Author: Will Ogden Moore, Grayscale; Compiled by: Tao Zhu, Golden Finance

Abstract

Compared with traditional alternatives, blockchain technology can guide capital to critical physical infrastructure more effectively project. These projects are called Decentralized Physical Infrastructure Networks (DePINs), including wireless connectivity platforms, AI-related resources such as computing and data, and other applications.

DePIN assets represent non-speculative use cases of real-world adoption of cryptocurrencies, as the network utilizes tokens as economic incentives to facilitate the sharing of these resources.

Solana is currently the leading blockchain for DePIN because it has some DePINs ranked high by market capitalization, usage and adoption, including Helium, Render, and Grass. [1] Solana is the preferred DePIN blockchain for its speed, cost-effectiveness and reliability. Grayscale offers Solana investment opportunities for qualified qualified investors* through the Grayscale Solana Trust.

According to market capitalization (48%), AI-related DePIN dominates the topic. [2] These assets include functions such as decentralized AI model training, computing, data crawling and storage. Grayscale Research is particularly interested in emerging decentralized model training use cases. The Grayscale AI Fund provides many AI-related DePIN projects for qualified qualified investors, including Bittensor (TAO).

In January this year, President Trump and First Lady Melania Trump launched memecoin on the Solana chain, which has recently attracted the attention of mainstream media. Although Solana is currently the dominant blockchain for memecoin transactions [3], it is also the carrier of a key non-speculative use case in cryptocurrencies: the Decentralized Physical Infrastructure Network (DePIN).

What is DePIN?

DePIN represents a unique emerging field: connecting blockchain to the real world. These physical systems include hotspot devices, wireless routers, physical Internet of Things (IoT) sensors, car dashcams and GPUs to meet the needs of Internet connectivity and computing resources. These devices provide value through the network in exchange for crypto token rewards. Unlike the traditional model that relies on companies such as telecom providers and data centers, DePIN is decentralized, dispersing ownership and management among different independent actors. In the Grayscale Research framework, DePIN assets belong to the utility and service encryption department as in EtherApplication layer protocols running on smart contract blockchains such as Fang or Solana.

DePIN Investment Outlook

Because DePIN distributes physical nodes among different actors, this helps eliminate single point of failure and enhances the resilience, efficiency, and sustainability of the infrastructure. In this way, DePIN can promote innovation, promote equity and provide greater flexibility while reducing operating costs. In short, DePIN may have several key advantages compared to traditional centralized models:

eliminate the need for centralized intermediaries

reduce single point of failure and improve fast response demand The ability to eliminate management costs and sometimes reduce consumer prices

Coordinate incentives between all participants (including resource providers and consumers)

Due to these advantages, DePIN is particularly related to industries with high capital requirements, high entry barriers, monopoly dynamics and insufficient resource utilization. The DePIN model is particularly relevant to areas such as Internet connection, data storage, and computing hardware for AI development. These industries are worth trillions of dollars, including some of the world’s largest companies, from AT&T to Amazon (AWS) and Nvidia.

So we think the opportunity is great; if DePIN accounts for a small portion of these markets, the potential impact may be huge. Grayscale Research believes this result is feasible because 1) DePIN provides a more efficient business model than traditional alternatives, and 2) there are already examples of DePIN successfully leveraging encryption incentives to guide network provisioning at scale.

Today, we consider Solana the best way to get this early and growing thematic investment opportunity as DePIN’s preferred blockchain (more on that later). However, we also believe that several DePIN projects have demonstrated impressive levels of adoption and are becoming potential category winners.

DePIN Current Situation

DePIN assets have a wide range of use cases, including wireless connectivity and the Internet of Things (Helium), geospatial data collection (Hivemapper), decentralized video streaming (Render), and AI-related use cases. Such as computing network (Akash). Today, AI-related DePIN projects dominate market capitalization (Figure 1). In the following sections, we will highlight some of these emerging use cases and projects.

