in building the world's fastest-growing data center hub. In the power industry, new infrastructure planning takes five to 10 years, according to Wood Mackenzie. Additionally, most state public utility commissions have little regulatory experience in a growth environment. It can be inferred that electric energy may become one of the biggest constraints on the development of AI in the next few years. Although the market is paying attention to innovative solutions such as controllable nuclear fusion, water from afar cannot quench the thirst for nearness, and it is inevitable to form short-, medium-, and long-term comprehensive solutions.

1.3 Multi-angle measurement: How much power does AIDC consume?

*Total power (GWh) = Total power (GW) × time (h)

*Total power (GW) = IT equipment power (GW) × PUE (energy efficiency ratio)

1.3.1 Calculation angle one (conservative): AI chips

Calculation logic: Calculation angle one is based on the number of chips and deduces to 20 In 30 years, we will use the number of chips * chip power consumption to predict the total power consumption. This does not take into account that the overall power consumption of the server will be greater than the number of single chips * the number of chips. It does not take into account the possible increase in single chip power consumption after future chip upgrade iterations. Therefore, we It is believed that the calculation angle 1 is a "conservative" calculation, and the calculation data is the smaller of several methods. The AIDC power demand in 2030 is 57GW.

Number of GPUs and TPUs in use: According to DCD reports, the total shipments of GPUs in the data centers of Nvidia, AMD and Intel in 2023 are estimated to be 3.85 million units. The number of TPUs produced for Google is expected to be 930,000. Further tracing the supply chain, TSMC predicts the year-on-year growth rate of demand for AI server manufacturing from 2024 to 2029.About 50%. Based on this calculation, GPU shipments in 2030 will be approximately 65.78 million, and TPU shipments will be approximately 15.89 million. According to NVIDIA’s official statement, the average service life of most H100 and A100 is 5 years. Therefore, we assume that the number of chips in use in 2030 is the sum of chip shipments in 26-30 years. Therefore, the number of GPUs and TPUs in use in 2030 is approximately were 171.36 million and 41.39 million.

GPU and TPU power consumption: The maximum power of H100 NVL can reach 800W. Then there are expected to be 171.36 million GPUs in 2030. Assume that GPU and TPU energy consumption account for 90% of the total energy consumption of IT equipment. Assume that the United States accounts for 34%, the utilization rate is 80%, and the PUE is 1.3. Calculated, in 2030 the United States AIDC GPU power requirement is approximately 54 GW (number of GPUs * GPU power consumption * U.S. share * PUE * utilization rate ÷ chip share = 171.36 million * 0.8kW * 34% * 1.3 * 80% ÷ 90% = 54GW);

According to Google’s official statement, TPU The average power of the v4 chip is 200W. Combined with the above estimate of approximately 41.39 million TPUs in use in 2030, we estimate that the total power consumption of TPUs in 2030 will be approximately 3.3GW (other indicators are assumed to be the same as GPUs).

Angle 1 Conclusion: The total AIDC power consumption in the United States in 2030 will be 57GW. The chip inventory from 23 to 26 years only takes into account the chip shipments after 23 years. The other calculation methods are the same as the above. The calculation methods from 27 to 30 are the same as the above calculation methods. Finally, the GPU and TPU power consumption is added up to get the U.S. The power capacity required by AIDC will reach 3/6/10/17/25/38/57GW respectively in 24-30 years.

Hypothesis 1: Chip growth rate is 50% per year (refer to TSMC’s statement) .

Assumption 2: Assume that the average chip life is 5 years (refer to the GPU life given by NVIDIA).

Assumption 3: The average power utilization rate of IT equipment is 90% (considering the power consumption of NVSwitches, NVLink, NIC, retimers, network transceivers, etc. in IT equipment, assuming that GPU and TPU energy consumption account for 90% %, other IT equipment energy consumption accounts for 10%).

Assumption 4: Considering that IT cannot run at full capacity and cannot run 24 hours forever, refer to Semi analysis and set the possible utilization rate to 80%.

Assumption 5: PUE is 1.3 (PUE is the total power consumption of the data center divided by the power used by IT equipment).

Hypothesis 6: The United States’ computing power demand accounts for 34% of the world’s computing power demand (as measured by the Institute of Information and Communications Technology, the United States’ share of global computing power is 34%).

1.3.2 Calculation angle two (optimistic): data center

Calculation logic: Calculation angle two is from the perspective of data center construction, referring to the global data center construction progress predicted by a third party ( compound growth rate of 25%), and since the forecast data ends in 2026, we assume 20 From 27 to 2030, the compound growth rate of 25% will still be maintained. We forecast the power demand of global data centers and assume the power consumption and proportion of AIDC. Therefore, we believe that the data obtained from this forecast perspective is relatively "optimistic". In the end, It is predicted that the AIDC power demand in the United States will reach a maximum of 91GW by 2030.

Research firm SemiAnalysis used analysis and construction forecasts of more than 5,000 data centers and combined these data with global data and satellite image analysis to predict data centers in the next few years. The growth of power capacity will accelerate to a compound annual growth rate of 25%, and the proportion of AIDC will further increase. In terms of data centers, according to the forecast According to measured data, global data center critical IT power demand will surge from 49GW in 2023 to 96GW in 26. We assume that the data center will continue to maintain a compound growth rate of 25% from 27 to 30 years (refer to the growth rate from 2023 to 2026, which is 25% %), then by 2029 and 2030, the key IT power demand of global data centers will increase to 188 and 234GW respectively; refer to Semi Analysis data, combined with the background of the booming development of AI computing power and the explosion of downstream applications, we believe that the proportion of AI in data centers is expected to continue to accelerate in the future, so we assume that the global AIDC proportion will reach 12%/16% respectively in 23-30 /30%/44%/56%/68%/78%/88%, thus calculating the global AIDC IT equipment power demand in 29 and 30 years to be 65GW and 91GW respectively.

Conclusion from the second perspective: Based on the US share of 34% and PUE of 1.3, the US AIDC power demand will reach 91GW by 2030.

< /p>

Hypothesis 1: Combined with the booming development of AI computing power, downstream applications are exploding. To give a broad background, we believe that the proportion of AI in data centers is expected to continue to accelerate in the future, so we assume that the proportion of global AIDC in 23-30 will reach 12%/16%/30%/44%/56%/68%/ 78%/88%.

Assumption 2: PUE is 1.3 (PUE is the total power consumption of the data center divided by the power used by IT equipment).

Hypothesis 3: The United States’ computing power demand accounts for 34% of the world’s computing power demand (as measured by the Institute of Information and Communications Technology, the United States’ share of global computing power is 34%).

1.3.3 Summary 1: AIDC accounts for the proportion of total electricity consumption in the United StatesImprovement

(1) AI power consumption accounts for an increase in the proportion of US power consumption, and the proportion is expected to exceed 10%

According to Statista forecast data, in 2022 , the electricity usage in the United States is about 4085 terawatt hours. It is expected that the electricity usage in the United States will continue to rise in the next few decades, reaching 4315 terawatt hours (corresponding to 493GW) by 2030. In 2050 it will reach 5178 terawatt hours. According to our previous "Calculation Angle 1", if the total power consumption of AIDC reaches a maximum of 57GW in 2030, then the proportion of U.S. electricity consumption will increase to 12% (57GW/493GW), which is significantly higher than the 4% in 2023.

1.3.3 Summary 2: AIDC consumption The power is expected to be comparable to Bitcoin mining

< /p>

In our report "AI Dongfeng has arrived, Bitcoin mines have begun the second growth curve" released on August 6, 2024, we made assumptions and predictions about the power consumption of Bitcoin mines. In this report, we predict The load of Texas Bitcoin mines in 2024/2025/2026/2027/2028 is 4.7/6.5/8.3/10.1/11.9GW respectively (assuming that the annual new load of Texas Bitcoin mines is 1.8GW), regarding the share of Texas Bitcoin mine load in the United States, we assume it remains unchanged at 28.5%. Therefore, according to our forecast, the annual load of Bitcoin mines in the United States is 17/23/29/36/42GW respectively.

For the convenience of comparison, we forecast the data to 2030, assuming: 1) The annual new load of the Texas Bitcoin mine is 1.8GW, 2) Assume that in 2029 and In 2030, the share of Texas mining farms will remain unchanged at 28.5%. Therefore, it is concluded that in 2024/2025/2026/2027/2028/2029/2030, the annual power consumption of U.S. Bitcoin mines is 17GW/23GW/29GW/26GW/42GW/48GW/54GW respectively.

