Bitcoin mining firms transition to AI infrastructure providers

Written by: Xiao Bing

In Abilene, Texas, eight H-shaped data centers are gradually coming online on a site spanning approximately 1,000 acres.

This is the first large-scale campus of OpenAI's Stargate project. The entire project is planned with a capacity of 1.2 gigawatts, and the first two buildings are already operational, with the remaining buildings under construction.

The park is operated by Oracle, and the developer behind it is Crusoe, which originally focused on Bitcoin mining.

Founder Chase Lochmiller was formerly a partner at the crypto fund Polychain Capital.

In 2018, he and his childhood friend Cully Cavness discovered that U.S. oil fields were flaring large amounts of stranded associated natural gas every day. The two moved power generation equipment and mining rigs to the wellheads to mine Bitcoin using the gas that would otherwise have been wasted.

The business logic is simple: find electricity in remote areas and convert it into computing power in a very short time.

Seven years later, the customer switched from the Bitcoin network to OpenAI. In 2025, Crusoe sold its Bitcoin business, comprising over 425 modular data centers, to NYDIG to focus exclusively on AI.

Crusoe’s transformation may seem vast, but its core capabilities have remained unchanged: finding electricity, building data centers, and operating and maintaining them.

Like Crusoe, many companies and individuals have made a successful transition from Crypto to AI; unlike the PayPal Mafia, they did not come from the same company or form a unified organization—what unites them are three types of assets left over from the previous Crypto cycle:

• Electricity, land, and grid connection permits controlled by mining companies;
• Engineers and entrepreneurs trained by crypto companies;
• Capital accumulated from the previous bull market.

After 2022, these three categories of assets began flowing into AI simultaneously.

The greatest contradiction facing humanity's future is the growing demand for data processing versus limited computing power.

In 2019, Wu Jihann, founder of Bitmain, wrote in the article "The Beauty of Computational Power" that this was precisely why Bitmain invested in AI chips.

At the time, this statement seemed like public relations rhetoric; six years later, it looks more like a prophecy written in advance.

In February 2026, Wu Jihhan's mining company Bitdeer announced it had liquidated its entire Bitcoin inventory to generate liquidity for building AI data centers, making a resolute shift from crypto to AI.

The transformation of Bitcoin mining companies into AI is often misunderstood by the public as "converting mining rigs into AI servers," but this is not the case.

Bitcoin mining rigs are mostly ASIC chips designed to perform only specific hashing algorithms and cannot be used to train large models. Even GPUs left over from Ethereum mining farms struggle to meet today’s large AI clusters’ requirements for networking, video memory, liquid cooling, and reliability.

The most valuable asset of mining companies is the data center already connected to the power grid.

Building an AI data center is often less about purchasing GPUs and more about securing hundreds of megawatts of reliable power, along with land, substations, transmission lines, and construction permits. This process can take years, and mining companies have already completed significant upfront work in North America, Northern Europe, and the Middle East to reduce mining costs and ensure compliance.

As Bitcoin mining profits decline and AI companies are willing to sign long-term high-price contracts, mining companies naturally begin switching clients.

CoreWeave was among the first to complete the transition.

In 2016, three commodities traders placed a GPU on a pool table in their Manhattan office and began mining Ethereum. When the crypto winter arrived, they took advantage of falling prices to acquire large quantities of used graphics cards, later expanding their business into film rendering and machine learning.

The company was originally called Atlantic Crypto and later renamed CoreWeave. Its filing shows that until 2022, the majority of its revenue still came from cryptocurrency mining; after that, the crypto business was completely discontinued.

Today, CoreWeave is a leading AI cloud company backed by NVIDIA, and its model is being replicated across the entire crypto mining industry.

In 2026, TeraWulf signed a 20-year data center lease agreement with Anthropic for approximately 401 megawatts, with an initial contract value of about $19 billion;

Cipher Mining signs a 15-year, 300-megawatt, approximately $5.5 billion agreement with AWS;

Core Scientific will provide a large amount of data center capacity to CoreWeave on a long-term basis.

