After Trump won the US presidential election, the market expected that he would relax the regulation of cryptocurrency. Trump once It said that Bitcoin mining will be concentrated in the United States, but this policy may have an impact on the AI ​​industry. In the early morning of November 23, the price of Bitcoin once reached $99,660, setting a new historical high again and heading towards the $100,000 mark.

Because both Bitcoin mining and AI training require a lot of energy and computing power. The simultaneous development of the two is bound to create competition in power and hardware resources. This means that the AI ​​training business may be affected by Bitcoin price fluctuations, especially as miners compete for limited hardware resources. In other words, the rise in the price of Bitcoin may increase the cost of AI training.

With the great success of ChatGPT , AI companies are racing to train and run their own models, hoping to surpass OpenAI’s flagship product. This creates a lot of demand: the reasoning process of the AI ​​model is much more complex than the indexing and retrieval process of a search engine, and a ChatGPT query consumes about ten times as much energy as a Google search.

This makes AI companies urgently seek cheap energy and large tracts of land to accommodate related equipment. In North America, some regions have implemented queuing systems waiting for large data centers to be connected to the grid. But once a company gets initial approval, building a data center from scratch can take years, millions of dollars, and go through a lengthy regulatory and administrative approval process.

Internationally, large-scale Bitcoin mining has historically been an extremely profitable business. However, it is also subject to the volatile cryptocurrency market. In the wake of the 2022 cryptocurrency market crash, many miners were forced into bankruptcy or shut down their operations entirely.

In 2023 and early 2024, mining companies that survived the market downturn reaped profits. However, this year’s Bitcoin halving, in which miners’ rewards are cut in half, did not trigger a sharp rise in Bitcoin prices to offset the reduction in rewards, as has been the case in previous cryptocurrency cycles. Since April this year, the long-term sideways movement in Bitcoin prices has squeezed miners’ profit margins, forcing some miners to seek to diversify their business models to hedge against the risk of cryptocurrency price fluctuations.

Four years ago, when data center and Bitcoin mining company IREN considered getting involved in AI training, they believed that from a business perspective, the volume of business at the time was not enough. But now, more and more large Bitcoin mining companies have begun to replace some of their mining equipment with equipment used to run and train AI systems. These companies believe that providing computing power to AI companies may provide a safer and more stable source of income than mining.

Today, artificial intelligence and bitsCooperation in the currency mining industry is natural. The needs of both parties are consistent. AI companies need existing sites for Bitcoin miners and access to cheap energy and infrastructure. Bitcoin miners, on the other hand, are pursuing the stable income brought by the AI ​​business and the potential profits brought by the current AI boom.

Some Bitcoin mining companies choose to lease their premises to AI customers. In June this year, Bitcoin mining company Core Scientific, which is on the verge of bankruptcy in 2022, announced that it would host more than 200 megawatts of GPUs for AI startup CoreWeave. Core Scientific said AI companies have begun buying mining farms at prices above the mining market, calling Bitcoin mining facilities the “power shell of the data center industry.”

Other Bitcoin mining companies operate GPUs themselves. Bitcoin mining company Hut 8 received $150 million in investment from Coatue Management to build artificial intelligence infrastructure. In some facilities of Australia-based mining company IREN, GPUs for AI and ASIC equipment for Bitcoin mining share the same space. Bitcoin offers instant income, but is highly volatile. AI relies on customers. Once it has customers, it will be more stable. Nasdaq-listed company BitDeer is also building its own AI data center in Singapore.

However, it can be realized There are only a few overseas mining companies that have made this transformation. Secondly, the equipment used to mine Bitcoin is called an ASIC, short for Application Specific Integrated Circuit, with “application specific” meaning it cannot be used for other tasks. Mining companies cannot "seamlessly connect" mining equipment to AI scenarios.

An AI Infra industry practitioner told Silicon Star, “For example, H100 is generally used for training models, while 4090 is used for mining.”

In other words, if you want To serve the AI ​​industry, Bitcoin miners must purchase brand new equipment, and the data center requirements for artificial intelligence and Bitcoin mining are different. It is already difficult to enter a new and highly complex industry, not to mention having to compete with Google and Amazon, Microsoft and other well-funded technology giants compete.

So, not all mining companies can replicate the high-dollar cooperation between Core Scientific and CoreWeave. Especially smaller miners, they actually don’t have much to offer the AI ​​industry.

Cryptocurrency mining has been banned in China, and there are no miners converting to AI. However, there are also companies in other industries who want to get a share of the AI ​​wave. They either quit directly or form computing power subsidiaries. Started a "computing power leasing" business. According to statistics, there are more than 100 listed companies with the concept of A-share computing power leasing, including the "Lottery Printing King" Hongbo Stockportion, "MSG King" Lotus Holdings and so on. On the video platform, there is even content such as "I demolished a house in my hometown, bought 800 graphics cards, and joined forces with my old junior high school classmates to rent computing power."

Ideally, the business model of computing power leasing only requires an initial investment in GPU server equipment, hosting the hardware in a professional intelligent computing center, and then leasing the computing power to end users for hardware operation and maintenance and software services. All are handled by the Intelligent Computing Center.

But in reality, this may not be a good business. The demand for computing power rental comes from the development of the AI ​​large model industry, and the rental cost of high-end hardware for training AI is declining sharply. Featherless.Ai CEO Eugene Cheah wrote an article stating that the overseas rental price of H100 was once as high as US$8 per hour, but has now fallen to less than US$2 per hour. Mainly because some companies signed computing power leasing contracts in the early stage and began to resell reserved computing resources in order to avoid wasting idle capacity. Most of the market chose open source models, resulting in reduced demand for new models.

The domestic computing power rental market is also experiencing a similar phenomenon of "excess computing power", but "the rental market is unlikely to reduce prices because they are all bought at a high price in the early stage," a practitioner in the intelligent computing industry told Silicon Valley stars.

There is a saying in the currency circle The famous saying "computing is power", computing power is power, has now spread to the AI ​​circle.

Behind computing power is energy. There is a close correlation between developed countries and high energy consumption. If we calculate the per capita The power generation (kWh) can be seen in the chart for comparison. In other words, access to energy surplus is a necessary condition for civilization to progress. After all, on top of basic subsistence levels such as agriculture are superimposed new layers such as manufacturing, transportation, public services, urbanization and computing, all of which require energy support.

In this dimension, the infrastructure originally established to serve cryptocurrency is providing solutions to the computing power needs of the AI ​​era. This is undoubtedly an opportunity for the overseas digital currency mining industry, which has always wanted to get rid of speculation, to prove its value. As long as this craze continues, leading manufacturers will benefit from the enthusiasm and liquidity brought by AI.

Every wave of technological innovation is always accompanied by a "gold nugget fever". For the speculators among them, what they pursue is always profit itself. It doesn’t seem to matter whether the subject matter is digital currency, artificial intelligence, or tulips from three hundred years ago.

After the Bitcoin halving, some miners are faced with a dilemma: either continue to mine and hoard coins and expect the price of Bitcoin to rise; Waves of fast money. Now that the currency price has reached a record high, some people have begun to sigh: "It's still this fast." Dan coinsThere is also a famous saying in the circle, "It is harder to keep money than to be widowed."

When people sway back and forth between the currency circle and the AI ​​​​circle, this repeated process reminds people of what Keynes said: Speculators do not care who is the most beautiful, only who cares about who will be in the beauty contest. is selected.

And this "beauty contest" will continue endlessly.