< /p>

Conclusion: Under conservative forecasts, US AIDC power consumption will catch up with Bitcoin mining power demand in 2030; under optimistic forecasts, US AIDC power demand will exceed Bitcoin mining in 2029.

< /p>

2. What is the solution to the dilemma: short-term "natural gas +" is the mainstream

2.1 The fastest implementation solution in the short term is natural gas

2.1.1 Substations have become the bottleneck of traditional electricity consumption

[Current status of data center power supply]

Purchasing power and substations: Data centers usually purchase power by signing a contract with the power company, which means that the power supply of the data center is the current generated from the power station It is transported to the data center through the transmission network. However, after power is transported over long distances, the voltage often needs to be adjusted through substations to ensure that the power meets the voltage needs of the data center.

Necessity of substations: Substations convert high-voltage electricity into low-voltage suitable for local use. Most power systems require voltage conversion and distribution through substations. Without a local substation, power cannot be used directly in the data center.

The construction of substations is difficult, takes a long time, and costs high: The construction of substations usually requires a large amount of capital investment, involving land, infrastructure construction, equipment procurement, and manpower. Reserves etc. In addition, substation construction takes a long time and needs to meet strict environmental and safety standards.

Conclusion: Under the current electricity purchase method, substations have become the bottleneck restricting AIDC's power consumption. As data center power demands continue to grow, building new substations or expanding existing substations takes a long time and requires significant approval and construction time, which may not keep up with data center demand quickly.

[Natural gas does not require substations, it is distributed The first choice for power supply】

Natural gas power generation does not rely on substations. Natural gas power generation generates electricity by burning natural gas. Natural gas power stations are usually connected to data centers through dedicated pipelines. The natural gas is directly transported to the power generation facility for combustion and power generation. The generated electricity is then supplied to the data center through the local power grid or dedicated lines. It can usually be It is completed in power generation facilities near the data center. Unlike traditional power transmission methods, natural gas power generation does not need to pass through the high-voltage power transmission network and therefore does not rely on remote substations and power transmission facilities. Natural gas power generation can build small natural gas power stations (such as distributed power generation systems) near data centers, reducing dependence on external power grids and shortening the response time of power supply.

2.1.2 There is a time lag between the rapid development of AI and the implementation of SMR nuclear power

Although nuclear power has advantages in many aspects, the most important demand in the North American computing power market is "rapid implementation". Quickly light up the GPU to obtain computing power, and natural gas has become the current first choice.

Although in February 2023 the U.S. Nuclear Regulatory Commission approved the design of the first nuclear power company Nuscale PowerSMR (Small Modular Reactors), and countries around the world such as China and Russia are racing to put SMR technology into practice. However, the commercialization of SMR will still take some time, and the safety approval process is complex and time-consuming. It can already be seen that SMR has aroused global interest in nuclear energy. The U.S. nuclear fission industry has received a boost from the Inflation Reduction Act, which includes a number of tax credits and incentives while providing $700 million in funding for the Office of Nuclear Energy to support the development of high-purity low-enriched uranium (SMR). supply of fuel); there are more than 70 commercial SMR designs under development around the world, and there are currently two SMR projects operating in China and Russia. But according to the U.S. Energy Regulator, nuclear reactors are extremely complex systems that must meet strict safety requirements and account for a wide variety of accident scenarios, and the licensing process is cumbersome and varied. This means that SMR requires certain standardization before it can enter the commercial market, so other solutions need to be found to solve the short-term energy shortage problem.

2.2 “Natural gas + multi-energy” combination More robust

The combination of natural gas + other multi-energy sources is currently the fastest implementation solution that can meet the power needs of AI. Compared with SMR nuclear power, an independent solution with high energy density but long deployment period, natural gas power generation can be used as a basic energy source to quickly respond to load demand due to its high efficiency and flexibility. At the same time, it can be integrated with renewable energy, fuel cells, and energy storage systems. Used in conjunction, they can effectively compensate for intermittency and lack of stability. This multi-energy combination can not only meet the needs of AI data centers for stable power supply, but also provide a balance between carbon emissions and costs, making it an important choice for current data center energy strategies.

Collaboration is not necessary, but for large-scale AI data centers that need to comprehensively balance stability, environmental protection and cost, collaborative use of multi-energy solutions is a more flexible and long-term choice. When there are clear goals (such as low cost, ultra-fast deployment), a single solution can also meet:

[Only natural gas power generation (single solution)]

Advantages: Natural gas power generation itself can be used as an independent power supply solution, which is suitable for scenarios with high requirements for stable power demand and rapid deployment, especially those that require high reliability. Scheduling AI data center;

Limitations: Although the deployment speed is fast, the carbon emissions will be high in the long run.

[The necessity of multi-energy coordination]

More stable and safe: AI data centers have extremely high requirements for power continuity (short-term power outages are not allowed) ), natural gas + energy storage system or fuel cell can be used as backup support;

More environmentally friendly: natural gas + wind energy, solar energy and other low-carbon energy combinations.

2.3 Natural gas solution: taking xAI as an example

Natural gas power generation technology path is mature, supporting equipment is complete, It is cost-effective and is the fastest option to solve the AI ​​power shortage problem in the short term. Tesla xAI uses a natural gas solution as emergency power supply. A natural gas generator is a generator that uses natural gas instead of gasoline or diesel. Compared to diesel, natural gas is less expensive to purchase and does not suffer from "wet accumulation" problems. Therefore, from the perspective of short-term energy solutions, natural gas generators have the advantages of cost-effectiveness, high operating efficiency, and more environmental protection than other generators using fossil fuels such as oil. According to DCD reports, Tesla CEO Musk has purchased 14 mobile natural gas generators from Voltagrid, each of which can provide 2.5 MW of power, to alleviate the power shortage problem in the data center of his startup xAI.

*Additional details 1: Musk xAI mainly uses NVIDIA H-series servers, and the cluster heat dissipation uses a liquid cooling solution. Each liquid-cooled rack in the xAI data center contains 8 Nvidia H100 GPU servers, with a total of 64 GPUs. The dense layout requires each computing node to dissipate heat efficiently. The traditional air cooling method is difficult to adapt to, so xAI chose AMD's Liquid cooling solution.

*Additional details 2: The xAI data center also uses the Megapack energy storage system. xAI said that when building the computing cluster, its team found that the AI ​​server did not run at 100% power all day long, but had many peaks and valleys in power consumption, so Tesla's battery storage product Megapack was added in the middle to buffer the fluctuations. Thereby improving the reliability of the overall system and reducing power loss.

2.4 Fuel Cell: Taking Bloom Energy as Example

Company Profile: Bloom Energy is focused on developing efficient, low-emission energy technologies through innovative solid oxide fuel cells (SOFCs) and solid oxide electrolysers. (SOEC) technology to promote global energy transformation. As a leading clean energy company, the company is committed to providing sustainable, reliable energy solutions for high-demand areas such as industry, commerce and data centers through its advanced hydrogen and fuel cell technologies. The company was founded in 2001, headquartered in California, USA, and has expanded its business globally.

Core technologies: The company's core technologies include solid oxide fuel cells (SOFC) and solid oxide electrolyzers (SOEC). The SOFC system provides efficient power generation when using 100% hydrogen. Electric power output and electrical efficiency are as high as 65%, far exceeding traditional energy systems. Bloom Energy’s fuel powerThe pool system can also integrate combined heat and power (CHP) technology, making the total energy efficiency up to 90%, thereby effectively reducing energy consumption and carbon emissions. In addition, SOEC technology can be used for efficient hydrogen production and is one of the key technologies in the clean energy transition.

Product application: The company's products are widely used in many fields, including industrial power Supply, commercial energy management and data center energy solutions. Especially in the field of data centers, as the demand for energy efficiency and carbon neutrality goals continues to increase, Bloom Energy's fuel cell technology is more prominent in its high efficiency and low emissions. Its hydrogen solutions can not only meet the needs of large-scale energy It can also provide enterprises with reliable backup power to ensure the continuity and stability of operations. At present, Bloom Energy's market has covered many regions such as North America, Asia and Europe, especially in South Korea's cooperation with SK Ecoplant. Bloom Energy's hydrogen fuel cell project is expected to be online in 2025. Additionally, the company has announced a gigawatt fuel cell purchase agreement with AEP to power AI data centers.

< /p>

3. Medium-term plan: SMR nuclear power stands out

3.1 Why nuclear power: more suitable for AI

3.1.1 Characteristics of AIDC: Distributed and High Density

Compared with traditional IDC data centers, AIDC computing power centers have two most significant features: The difference is also an important feature of AIDC.