Hut 8 has signed two 15-year leases at the Beacon Point campus in Texas, each with an initial contract value of approximately $9.8 billion.

IREN disclosed additional contracts totaling $2.8 billion in July 2026, following its $9.7 billion cloud services agreement with Microsoft.

According to CoinShares, as of Q1 2026, publicly traded mining companies have announced over $70 billion in AI and high-performance computing contracts. Meanwhile, the revenue per unit of hash rate for Bitcoin mining dropped to approximately $30 to $35 per PH/s per day, pushing mines using outdated equipment or facing higher electricity costs close to profitability thresholds.

Mining companies have transformed from computing facilities of the crypto era into the infrastructure of the AI era, remaining at the forefront.

Beyond mining farms, people in the crypto industry are also moving toward AI.

Alex Atallah is the co-founder and former CTO of OpenSea. At the height of the NFT boom, OpenSea’s monthly trading volume once exceeded $4 billion. In July 2022, Atallah left the company to prepare for his next venture.

In 2023, he founded OpenRouter.

OpenRouter solves a straightforward problem: with an increasing number of large models, each differing in price, speed, and capability, developers don’t want to integrate APIs individually for every model provider. With OpenRouter, they only need to connect to a single interface to access hundreds of models, with requests automatically routed based on price, performance, and availability.

In 2025, OpenRouter completed a total of $40 million in funding, with a valuation of approximately $500 million.

In May 2026, it completed a $113 million Series B round led by CapitalG, raising its valuation to approximately $1.3 billion. Over the past six months, the platform’s weekly token volume has grown from $5 trillion to $25 trillion.

OpenRouter and OpenSea do not do exactly the same things, but their business structures are quite similar.

OpenSea brings together NFT buyers and sellers, while OpenRouter aggregates models, compute providers, and developers. The former facilitates digital asset trading; the latter matches inference requests. The product has changed, but the ability to build markets and integrate fragmented supply remains the same.

Some crypto traces remain directly embedded in the product. On the OpenRouter registration page, MetaMask is still displayed alongside Google and GitHub login options, and the platform accepts USDC payments.

Fal.ai is another example.

Founder Burkay Gur previously worked at Coinbase building machine learning platforms and founded the company in 2021, initially developing machine learning data pipelines and deployment tools.

After Stable Diffusion was open-sourced, they found that while image and video models were becoming increasingly numerous, they suffered from slow inference speeds, difficult deployment, and low GPU utilization. As a result, Fal.ai shifted its focus to generative media inference.

This choice was quickly rewarded.

By mid-2025, Fal.ai's annual revenue had approached $950 million. In December of the same year, the company completed a $140 million Series D funding round led by Sequoia Capital, achieving a $4.5 billion valuation. Companies such as Adobe, Canva, and Perplexity are using its generative media infrastructure.

Mining companies provide electricity and data centers to AI, while capital accumulated during crypto cycles enters AI in another way.

The most direct example is Jed McCaleb.

He created the cryptocurrency exchange Mt. Gox, later co-founded Ripple and Stellar, and became one of the earliest billionaires in the crypto industry.

In 2023, McCaleb’s Navigation Fund invested approximately $500 million to purchase 24,000 NVIDIA H100 GPUs in a single transaction and established Voltage Park to rent GPUs to AI companies and research institutions.

Instead of creating another public blockchain, he converted the money he made from Crypto into the most scarce asset in the AI industry.

In 2026, Voltage Park merged with the AI development platform Lightning AI, with the combined entity valued at approximately $2.5 billion. The wealth accumulated during the previous crypto cycle thus became the balance sheet of an AI cloud company.

The investment portfolio left behind by SBF, the founder of the collapsed cryptocurrency exchange FTX, provides a more dramatic case.