Figure 1: Artificial intelligence-related DePIN projects dominate by market value

Helium — Wireless Connectivity and the Internet of Things< p>Helium, headquartered in Solana, is DePIOne of the early pioneers of N and a classic example of real-world cryptocurrency use. Helium enables individuals to monetize their unused network capacity in exchange for Token Rewards (HNTs), creating a decentralized hotspot network. Users access internet connections through this hotspot network at much cheaper prices than traditional alternatives. Helium currently has over 129,000 mobile users and offers an unlimited traffic package of $30 per month — well below the US average of $144 per month. [4] It is worth noting that Helium recently announced its free mobile package with limited data, text messages and call time. [5]

Helium also works with telecom providers such as T-Mobile. Through this partnership, T-Mobile users can seamlessly access Helium’s wireless connections in high traffic areas, reducing congestion and cost. T-Mobile pays Helium network based on usage, which has a deflationary effect on the supply of HNT. Thanks in large part to the efforts of the enterprise, Helium has provided wireless internet access to approximately 450,000 people and achieved an increase in overall network usage in the second half of 2025 (Figure 2). [6]

In addition to mobile connectivity, Helium also supports IoT devices, allowing sensors to transfer data through their distributed networks at a lower cost than traditional Wi-Fi solutions. Similar DePIN projects focused on IoT solutions include Geodnet, Hivemapper, and NATIX, which use IoT sensors in your car or home to collect real-time geospatial data. This data is very valuable for today’s navigation services and can be used to improve the accuracy of artificial intelligence systems such as autonomous cars and robots. Geodnet, in particular, saw an astonishing increase in annualized network fee revenue in January 2025, reaching US$3 million (a year-on-year increase of approximately 518%). [7]

Grayscale Research believes that if Helium can 1) continue to expand its usage and geographical coverage in its current partnership with partners such as T-Mobile and Telefonica; 2) through its new free The mobile program has gained wider consumer recognition, then its opportunities may make sense.

Figure 2: Helium network usage will increase by the second half of 2024

Data crawling—Grass< p>DePIN has a great overlap with AI use cases; one example is data acquisition. Large language models (LLMs) require a large amount of high-quality data. OpenAI and GeExisting AI services such as mini continue to access real-time and multi-modal data through Bing and Google, respectively, thus having a significant advantage over competitors. Grass aims to level the playing field by enabling new data acquisition models powered by token incentives.

Through a simple desktop application, Grass allows anyone to share their unused internet bandwidth in exchange for GRASS tokens. This bandwidth is used to grab online data and then sell it to AI companies and developers for model training. Therefore, individuals on the Grass network can monetize their bandwidth, while AI companies can access data that would otherwise be trapped in silos.

Grass is one of DePIN success stories, leveraging token incentives to guide supply, expanding its contributor base to 2.5 million nodes in 190/region, more than any other DePIN project. [8]Grass has crawled 20% of YouTube’s data, totaling over 7000 TB (Figure 3). The Grass network currently provides data for AI labs such as Laion and Ontocord, and there could be a huge upside if it can work with some of the larger AI labs.

Competrators on the topic include Masa, which provides similar decentralized data crawling services to the Bittensor ecosystem.

Figure 3: Total data captured by Grass since November last year

Artificial Intelligence Model Training——Prime Intellect and other

Artificial intelligence companies can use this data to train LLM (large language models). One of the earliest use cases for DePIN (but probably still one of the most promising ones) is decentralized training.

Artificial intelligence model training requires huge computing power, usually concentrated in centralized data centers. DePIN is changing this situation. By allocating workloads on global GPU networks, these projects unlock decentralized training. Last November, Prime Intellect successfully trained a 10 billion parameter language model on 1 trillion tokens on five and three continents. [9] Prime Intellect achieves this using their DiLoCo framework, which enables efficient training across loosely connected devices by significantly reducing communication requirements. [10] Competitors include Nous Research, which built the Hermes open source model and recently chose Solana as its blockchain.Another emerging player is Flock.io (FLOCK), which enables financial companies such as Animoca Brands and GSR, as well as healthcare companies to train AI models in a privacy-protected manner. While Prime Intellect and Nous Research have not yet launched tokens, Grayscale Research believes some of these decentralized training protocols can reduce costs and improve access to AI developers.