[AIDC Feature 1: Distributed Deployment]

The application scenarios and task requirements of AI determine that AIDC needs to adopt a distributed deployment method. There are significant differences between AIDC and traditional IDC in terms of computing requirements, application scenarios, resource consumption, etc. AIDC tasks are usually computationally intensive, especially large-scale deep learning, machine learning, data analysis and other tasks in the AI ​​field. A single calculation Nodes cannot carry all tasks. Therefore, AIDC needs to split computing tasks into multiple small tasks and distribute the tasks to multiple nodes for parallel computing through a distributed computing framework. This requires data centers or computing nodes in multiple geographical locations. Work collaboratively.

[AIDC Feature 2: 24-hour high-density computing]

The persistence and high load of AI computing tasks determine that AIDC must be high-load 24 hours a dayOperation requires higher power resources and cooling support. AI model training is often a long-term process that requires continuous computing power support. Therefore, AIDC usually performs long-term and continuous computing tasks; the load of traditional IDC generally fluctuates according to business needs, and many applications do not require such long-term computing tasks. , uninterrupted computing support. Therefore, AIDC's high-power computing hardware requires round-the-clock strong power supply and cooling support.

AIDC’s distributed deployment + high density The computing characteristics determine that other energy sources are difficult to meet the needs of adaptation, and small nuclear power SMR best meets the power supply needs.

Thrust - other energy sources are not suitable for AI needs, and the stability and geographical location are difficult to meet AIDC:

The seasonality of hydropower is obvious , it is difficult to meet the stable and large power supply needs, and at the same time, the geographical location with abundant water resources The location is fixed, making it difficult to meet the distributed deployment requirements of AIDC. At the same time, hydropower requires a distribution network to transmit power, which has higher overall costs and higher construction costs and time;

The fuel cost of thermal power is high, and Carbon emission restrictions are strict. Even if you purchase carbon emission indicators, It will make the overall cost of thermal power higher, so it is not suitable for AIDC that requires a lot of power. At the same time, thermal power also faces the problem of higher costs caused by the distribution network;

Other new energy sources (such as solar energy, wind energy, etc. ), although clean, its power generation capacity is affected by weather conditions and location Domain restrictions have a greater impact. Intermittency and instability make them unable to ensure stable operation of AIDC around the clock under high load. In addition, the conversion efficiency of some new energy sources such as photovoltaics is still low, and the later operation and maintenance costs are high. From the perspective of cost-effectiveness The point of view does not apply to AIDC either.

Attraction - SMR nuclear power has stronger comparative advantages, its modular design is suitable for distributed deployment, and it also adapts to the environmental protection requirements of carbon emission reduction. The modular characteristics of SMR technology enable it to be flexibly applied in distributed deployment scenarios. Modules can be flexibly added or reduced according to the needs of different regions, ensuring that the power supply of AIDC distributed data centers is not affected by geographical location, weather, and energy price fluctuations. Moreover, as a clean energy source, nuclear power is in line with the global trend of reducing carbon emissions and is suitable for AIDC’s demand for green energy. Therefore, nuclear power SMR is suitable as the main power supply source for AIDC.

3.1.2 Nuclear power SMR has the fastest landing speed Quick

What is SMR - a modular, smaller, and easier to deploy nuclear reactor. SMR (Small Modular Reactor) is a new development of nuclear energy technology. SMR is a type of nuclear power plant, but it is significantly different from traditional nuclear power plants. SMR is a small, modularSMRs are designed to provide smaller-scale power output and use modular components during construction to facilitate factory production and transportation. Generally, the output power of SMRs is smaller than that of traditional large-scale nuclear reactors. Before the advent of AIDC, SMR was often used in remote areas, small islands, and military bases far away from the power grid, or as a supplementary source of industrial power.

Compared with traditional nuclear power plants, SMR has Small scale, short construction time, lower cost (construction and maintenance), higher safety, cleaner and greener, longer life, etc.:

< p> Small module output power: The output power of SMR is smaller than that of traditional nuclear power plants, usually between tens to hundreds of megawatts, while the scale of traditional nuclear power plants is usually more than 1,000 megawatts. For example, NuScale's SMR module can provide 77MW per unit Power, after assembling 12 modules, it can provide up to 924 MW of electricity;

Shorter construction time: Because SMR adopts a modular design, allowing factory prefabrication and rapid assembly, for example, NuScale’s SMR nuclear power plant only takes 36 months (3 years), while that of traditional nuclear power plants The construction period is usually longer and may take more than five to ten years.

Small footprint: Traditional nuclear power plants occupy a larger area, usually larger than 1 square mile (approximately 2.6 square kilometers), while modular SMRs usually occupy a smaller area. SMR nuclear power plants predicted by NuScale It covers an area of ​​0.06 square miles, close to the size of a small park.

Lower costs: The construction costs of traditional nuclear power plants are usually high and are affected by scale effects, but the construction costs of SMRs are relatively low, partly due to the use of standardized and modular designs, which allow each module to be installed in batches production, reducing the construction and maintenance costs of individual reactors.

Higher safety: SMR designs tend to have higher passive safety features and disaster resistance, and can automatically shut down the reactor in the event of a fault without human intervention. SMR reactors are smaller in size, so they have greater High security and reliability.

Cleaner: SMR adopts advanced reactor design, which can use fuel more efficiently and reduce the generation of nuclear waste, more in line with the requirements of clean energy;

Longer life: SMR The design life span is as long as decades without changing the fuel, and the life span far exceeds the traditional power generation mode. For example, Nuscale's SMR has a design life of up to 60 years.

The principle of SMR - and large nuclear reactors Basically the same, it still generates heat energy through nuclear fission reaction to form steam, which then drives the generator to generate electricity. (1) Nuclear fission reaction: Like traditional nuclear power plants, the core of SMR is a nuclear reactor, which generates heat through nuclear fission reaction. Fissionable materials (such as uranium or plutonium) such as uranium-235 in the reactor absorbFission occurs after collecting neutrons, and the fission process will release a large amount of heat energy and neutrons; (2) Heat exchange and steam generation: The heat generated by the fission reaction in the reactor can be used to heat the coolant, and the coolant flows in the nuclear reactor. The heat is taken away and transferred to a steam generator or directly to water through a heat exchanger to form steam; (3) Steam-driven generator: the generated steam is introduced into the turbine, The rotation of the turbine drives the generator, which then converts mechanical energy into electrical energy and supplies it to the power grid or users; (4) Cooling system and safety mechanism: SMR usually uses a natural circulation cooling system or a passive safety system, utilizing natural physical processes ( Such as thermal convection) to keep the reactor cool, thereby reducing reliance on external power and equipment, these systems can automatically shut down the reactor and cool down in the event of a failure.

The composition of SMR - usually includes multiple Each module uses standardized components and can be quickly assembled and deployed. (1) Reactor core: contains nuclear fuel, nuclear fission occurs, and generates a large amount of heat energy; (2) Cooling system: heat is taken away from the reactor core by circulating coolant, which can be liquid metal (such as sodium), gas (such as carbon dioxide) or helium), or water, some The SMR design uses natural convection or a passive safety system and does not rely on external power to maintain cooling, which enhances the safety of the system; (3) Steam generator: transfers the heat-exchanged coolant to water to generate steam, which is introduced into the turbine. Drive power generation; (4) Vortex Turbines and generators: convert mechanical energy into electrical energy; (5) Control system: SMR adopts digital control system, and some also introduce AI technology; (6) Safety system: adopts passive safety system, that is, it operates without external power supply or operator intervention. situation, the system can automatically Cooling reactors. Common designs include natural convection cooling, thermal storage devices, etc. These designs can maintain the safety of the reactor through physical principles (such as thermal convection or gravity) in the event of an emergency; (7) Nuclear waste processing system: storage or Disposal of nuclear waste and radioactive materials.