In 2022, SBF invested $500 million in Anthropic, then relatively unknown, acquiring approximately 13.5% ownership. After FTX’s collapse, the liquidation team sold these shares in batches in 2024, recovering about $1.3 billion. Today, Anthropic’s post-money valuation has reached $965 billion. Had FTX not sold its shares, its stake would likely still be around 6.7%, worth approximately $65 billion—roughly 130 times the original $500 million investment.

The story of Cursor is even more extreme.

In April 2022, Alameda, a fund under SBF, participated in Anysphere’s early-stage financing with $200,000; the company later launched the AI programming tool Cursor. After FTX entered bankruptcy proceedings, the liquidation team sold this stake in April 2023 for $200,000—nearly breaking even.

In June 2026, SpaceX announced the acquisition of Anysphere in a $60 billion all-stock transaction; according to public reports, Alameda initially received approximately a 5% stake. Ignoring any subsequent dilution from Anysphere’s later financings, this stake would have a paper value of $3 billion, representing a 15,000-fold return on the original $200,000 investment.

This should not be simplistically attributed to SBF being an investment genius; a more accurate understanding is that, prior to the release of ChatGPT, the most aggressive and risk-tolerant capital in the crypto bull market had already begun seeking AI projects.

During a crypto market boom, substantial capital believed in two assumptions: computing power would become increasingly valuable, and software networks could scale globally in a short time. AI恰好符合这两个条件。

Therefore, as crypto capital enters AI, it is not only purchasing GPUs but also funding new technologies and organizational experiments.

Nous Research is a prime example.

Hermes Agent, developed by Nous, is an open-source AI agent capable of accumulating long-term memory and automatically generating skills. According to OpenRouter statistics, Hermes Agent has the highest global token usage, surpassing Claude Code.

In 2025, crypto investment firm Paradigm led the $50 million Series A funding round for Nous Research.

According to reports at the time, this funding round valued the company at approximately $1 billion based on its unissued tokens, with previous investors including crypto VC Distributed Global and former Coinbase CTO Balaji Srinivasan.

In addition to Hermes, Nous is developing Psyche, a distributed model training network built on Solana.

Traditional AI labs need to concentrate large numbers of GPUs in a single data center. Psyche wants to test an alternative approach: connecting GPUs distributed across different regions and owned by various participants to jointly train models, while using smart contracts to coordinate training progress, verify participants, and distribute rewards.

At this stage, Psyche remains an experiment, and the testnet tokens are explicitly labeled by the official team as having no economic value, but they represent another impact of crypto capital entering the AI space.

OpenAI also seriously considered a similar direction in its early days.

OpenAI was established in 2015 as a nonprofit organization, but the funding required for advanced models quickly exceeded what the donation model could support. By the end of 2017, Sam Altman and Greg Brockman had begun discussing new funding structures, one of which was token issuance.

Later-released internal emails showed that the team seriously considered issuing a token in early 2018. Musk explicitly opposed it, arguing that a token launch would severely damage OpenAI’s reputation. OpenAI later added that by the end of January that year, the team themselves had gradually lost interest in the idea.

OpenAI ultimately chose to form a for-profit entity and received substantial investment from Microsoft, but Sam Altman did not leave Crypto.

In 2019, he co-founded Worldcoin with Alex Blania and Max Novendstern, a project that uses the iris recognition device Orb to verify that a user is a real and unique human, establishing an identity and payment network through World ID and WLD tokens.

From Crusoe and CoreWeave to OpenRouter, Fal.ai, and Nous Research, these anecdotal stories, tainted by survivorship bias, do not mean that crypto companies have a higher chance of successfully transitioning into AI.

Mining companies left behind electricity, land, and grid connection permits; exchanges and Web3 companies cultivated a cohort of engineers familiar with distributed systems, GPU scheduling, and globalized products; and the wealth generated by token price increases became capital for purchasing GPUs, investing in model companies, and funding technological experiments.

Crypto didn't magically turn into AI; it simply redirected the resources left over from the previous cycle to the next industry that needed them more.