Artificial Intelligence Model Inference — Akash

Once the LLM is trained, it can be queried by consumer applications such as chatbots (e.g. ChatGPT). Each time a user asks ChatGPT, inference costs associated with generating output are incurred. This inference process requires complex GPU hardware (such as the NVIDIA H100) to perform the necessary calculations. These highly demanding resources are often sold at high prices and long-term contracts, which makes it difficult for startups, researchers or organizations without large amounts of money to afford. Meanwhile, many data centers with these GPUs have up to 30% underutilized capacity. [11]

Akash is a decentralized cloud computing GPU rental market. It enables GPU vendors to monetize idle supplies and enables consumers to access high-performance GPUs in flexible time periods, at a cost that is often lower than other options.

Akash’s GPU rental number increased from less than 200 before November 2024 to over 600 in January 2025, including 398 NVIDIA H100s. [12]As a result, Akash increased its network expense revenue to an annualized $4.6 million in January this year (Figure 4). [13] brev.dev (acquired by Nvidia), Venice.ai, ElizaOS (formerly ai16z) and UT Austin all use Akash GPUs. [14] Although Akash and similar decentralized inference platforms may not compete directly with major cloud providers, they have the opportunity to seize market share when serving startups, decentralized AI projects, independent researchers, and universities. . Akash competitors that offer similar services include io.net, Nosana, and Hyperbolic.

Figure 4: Akash expense revenue and GPU rental growth accelerated in November 2024

Currently the largest market value The DePIN asset is Bittensor (Figure 5), which is an AI we have introduced in depth before.Related platforms. As a platform for the "AI Internet", it hosts a subnet dedicated to performing tasks such as data crawling and model inference computing (two use cases mentioned earlier). Each subnet acts as a separate entity in the broader Bittensor ecosystem, which operates similarly to an interconnected DePIN network. On February 13, Bittensor underwent a major upgrade called “Dynamic TAO” that made a single subnet investable for the first time.

Figure 5: Example of DePIN by category, blockchain and market capitalization

Solana as the leading DePIN block Chain

While many DePIN projects choose to build their own standalone layer 1 blockchain (“application chain”), Solana has become the main blockchain for DePIN applications. By market capitalization, Solana is the leading DePIN blockchain and the preferred blockchain for many top DePIN projects (Figure 6). A notable example is Helium, which was initially built on its own tier 1 and then moved to Solana. Similarly, video transcoding network Render and one of the larger DePIN assets by market capitalization migrated from Ethereum to Solana in 2024. Other DePIN projects built on Solana include Grass, Hivemapper, io.net, Nosana, and more.

Grayscale Research believes that Solana is the preferred DePIN chain for several reasons. First, Solana is a fast and low-cost chain that allows DePIN projects to transfer value and operate for suppliers and consumers in an efficient way. Second, Solana provides the largest network security, liquidity, and history in high-throughput and low-cost chains such as BASE and SUI. Finally, by building on Solana instead of a standalone Tier 1 application chain, DePIN project developers can focus on their core competitiveness at the application layer without allocating resources for scaling and infrastructure-level tools and maintenance.

Figure 6: Solana is the leading blockchain for DePIN by market value

Conclusion

DePIN may represent A promising frontier of blockchain technology, providing innovative solutions for global resource allocation. While some may claim that the crypto industry is limited to speculative or financial use cases, many DePIN projects provide compelling evidence to the contrary.