At present, the main small modular reactors SMR are There are several different technology routes, the most mainstream of which is light water reactor (LWR-SMR), because the technology foundation is mature and it is easy to obtain regulatory approval. As of 2021, countries around the world have proposed more than 70 different SMR nuclear power solutions, including pressurized water reactor solutions, helium gas-cooled reactor solutions (HTGR), high-temperature gas-cooled practical reactor solutions, and sodium-cooled fast neutron reactor solutions (SFR). About half of the plans are light water reactor reactions, which evolved from second-generation nuclear power technology. The technology is highly adaptable and can be commercialized quickly. However, due to the Fukushima Nuclear Power Plant issue in 2011, the technology tree selection for nuclear power has become more complicated, and safety concerns about light water reactors have become more prominent. Safer non-light water reactor solutions have been favored, and high-temperature gas-cooled reactor solutions have also become more popular.Gradually becoming popular:

Light water reactor (LWR-SMR): based on mature light water cooling technology, such as NuScale’s design, the most mainstream and close to commercialization;

High-temperature gas-cooled reactor (HTGR): cooled by an inert gas (such as helium), suitable for high-temperature process heat requirements, such as Huaneng's high-temperature gas-cooled reactor;

Liquid metal-cooled reactor (such as sodium-cooled reactor): such as the Natrium reactor developed by TerraPower, which has efficient heat dissipation capabilities;

Molten Salt Reactor (MSR) : Use high-temperature lava as cooling fast neutron reactor (FNR): Use fast neutrons as high-efficiency fission fuel, such as the Russian BREST reactor type.

< /p>

3.3 Current status and industrial chain of SMR nuclear power

3.3.1 Cloud giants vigorously deploy nuclear power

With power shortage, various cloud giants have The deployment of SMR nuclear power, on the one hand, is that data centers have huge demand for power. SMR provides long-term and stable clean energy, which can reduce the need for On the other hand, in the long run, SMR can reduce the risk of electricity price fluctuations, optimize long-term operating costs, and help the company achieve its carbon neutrality commitment:

Amazon: As early as March this year, it began to look for nuclear power support solutions and acquired a company located in Pennsylvania for US$650 million. The Talen Energy data center park next to the Susquehanna Steam Electric Station nuclear power station; and announced three major nuclear power investment agreements in October this year, collaborating with Energy Northwest and Dominion Energy to build 960MW and 300MW SMRs in Washington and Virginia respectively; leading the investment in nuclear energy start-ups CompanyX-energy Received US$50 billion in Series C-1 financing;

Microsoft: Its support for nuclear power is also significant. In June this year, Bill Gates said that he would continue to support Wyoming in the United States through the start-up company he founded, TerraPower LLC. The state is investing billions of dollars in "next-generation" nuclear power plants, with the first commercial reactor expected to be completed in 2030; it reached a strategic agreement with Constellation Energy in September to restart Three Mile Island (Three Mile Island) nuclear power plant provides approximately 835 megawatts of power for Microsoft's data center.

Google: In October it said it had agreed to buy nuclear energy from a small modular reactor being developed by a startup called Kairos Power to develop more than 500MW of electricity and expected the first reactor to be operational in 2030 Run;

Oracle: Founder Larry EyreLeeson said in September that Oracle planned to build a 1GW data center park supported by three SMRs;

Meta: is actively soliciting proposals from nuclear power developers, aiming to promote its artificial intelligence by increasing nuclear power generation capacity It plans to add 1 to 4 gigawatts of U.S. nuclear power generation capacity by the early 2030s as the technology develops and meets environmental goals.

The huge power gap caused by AI data centers and the urgent power requirements faced by CSPs have made the trend of SMR nuclear power industry more and more obvious. It is expected that more SMR layouts will be announced in the future.

3.3.2 SMR nuclear power upstream and downstream< /p>

The SMR nuclear power industry chain covers all aspects from upstream fuel uranium mines, midstream R&D and construction, downstream operations and waste processing. Relatively speaking, upstream design and manufacturing have higher thresholds for professionalism and technical barriers, so upstream manufacturers have higher bargaining power. Due to the long and stable operation cycle of the downstream operation and maintenance links, they can bring long-term cash flow and are also relatively profitable. The profit margin of midstream project construction is subject to factors such as construction cost, project cycle and engineering risks, and the profit margin is relatively less stable than that of upstream or downstream.

[Upstream: Raw Materials and Processing]

The upstream industrial chain mainly involves the supply of basic raw materials, key equipment and nuclear fuel required for nuclear energy development, mainly including Uranium mining and uranium enrichment.

(1) Uranium mining and uranium processing

Uranium mining: The global uranium supply market is highly concentrated, and the United States mainly relies on imports of uranium. Global uranium mines are mainly dominated by Kazakhstan, Canada and Australia.

Main uranium mining and typical companies that mine locally: Kazatomprom in Kazakhstan, Canada's Cameco and Orano (formerly Areva, a French company but mines uranium globally) and Denison Mines, Australia's BHP (BHP Billiton) and Rio Tinto (Rio Tinto Group), Russia's Rosatom, etc. In addition, there are also some uranium mining companies in the United States, such as Energy Fuels (NYSE: UUUU), Uranium Energy (NYSE: UEC), etc.

Uranium Processing: Uranium Enrichment Technology for Safety Safety, cost and technical requirements are very high, so it is mainly dominated by a few multinational companies. Natural uranium is mainly composed of uranium-235 and uranium-238. When neutrons collide with uranium-235, a huge amount of energy will be released through a fission reaction. The fissionability of uranium-238 is smaller than that of uranium-235. Natural uranium only contains Approximately 0.7% uranium-235, so isotope separation (uranium enrichment) is required to increase its contentAs high as 3% to 5% for use as fuel in light water reactors. Concentration methods include gas diffusion, laser concentration and centrifugation.

*Principle of centrifugation method: The gaseous uranium compound uranium hexafluoride is fed into the rapidly rotating rotor of the centrifuge to separate U-235 and U-238, and the heavier ones The isotope U-238 is pushed outward, while the lighter isotope U-235 is concentrated in the center of the rotor. The gas with a higher concentration of U-235 is extracted and fed into another centrifuge, where the process is repeated several times to produce uranium with a higher concentration of U-235.

Main uranium enrichment companies: Centrus Energy (NYSE: LEU, United States, dominates the global market), Orano (France, deploys both mining and processing), Rosatom (Russia), Urenco (Europe).

< /p>

(2) Nuclear fuel assembly manufacturing

The fuel used in SMR reactors includes uranium fuel rods, fuel elements and control rods, etc. Components must meet specific standards to ensure safe and efficient operation of the reactor.

Participants: Such as Westinghouse, Orano, etc., providing nuclear fuel components and technical support.

(3) Reactor component manufacturing

Reactor components are an important part of SMR, including reactor pressure vessels, cooling systems, control systems, cores and Other related facilities, these components require a high degree of radiation resistance, high temperature resistance and reliability. Due to the modular design of SMRs, reactor components are usually mass-manufactured in factories and then transported to sites for rapid assembly, reducing on-site construction time.

Participants: Such as NuScale Power, Rolls-Royce, etc.

[Midstream: Design, R&D and Construction]

(1) SMR design and R&D

Design and R&D: Design Company Responsible for the technology development and design standardization of SMR reactors. SMR R&D usually includes nuclear reactor structural design, cooling system design, control system integration, etc. The design and R&D company works closely with departments and regulatory agencies to ensure that the design complies with nuclear safety standards.

Participants: SMR design and R&D companies such as NuScale Power, OKLO, TerraPower, Rolls-Royce, etc.; institutions such as the U.S. Department of Energy (DOE), which provide financial support and support for SMR Design supervision and verification.

(2) Reactor construction and installation

The modular design of the SMR allows most components to be prefabricated in the factory,Then transported to site for quick installation. The construction phase is simpler than that of traditional nuclear power plants because SMRs are smaller in scale and highly modular, and can be put into operation without large-scale construction. For example, the construction company is responsible for assembling the various modules of the SMR reactor into a complete For nuclear power plants, complete on-site installation and factory-prefabricated components will greatly shorten the on-site construction cycle.

Participants: Construction companies such as Bechtel, Fluor, etc., are responsible for the construction and assembly of SMR power plants.

[Downstream: operations, sales and scrap Processing]

(1) SMR nuclear power plant operation

The operator is responsible for the long-term management and maintenance of the power station, monitoring the reactor operation, and ensuring that the reactor is in a safe state. The operation and management complexity of SMR nuclear power plants is lower than that of traditional nuclear power plants. In addition, operators are also responsible for the regular maintenance of the SMR system, including fuel replacement, equipment inspections and technical upgrades.

Participants: Such as American Electric Power Company (AEP), British Electric Power Company (EDF), Southern Company, Exelon Corporation, Duke Energy (NYSE: DUK), Entergy Corporation (NYSE : ETR), PSEG (Public Service Enterprise Group, NYSE: PEG), Dominion Energy, etc. Some operators may purchase SMR power stations and operate them; management and monitoring companies will provide intelligent monitoring, data analysis and system optimization services.

(2) Electricity sales and grid connection

The electricity produced by SMR power stations is sold to grid companies or industrial users through power purchase agreements (PPA). SMR is suitable for small grids and is particularly suitable for specific markets such as remote areas, remote cities or industrial projects.

*Power Purchase Agreement (PPA): Operators sign long-term contracts with power purchasers (such as power grid companies, large industrial users, etc.) to ensure stable cash flow and profitability model.

Participants: Power purchasers such as local power grid companies, large industrial enterprises, institutions, etc.

(3) Waste and nuclear power decommissioning treatment

SMR reactors require waste management after the life cycle. Long-term storage and processing of nuclear waste is an important part of the nuclear power industry. Waste management companies are responsible for the safe handling, transportation and storage of waste to ensure compliance with nuclear safety standards. .

Participants: Waste disposal companies such as Waste Control Specialists, which specialize in the disposal of nuclear waste.

< /p>

4. Long-term outlook: Controllable nuclear fusion

Nuclear fusion is a process in which two light nuclei combine to form a heavier nucleus and release a large amount of energy. The energy released by the controllable nuclear fusion reaction is about 4 million times higher than burning coal, oil or natural gas, and 4 times more than nuclear fission. If the nuclear fusion process can be replicated on an industrial scale, it can provide unlimited clean and cheap energy. Currently, more than 50 countries are conducting nuclear fusion research. However, due to the strict conditions for nuclear fusion to occur, breakthroughs in new materials and new technologies are still needed to achieve controllable nuclear fusion. How long it will take to achieve controllable nuclear fusion will depend on the technology development progress of the industry. At the same time, it is necessary to develop the necessary infrastructure and formulate management requirements and standards for the technology. According to space reports, the British company Tokamak Energy has heated hydrogen plasma to 27 million degrees Fahrenheit for the first time in a new reactor, which is higher than the core of the sun. The company says using nuclear fusion to produce commercial electricity could be possible by 2030.

< /p>

5. Business models and participants in the energy war

5.1 SMR Nuclear Power US Stocks

5.1.1 SMR (NuScale, R&D manufacturer)

Company profile: NuScale Power is the first SMR nuclear power manufacturer to go public. The company originated from the SMR research project jointly carried out by the Idaho Laboratory and Oregon State University in 2002, and received support from the U.S. Department of Energy (DOE). NuScale Power. LLC was established in 2007 and became the first to obtain NRC ( The SMR has been designed and approved by the U.S. Nuclear Regulatory Commission and will become the first SMR technology provider to go on the market in 2022.

Core products: The company's core product SMR power module. The NuScale Power Module is the smallest light water SMR, measuring 76 feet tall,At 15 feet in diameter, a single module can generate 77 MW of electricity. The modules, including seals, are completely manufactured in the factory and transported to the factory site by truck, rail or barge, eliminating the need for on-site fabrication or construction, reducing the schedule and cost risks associated with on-site construction.

Competitive advantage: The company has its own nuclear power plant Plant – VOYGR Plant Models. VOYGR Plant Models is a standardized nuclear power plant designed by NuScale for its small modular reactor SMR. It has flexible power output and higher operating efficiency and can meet power needs of different sizes. It is the first and only design approved by the U.S. Nuclear Regulatory Commission (NRC) Approved small modular reactor.

VOYGR Plant Models different parameter modules:

VOYGR-4: consists of 4 NuScale Composed of SMR modules, providing approximately 308 MW of power output, suitable for providing power to small and medium-sized communities and industrial applications;

VOYGR-6: Contains 6 modules, providing approximately 462 MW of power, suitable for medium-sized power needs applications, such as small cities or larger Industrial facilities;

VOYGR-12: Consisting of 12 modules totaling approximately 924 MW of power output, this is NuScale’s largest capacity VOYGR layout and is suitable for urban and industrial applications that meet large-scale power needs , or even as Level grid baseload power, even in the event of a catastrophic loss, the VOYGR-12 can operate at 154 MW for 12 years without the use of new fuel.

Business layout: The company provides downstream customers with Full service support from license application, construction and commissioning to operation and maintenance. The services provided by the company can be divided into two categories: pre-commercial application (COD) and post-commercial application:

Pre-commercial application: startup and testing, ITAAC management (inspection, testing, analysis and acceptance criteria), COLA management (Joint License Application for VOYGR™ Power Plant);

Post-Commercial Application: Design Engineering Management, O&M Engineering Project Management, Requalification Training and Simulator Support, Procurement and Spares Management, Nuclear Fuel and Fuel Outage , system verification and validation.

Project progress: Already cooperated with many projects around the world Customers cooperate with SMR nuclear power projects. So far, the company has cooperated with RoPower Nuclear S.A. (Romania), KGHM Polska Miedź S.A. (Poland), Kozloduy Power Plant (Bulgaria), Standard PoWer (Ohio and Pennsylvania), Prodigy Marine Power Station (Canada), Indonesia Power (Indonesia), and GS Energy (South Korea) have project cooperation.

"Soft power": Focus on scientific research and cultivating talents, and open E2 nuclear energy exploration center laboratories in many universities around the world. In addition, the company has also set up an E2 Center (Energy Exploration Center) to provide users with practical opportunities to apply nuclear science and engineering principles through simulated real nuclear power plant operating scenarios. E2 has center points in multiple universities and regions around the world. Such as Texas College Station, Bucharest Polytechnic University (Romania), Seoul National University in South Korea, Oregon State University, etc.

Financial analysis: The company's financial situation is currently in a volatile stage, with abundant cash flow and no debt, and excellent results in cost reduction and efficiency increase. The company's latest third quarter report shows that in the third quarter of 2024:

Revenue: The company's operating income was US$500,000, compared with US$7 million in the same period last year. The decrease in revenue was mainly due to the termination of the CFPP contract ( 2023 On November 8, 2020, UAMPS and NuScale announced that both parties agreed to terminate the carbon-free power project CFPP);

Net profit: The company's net loss was US$45.5 million (of which US$7.2 million was related to the fair value of the issued warrants) non-cash expenses), the company had a net loss of US$58.3 million in the same period last year, and the net loss further narrowed;

Expenses: Operating expenses were US$41.2 million, compared with US$93.9 million in the same period last year. Operating expenses decreased by US$52.7 million year-on-year. The company further improved its ability to reduce costs and increase efficiency;

Cash: As of 2024 According to the third quarter report, cash, cash equivalents and short-term investments were US$160 million (of which US$5.1 million was restricted), and there was no debt.

Consortium background:

< p>

Fluor Corporation: A world-renowned engineering and construction company, it is a major shareholder and owns a large number of shares. Its investment in NuScale began in 2011, helping the company gain support in technology research and development and commercialization;

p>

U.S. Department of Energy (DOE): United States Provided a large amount of financial support (more than 300 million U.S. dollars) for NuScale's research and development through the Department of Energy to support the research and development and deployment of SMR technology;

Japanese trading company JGC Group;

Public and private equity Equity: 2021 NuScale Announces Partnership with Spring Valley Acquisition Corp. merger. Through this merger with SPAC (Special Purpose Acquisition Company), NuScale entered the public capital market, bringing NuScale approximately US$235 million.Yuan of funds;

South Korean company Doosan Heavy Industries: the world's leading heavy industry company, not only participates in investment, but also plans to provide some parts and manufacturing support for NuScale's reactor;

p>

5.1.2 OKLO (R&D Manufacturer)

Company Profile: The company is founded by Jacob DeWitte and Caroline Cochran (The founders all have nuclear energy engineering backgrounds) It was formally established in 2013, focusing on the development of small modular reactors (SMRs). It is headquartered in California. In 2014, OKLO entered the well-known start-up accelerator Y Combinator and received start-up funds. In September 24 OKLO received site authorization for a mini-reactor in Idaho in March and plans to deploy it in 2027. The company's Aurora microreactor uses metal fuel (different from other nuclear reactors that use uranium fuel), and currently the company mainly provides 24/7 clean energy for data centers, factories, industrial sites, communities and defense facilities.

Core product: The company's core product "Aurora Microreactor" has a single module power of 1.5 MW. The Aurora module is refueled every ten years (so the main expected downtime is power conversion system maintenance), the Aurora power plant provides power ranging from 15 MW to 50 MW. The power plant covers an area of ​​only a few acres, has low operating and maintenance costs, and the plant can be located where customers need power, avoiding expensive and Long power line transmission.