Most DePIN projects are less than five years old, but some have shown encouraging signs of adoption, including working with and serving some of the world's largest organizations (for example, through brev.dev and T -Mobile and Nvidia collaborate). Therefore, we are particularly excited about certain use cases, including internet connectivity, decentralized training, and networks that collect social data or real-world physical video data for AI training. Additionally, the potential future growth of DePIN may create incremental tailwinds for Solana. As the leading DePIN blockchain, Solana offers the speed, cost-effectiveness and security that these projects need to thrive.

Ultimately, we believe that DePIN may represent a compelling investment opportunity. Investors interested in early-stage, high-betaDePIN investments can consider Helium or Bittensor, for wireless/Internet of Things and AI investments, respectively. For those who prefer to hold large-cap options, we believe Solana is the preferred investment option to grow with DePIN.

Noun Index

Internet of Things (IoT): Internet of Things (IoT) refers to interconnected physical devices, vehicles, appliances and other objects that are embedded with sensors, software and connectivity functions. networks that enable them to collect and exchange data over the Internet.

DePIN: A decentralized physical infrastructure network that uses blockchain and token rewards to incentivize individuals to contribute physical infrastructure resources.

Blockchain: A distributed digital ledger that securely stores records across computer networks in a transparent and tamper-proof way.

Helium: A blockchain network designed to promote wireless communications of IoT devices around the world using a unique coverage proof consensus mechanism.

Render: A practical token native to the Render Network that allows users to share GPU capabilities to render visual projects.

Grass: A protocol on the Solana blockchain that allows users to share unused Internet bandwidth for network data collection, supporting AI development.

Solana: An encrypted computing platform designed to achieve high transaction speeds without sacrificing decentralization, using innovative approaches such as the “historical proof” mechanism.

Decentralized AI: Integrating AI algorithms and processes in decentralized networks, often using blockchain technology to promote transparency, security and collaboration

Data crawling : A technique in which a computer program extracts data from the readable output of another program.

Bittensor: A platform that connects artificial intelligence and blockchain, allowing users to create or participate in competition-based markets (called subnets)

Memecoin: A platform that connects artificial intelligence and blockchain to blockchain, allowing users to create or participate in competition-based markets (called subnets)

Memecoin: A platform that connects artificial intelligence and blockchain to blockchain, allows users to create or participate in competition-based markets (called subnets)

Cryptocurrencies created for comedy, usually based on memes or humor rather than serious technology development.

Ethereum: A decentralized blockchain platform with smart contract capabilities, using Ethereum (ETH) as its native cryptocurrency.

Data storage: In the context of a decentralized network, data is encrypted and stored in systems on multiple locations or nodes run by individuals or organizations that share additional disk space.

Computing hardware: The physical components and devices that make up a computer system, enabling it to perform various tasks and run software.

Geospatial data collection: The physical components and devices that make up computer systems enable them to perform various tasks and run software.

Decentralized video streaming: A video platform that uses blockchain technology to distribute content storage and streaming between multiple nodes in a peer network, reducing censorship risks and enhancing privacy.

Akash: A decentralized cloud computing market that centralizes and redistributes idle computing processing power to customers who need it, using AKT as its native utility and governance token.

Market Value: The total value of a cryptocurrency is calculated by multiplying the total number of coins in circulation by the current price of a single coin.

LLM: Large language model, an advanced artificial intelligence system, trained by a large amount of text data, can understand and generate human-like text.

Token Rewards: Provide rewards to participants in the blockchain network in the form of cryptocurrency tokens to recognize their contributions or activities.

Comments

[1]Artemis, Grayscale Investments, data as of Feb 18, 2025.

[2]Artemis, Grayscale Investments, data as of Feb 18, 2025.

[3]Artemis, as of Feb 18, 2025

[4]Blockworks

[5]Blockworks

[6]Messari

[7]Artemis, data as of Feb 18, 2025.

[8]Messari

[9]Prime Intellect

[10]Prime Intellect

[11]CBRE

[12]Akash Network

[13]Artemis

[14]Akash Network