Competitive advantages (fuels are different from others):

Microreactors are better suited for distributed needs: OKLO’s Aurora microreactor It is a medium-sized SMR, and the power plant usually has a power of 50 At about MW, it has a competitive advantage in meeting medium-sized distributed, remote and independent power needs. In comparison, NuScale is larger and more suitable for grid-level energy solutions;

Fuel and cooling technology Cleaner, more environmentally friendly and cheaper: Oklo’s The Aurora reactor uses metal fuel instead of traditional light water reactor fuel, and its cooling system is also different from common water cooling, using liquid sodium as the coolant. On the one hand, such a fuel and cooling design can improve the thermal conductivity and efficiency of the reactor, and on the other hand, it can reduce the output of nuclear waste, thereby reducing the cost and environmental impact of nuclear waste treatment. In contrast, NuScale uses traditional light water as the cooling medium and uranium as the fuel, which is more suitable for existing nuclear power plant technology and supply chain.

Financial analysis: The company continues to expand investment in preparation for initial commercialization. The company’s current cash flow is relatively abundant. The company’s latest third quarter report shows that the third quarter of 2024:

Expenses: Operating expenses of $12.28 million, while it was US$4.66 million in the same period last year, the company continues to expand investment;

Net profit: The company's net loss was US$9.96 million, compared with US$8.67 million in the same period last year. The expansion of net profit loss was mainly caused by continued investment;

p>

Adequate cash: As of the third quarter report of 2024: total cash and marketable securities were US$290 million, including US$91.8 million in cash and cash equivalents, and US$197 million in marketable securities.

Capital background:

Sam Altman (founder of Open AI): one of Oklo’s main funders. In 2014, Altman included Oklo Y Combinator incubator. In 2024, Altman further helped Oklo successfully list on the New York Stock Exchange through a merger with his special purpose acquisition company (SPAC) AltC Acquisition Corp., raising approximately $306 million in funding to support its nuclear energy projects. Commercialization and future development;

Y Combinator: OKLO is Y Startups incubated by Combinator, early financing mainly came from YC's incubation projects, and received start-up capital support. After Oklo merged with AltC Acquisition Corp, Oklo went public with a pre-investment valuation of approximately US$850 million. Early backer Y Combinator may retain Indirect stake in Oklo, but has not announced additional investment in the post-IPO stage;

DCVC (Data Collective): a well-known venture capital firm that focuses on investments in technology and deep technology fields. It has provided financial support to OKLO to help its technology development and market expansion;

U.S. Department of Energy (DOE): DOE’s investment in OKLO's research and development provides funding for the commercialization of advanced fuel cycle and manufacturing technologies, and DOE's funded projects have played a key role in advancing the maturation and validation of OKLO's technology.

5.1.3 NNE (NANO, R&D and manufacturing + fuel processing)

Company profile: Nano Nuclear Energy’s main business covers The four pieces of SMR related content cover manufacturing, fuel, transportation and other aspects, aiming to create a diversified vertically integrated industrial chain. NNE is an American start-up. Its founder and chairman, Jay Jiang Yu, was an analyst at the investment banking department of Deutsche Bank. James Walker, its CEO and chief R&D nuclear physicist, worked on the project of the new Rolls-Royce nuclear chemical plant. Principal, the company is focused on developing small modular reactors and is committed to becoming a commercially focused, diversified and vertically integrated company spanning four business lines:

Micro Nuclear reactor technology development: NANONuclear's main products include the solid core battery reactor "ZEUS" and the low-pressure coolant reactor "ODIN";

Nuclear fuel manufacturing: Established a nuclear fuel subsidiary, HALEU Energy Fuel Inc. (HEF), to provide HALEU nuclear fuel (a Advanced nuclear fuel containing 5%-20% uranium 235), the fuel can be used by oneself or supplied from outside;

Nuclear fuel transportation: Establish a transportation subsidiary Advanced Fuel Transportation Inc. (AFT), provides HALEU nuclear fuel for small modular reactors, microreactor companies, laboratories, the military, U.S. Department of Energy projects, etc.;

Nuclear energy industry consulting services;

Others Subsidiaries: NNE has also established a space business subsidiary, NANO Nuclear Space Inc. (NNS), to explore the potential commercial applications of NNE micronuclear reactor technology in space.

Core products (manufacturing side): NNE microreactors can provide 1-20MW of thermal energy, among which the Zeus nuclear microreactor has a completely sealed core and relies on high conductivity moderation The Zeus reactor core and power conversion system are installed in a standard container for easy transportation, and Can run for 10 years. The ODIN nuclear reactor is the second advanced nuclear reactor (ANR) being developed by NNE. It uses a low-pressure coolant. The operating temperature of the reactor will be higher than that of traditional water-cooled reactors, which can maximize the use of natural convection of the coolant.

Core Products (fuel side): Subsidiary HALEU Energy is focused on developing and manufacturing high-purity low-enriched uranium, HALEU, for its reactors and other SMR and microreactor companies and has been selected for the U.S. Department of Energy’s new High-Purity Low-Enriched Uranium Alliance (HALEU Alliance to be established on December 7, 2022) Official founding member. HALEU is enriched uranium with a concentration of the fissile isotope U-235 ranging from 5% to 19.9% ​​of the fuel mass. Compared with traditional uranium fuel, HALEU has many advantages - the reactor does not need to be refueled frequently, reduces the amount of waste, can be used as the next generation fuel of existing reactors, and is more economical and safer. According to NNE data, nearly 600 metric tons of HALEU will be needed by 2030 to bring new reactors to market.

Company Finance: The company is currently in the process of project In the development stage, cash flow is relatively sufficient. In the second quarter of 2024:

Expenses: Operating expenses are 4.32 millionUS dollars, compared with US$2.7 million in the same period last year, investment expanded significantly

Net profit: The company's net loss was US$4.67 million, compared with US$2.7 million in the same period last year. The expansion of net profit losses was caused by continued investment

Cash: As of the end of the second quarter of 2024, cash and cash equivalents were US$13.79 million, mainly from listing financing in May 2024.

Capital background:

Citizens Financial Group Inc: Citizens Financial Group Inc holds shares in NNE through open market investments.

BlackRock: One of the world's largest asset management companies, BlackRock holds shares in NNE through public market investments.

5.1.4 LEU (Centrus Energy, fuel processing)

Company Profile: Centrus Energy is positioned as a supplier of nuclear fuel and services (located in the middle reaches of the industrial chain), focusing on providing high-purity low-enriched uranium and high-efficiency nuclear fuel solutions for the global nuclear energy market. Headquartered in the United States, the company's business covers the research and development of uranium enrichment services and related technologies. It is particularly at the forefront of the market in the field of advanced fuels (such as HALEU) and supports the commercialization of small modular reactors (SMRs) and next-generation nuclear energy projects. In the field of traditional nuclear energy, Centrus designs, manufactures and successfully operates gas centrifugal enrichment technology - American centrifuges, which have passed the test of the US Department of Energy. Currently, the company is gradually expanding from traditional nuclear fuel business to more advanced fuel business.

Main business: The company's main business has three categories, 1) Nuclear fuel supply: providing low-enriched uranium (LEU) and highly-enriched uranium (HALEU) to serve the nuclear energy and new reactor markets ; 2) Advanced manufacturing: Use high-precision engineering technology to manufacture complex components. Specific products include efficient equipment for the nuclear fuel cycle, ultra-high-precision mechanical components, and complex modules for nuclear energy and safety systems, etc., providing services for energy, national defense, and aerospace. industry to provide support; 3) Defense: for the U.S. Provide nuclear fuel technology and related services to ensure the safety of nuclear energy infrastructure.

Company Finance: The company's revenue maintained a growth trend, mainly driven by the HALEU operation contract signed with the Department of Energy (DOE), but the gross profit declined as the number of SWU sales decreased. dropped to a large extent. The company's latest third quarter report shows that in the third quarter of 2024:

Revenue: The company achieved revenue of US$57.7 million in the third quarter, compared with US$51.3 million in the same period last year. Revenue has increased steadily, mainly due to the company's revenue in 2022 The HALEU operation contract signed with the Department of Energy (DOE) is transitioning from the first phase to the second phase at the end of 2023 to bring revenue expansion;

Gross profit: The company’s gross profit is atTo $8.9 million and $11.3 million for the three months ended September 30, 2024 and 2023, respectively, the decrease for the three months ended September 30, 2024 was primarily attributable to a decrease in LEU segment gross profit, which was primarily due to The reduction in the number of SWUs sold leads to an increase in SWU unit costs;

Expenses: Operating expenses were US$16.5 million, compared with US$14.2 million in the same period last year. Operating expenses increased by US$2.3 million year-on-year, mainly due to an increase in sales and administrative expenses;

Net profit: The company had a net loss of US$5 million, Net income for the same period last year was US$8.2 million, mainly due to a decrease in gross profit.

Capital background:

U.S. Department of Energy (DOE): Centrus Energy has received strong support from the U.S. Department of Energy, especially in highly enriched and low-enriched uranium. (HALEU) production to facilitate the development and deployment of advanced nuclear fuels.

BlackRock: One of the world's largest asset management companies, BlackRock holds shares in Centrus through public market investments.

TerraPower: A nuclear energy company founded by Bill Gates, working with Centrus to develop advanced nuclear fuel for small modular reactors (SMRs).

X-Energy: An American nuclear energy company that cooperates with Centrus to develop fuel technology for high-temperature gas-cooled reactors.

5.1.5 UUUU (Energy Fuels, raw material mining)

Company Profile: Energy Fuels is a mining and energy company headquartered in the United States (located upstream in the industry chain, directly storing fuel resources). It was formally established in 2006 and listed on the New York Stock Exchange in 2013. It focuses on the production of natural uranium and thorium and is a leader in nuclear energy and advanced fuels ( Such as HALEU) important raw materials for technology, it is also involved in the separation and refining of rare earth elements (REE). The company has multiple production facilities in the United States and is an important player in the North American nuclear fuel supply chain.

Main business: The company's main business includes the mining and processing of natural uranium and thorium, providing key fuels for the nuclear energy industry, and refining rare earth oxides through its facilities for wind power generation , electronic equipment and other clean energy and high-tech applications. The company's business model revolves around the extraction, processing and sale of mineral resources and establishing long-term supply relationships with customers in the energy and technology sectors.

Corporate Finance: The company continues to promote various In the mining of rare earth elements, production has recently been lower than expected due to transportation problems. The company successfully completed the acquisition of Base Resources, which included the advanced, world-class Catulia titanium and zirconium project in Madagascar, ensuring the company's leading position in the titanium and zirconium mineral industry. The company's latestThe third quarterly report shows that in the third quarter of 2024:

Revenue: The company achieved revenue of US$4.04 million in the third quarter, compared with US$11 million in the same period last year, mainly due to the shift from Pinyon Plain mine to Whi TeMesaMill's ore transportation has been delayed, and the problem is expected to be resolved in the fourth quarter of 2024;

Net profit: The company's net loss attributable to the company in the third quarter was US$12.06 million, compared with net income of US$1 in the same period last year 0.56 million US dollars, mainly due to transaction and integration costs related to advancing the Donald project, costs related to the acquisition of Base Resources and recurring operating expenses;

Gross profit margin: The company's gross profit from the uranium mine business in the third quarter was US$2.15 million , with a gross profit margin of 54%;

Expenses: The company's operating expenses in the third quarter were US$14.11 million, compared with US$12.38 million in the same period last year. The increase was due to the increase in new project integration costs and M&A transaction costs.

Capital background:

BlackRock: a world-renowned asset management company that holds publicly traded shares of Energy Fuels;

Vanguard Group: another large asset management company , investing in Energy Fuels through the open market;

State Street Corporation: As a large financial services institution, State Street holds part of the shares of Energy Fuels.

5.1.6 Others

The small and micro nuclear power industry chain is huge. In addition to innovative companies focusing on SMR technology, traditional Nuclear power companies and power operators are also involved:

UEC (Uranium Energy): a company focused on the exploration, development and production of uranium resources in North America, mainly using in-situ leaching (In-Situ Recovery, ISR) technology, the mining method is lower cost and more environmentally friendly. The company currently operates two major ISR platforms: one is located in Texas, supported by the Hobson plant; the other is located in Wyoming, relying on Irigaray and Christensen Ranch (former Willow Creek project ) support, these platforms manage multiple uranium mining projects and have a high degree of production readiness. In addition, the company owns high-grade traditional uranium projects in Canada such as Henday Lake and Carswell.

CCJ (Cameco): Canada's uranium mining and supplier, focusing on upstream uranium mining and processing, is one of the world's largest uranium suppliers for nuclear fuel The market provides raw materials.

BWXT (BWX Technologies): Focus on nuclear reactor component manufacturing and nuclear energy technology, working with the largest companies such as SMR/OKLOThe difference is that BWXT is a large equipment supplier and technical service provider. It mainly provides nuclear reactor components, nuclear fuel, and defense-related nuclear technology for the military and commercial fields. Its customers include the United States (such as providing nuclear reactors for Navy nuclear submarines).

DUK (Duke Energy), CEG (Constellation Energy Group), EXC (Exelon Corporation), ETR (Entergy Corporation): A large integrated power company in the United States that operates traditional nuclear power plants and provides power services. Its core business includes power generation, transmission and distribution services. DUK focuses on the southeastern region and has a relatively balanced power generation portfolio, including natural gas, coal and renewable energy. ; With clean energy as its core, CEG operates the largest carbon-free nuclear power plant group in the United States, focusing on carbon emission reduction; EXC focuses on nuclear power generation and is the largest nuclear power operator in the United States, covering multiple states; ETR serves the southern United States, with nuclear power and natural gas Mainly power generation, focusing on high reliability power supply.

5.2 Competitive landscape and advantages

From the perspective of industry competition landscape, since SMR nuclear power is still in the early stages of development, competition The pattern has not yet stabilized, and each company’s market share The gap is not big. The main differences are reflected in the technical path, business model and market layout:

1) Differences in technical routes

Pressured Water Reactor (PWR) ) Dominance: Currently, NuScale The pressurized water reactor technology promoted by Power and other companies occupies the mainstream market because of its high technological maturity and clear regulatory approval path, making it easier to gain the trust of investors;

The rise of innovative technologies: such as X-energy's High Temperature Gas-cooled Reactor (HTGR) and Terrestrial Energy The Molten Salt Reactor (MSR) represents the next generation of innovative nuclear energy technology, providing higher efficiency and flexibility, but faces challenges such as long R&D cycles and complex supervision;

Key differences: traditional technology The robustness and potential breakthroughs of advanced technologies have formed a differentiated competitive landscape.

2) Differences in business models

Modularity and scalability: NuScale Power and other companies focus on modular design, making reactors easier to produce, transport and assemble, thus reducing costs Construction and operating costs;

Specific market positioning: Nano Nuclear Energy targets the small, high-efficiency reactor market in remote areas and military bases to provide more flexible power solutions.

3) Regional differences in market layout

United States: Thanks to support (such as the Inflation Reduction Act) and technological advantages, American companies (NuScale, X-energy) have First-mover advantages in technology leadership and capital acquisition;

Russia: Rosatom has achieved commercialization with RITM-200 and is a leader in floating nuclear power plants and the polar market;

: Promoted the successful connection of the Shidao Bay high-temperature gas-cooled reactor to the grid, providing energy security Provide support while demonstrating potential in export markets.

Looking to the future, technology maturity, cost competitiveness, support, and market positioning are several important factors that determine the success or failure of SMR participants:

p>

1) Technology maturity and safety

The core of the nuclear power industry is the safety and maturity of technology, which is the primary threshold for entering the market. NuScale has obtained design certification from the U.S. Nuclear Regulatory Commission (NRC) and is the first SMR company in the world to obtain this certification; important development direction of nuclear fuel.

2) Cost advantage

SMR needs to prove that it is superior to traditional nuclear power plants and other forms of energy in terms of full life cycle costs (construction, operation, decommissioning). Modular design is the key to reducing costs. key. NuScale, OKLO, etc. all reduce single stack costs through standardized manufacturing and mass production.

3) Financial support

Support and initial capital investment are important factors that determine whether the SMR project can be implemented. The United States supports the renaissance of nuclear energy through various incentives, including direct grants and tax incentives. In addition, companies can also raise funds through international cooperation, such as NuScale's agreements with Romania and Poland to promote global deployment.

4) Market positioning and application scenarios

Diversified applications are an important competitive advantage of SMR, covering power generation, industrial heat supply, seawater desalination, hydrogen production, etc. For example, NuScale's target market focuses on public power supply and industrial power markets, while Nano Nuclear targets remote areas and special-purpose markets, emphasizing miniaturization, mobility and rapid deployment capabilities.

5) Internationalization and customer resources

International market competition will become the key to the future, and companies need to prove the applicability of their technologies under different regulatory, geological, and economic conditions. At present, NuScale has signed supply agreements with multiple companies to seize the global market. With the support of the British local market, Rolls-Royce plans to expand to Europe.

Horizontal comparison of financial data shows that nuclear fuel mining companies located in the upper reaches of the industry chain, such as Energy Fuels and Centrus Energy, are progressing rapidly in commercialization, and will achieve positive revenue and profits in 2023, while micro companies located in the middle reaches of the industry chain Reactor manufacturing companies such as NuScale Power, OKLO, and NANO Nuclear Energy are generally still in the business model verification stage, and high R&D and production expenses have resulted in net profits.Negative, among which NuScale Power is making rapid progress. Through SMR projects in cooperation with many countries, it will achieve revenue of US$22.81 million in 2023. Specifically, the competitive advantages of each company include:

NuScale Power: 1. The company has its own nuclear power plant - VOYGR Plant Models, which is the first and only company to obtain U.S. nuclear power A small modular reactor designed and approved by the National Regulatory Commission (NRC); 2. The company has currently cooperated with many companies around the world customers to cooperate with SMR nuclear power projects;

OKLO: 1. OKLO’s Aurora microreactor is a medium-scale SMR and has competitive advantages in meeting medium-sized distributed, remote and independent power needs; 2. Oklo’s The Aurora reactor uses metal fuel instead of traditional light water reactor fuel, which is cleaner, more environmentally friendly and lower cost;

NANO Nuclear Energy: The business coverage is wide. The company's main business covers 4 SMR related contents, including manufacturing, fuel, transportation and other links, aiming to create a diversified vertically integrated industrial chain;

Energy Fuels: The production capacity is relatively large Large, the company has multiple production facilities in the United States and is an important participant in the North American nuclear fuel supply chain. It also recently successfully completed the acquisition of Base Resources The acquisitions, including the advanced world-class Catulia titanium and zirconium project in Madagascar, ensure the company's leading position in the titanium and zirconium mineral industry;

Centrus Energy: 1. Focusing on providing high-purity low-enriched uranium and efficient nuclear fuel solutions for the global nuclear energy market, it is especially at the forefront of the market in the field of advanced fuels (such as HALEU); 2. The company has received strong support from the U.S. Department of Energy, especially in the field of high-enriched and low-enriched uranium uranium (HALEU) production to facilitate the development and deployment of advanced nuclear fuels.

6. Investment advice

[From computing power to energy: why the energy infrastructure track is recommended now]

< p>Existence of expectation gap: The core logic of currently recommending energy tracks stems from the fact that the technology industry chain driven by AI is extending from computing power ecology to energy IT infrastructure, and the market currently has no idea of ​​the mid- to long-term value of this key link. Fully aware. At the same time, the market believes that China's power infrastructure is complete, AI accounts for a small proportion, and it is difficult to be flexible. However, we believe that the increase in global computing power has become a trend, and the upward trend in computing power consumption is inevitable. Our advantages in the field of IT infrastructure can better take advantage of this east wind to achieve overseas deployment.

1. From computing power to energy: the inevitable path driven by the industry chain

Under the accelerated development of the AI ​​industry, GPU and CPU to storage, communication, copper cable Various subdivisions, such as computing power, have become hot topics in the current market. However, behind these computing power ecology, there is a strong reliance on the continuous supply of energy and infrastructure. From downstream AI application scenarios (whether it is games, finance, or medical care, etc.) to upstream basic supporting facilities (including cooling, IDC, energy, etc.), every link is interconnected. However, the current market is in the context of fierce competition in computing power. Under the current situation, more focus is placed on the mid-stream and downstream links (hardware and applications), while the key role of infrastructure in the long-term sustainable development of computing power is ignored. At present, power and IT infrastructure have become the bottleneck of the North American computing power market. There is no doubt that computing power software and hardware has become a hot spot in the market. Looking forward to the next 3-5 years, to explore more computing power-related opportunities, it is even more necessary to study the upstream infrastructure links in advance, especially the energy link.

2. The market’s neglect of mid- and long-term infrastructure planning has created an investment window for the energy track

1) The contradiction between supply and demand has become increasingly prominent

The power supply in North America is in a tight balance, and the demand for AI computing power is increasing rapidly. It is expected that the installed power demand of global data centers will increase from 40GW to 140GW by 2029. This exponential increase in energy consumption has exposed the shortcomings of current infrastructure planning. The expansion of power infrastructure lags behind, and the production cycle of key equipment such as transformers restricts energy supply capabilities.

2) Long-term hidden dangers caused by short-term market behavior

The current capital market is paying attention to computing power-related tracks (such as GPU, storage, communications) The degree is extremely high, but insufficient attention is paid to the mid- to long-term layout of infrastructure such as cooling, IDC, and energy. These infrastructure links are the core of promoting the sustainable development of computing power ecology. Take liquid cooling as an example. When we released an in-depth report on the liquid cooling industry in early 2024, industry changes were not taken seriously in the capital market. Its market popularity gradually emerged after the demand for AI power consumption exploded. Similar logic is the same. for the energy racetrack, and it's just getting started.

3. Why is the energy track worth planning in advance?

Energy is the core element of competition in the next stage of the technology industry, but the market’s planning and understanding of energy are still in its infancy:

Strategic scarcity Resources: Sam Altman mentioned that the most important resources in the future are computing power and energy. The energy track is not only the support of computing power, but also the basis for achieving sustainable technological development.

Initial investment window: The current AI energy track is in its infancy, the investment valuation is relatively reasonable, and there is large room for upward demand revision in the future.

Driven by collaboration with technology: Take SMR nuclear power as an example. It has the characteristics of low carbon, environmental protection, and efficient power supply, and is highly consistent with the global carbon neutrality goal. Natural gas, as a transitional energy source, will alsoBenefit from the expansion needs of data centers in the short term.

To sum up, the logic of the energy track , similar to the logic we used when recommending the liquid cooling industry last year: in the early days of the industry’s outbreak, the market valuation was not low, but investment was in long-term growth, not short-term cheap prices. Energy is the next battle in technological competition. Just as liquid cooling has evolved from optional to mandatory, the AI ​​upstream infrastructure track is also moving from traditional industries to core technology supporting equipment. Seizing the opportunity for deployment is the key to winning in the future.

6.1 SMR Nuclear Power U.S. Stocks

SMR nuclear power can meet power supply needs with a single solution, so we have sorted out the upstream, midstream and downstream players in the industry chain:

6.2 Natural Gas + Multi-Energy US Stocks

In addition to SMR nuclear power, there are many ways to deal with energy challenges, including natural gas power generation, renewable energy (such as solar energy, wind energy), and energy storage systems , and the use of innovative technologies such as fuel cells, usually using a composite solution of natural gas + other energy sources, so we have sorted out the main players in each link:

Natural gas power generation: NextEra Energy (NYSE : NEE), Dominion Energy (NYSE: D), Cheniere Energy (NYSEAMERICAN: LNG), etc.

Renewable energy (solar and wind energy): First Solar (NASDAQ: FSLR), Enphase Energy (NASDAQ: ENPH), Brookfield Renewable Partners (NYSE: BEP), etc.

Energy storage technology (balancing the intermittency of renewable energy): Tesla (NASDAQ: TSLA), Fluence Energy (NASDAQ: FLNC), etc.

Fuel cells and distributed power generation (fuel cells are fueled by natural gas or hydrogen): Bloom Energy Corporation (NYSE: BE), Plug Power (NASDAQ: PLUG), etc.

Core energy efficiency technologies (data center cooling): Vertiv (NYSE: VRT), Schneider Electric, etc.

6.3 A-share related targets

7. Risk warning

1. Technology and regulatory risks.

SMR technology is still in the research and development and early deployment stages. Many designs have not yet received comprehensive regulatory approval. The development cycle is long and there are technical uncertainties, such as safety testing, material performance verification, etc. Any technical Failures or regulatory delays could significantly increase costs and impact the commercialization process.

2. High capital demand and financing pressure.

The development and deployment of SMR requires huge capital investment, including design, construction and approval costs. Many start-ups rely on external financing to maintain operations. Once the capital chain is broken, the project may be suspended. In addition, the investment return cycle is long and it takes decades to recoup the initial investment.

3. Market demand and competition risks.

Market acceptance and demand may be affected by energy sources, public attitudes and technological substitution (such as energy storage technology and green hydrogen). If market demand is insufficient or support is reduced, SMR companies may face profitability difficulties.

This article is excerpted from the report "Guosheng Communications丨A New Perspective on AI: From the Battle of Computing Power to the Battle of Energy" released by Guosheng Securities Research Institute on December 19, 2024 ——Communication Strategy: AI Infrastructure", please see the relevant report for details.

< /mp-style-type>