On a roughly 1,000-acre construction site in Abilene, Texas, eight H-shaped data centers are being brought online in phases. The site is the first large campus under OpenAI’s Stargate plan. Total planned capacity is 1.2 gigawatts. The first two buildings are already online, and the rest are still under construction. Oracle operates the campus. The developer behind it is Crusoe, a company that started out in Bitcoin mining.
Crusoe co-founder Chase Lochmiller was once a partner at crypto fund Polychain Capital. In 2018, he and childhood friend Cully Cavness saw a business in stranded natural gas that was being flared at U.S. oil fields because it could not be transported. They brought power equipment and mining machines to the wellhead and used gas that would otherwise be wasted to mine Bitcoin. Seven years later, the customer base had changed from the Bitcoin network to OpenAI. In 2025, Crusoe sold its Bitcoin business, including more than 425 modular data centers, to NYDIG and focused on AI.
The article’s argument is that crypto did not magically become AI. What happened instead was a transfer of resources built during the last crypto cycle into a sector that now needs them more. Those resources fall into three groups: power, land and grid interconnection permits controlled by miners; engineers and founders trained inside crypto companies; and capital accumulated during the last bull market. According to the piece, all three began flowing into AI after 2022.
What miners are really selling to AI is electricity, not mining rigs
In 2019, Bitmain co-founder Jihan Wu wrote in an essay titled The Beauty of Computing Power that “the biggest contradiction in humanity’s future is between growing data-processing demand and limited computing power.” He said directly that this was why Bitmain was investing in AI chips. PANews says the line looked like public relations language at the time, but reads more like an early forecast six years later.
In February 2026, Bitdeer, the mining company under Wu, said it had cleared out all of its Bitcoin holdings to provide liquidity for AI data center construction. The move signaled a firm pivot from crypto to AI.
The article pushes back on the common idea that a miner’s AI transition simply means converting mining machines into AI servers. Most Bitcoin mining machines use ASIC chips designed for specific hashing algorithms and cannot train large AI models. Even GPUs left behind by Ethereum mining sites are hard to plug directly into today’s large AI clusters, which have demanding requirements on networking, memory, liquid cooling and reliability.
What miners do have, the article says, is data center infrastructure already connected to the grid. For AI data centers, the hard part is often not buying GPUs. It is securing hundreds of megawatts of stable power and obtaining land, substations, transmission lines and construction permits. That process can take years. Miners had already done much of that groundwork in North America, Northern Europe and the Middle East as they chased lower costs and regulatory compliance.
Once Bitcoin mining economics weakened and AI companies proved willing to sign long-term contracts at high prices, many miners started switching customers. CoreWeave is presented as one of the earliest examples. In 2016, three commodities traders put a GPU on a pool table in a Manhattan office and began mining Ethereum. During the crypto winter, they bought large volumes of second-hand graphics cards at depressed prices, then expanded into rendering for film and television and into machine learning. The company was first called Atlantic Crypto and later renamed CoreWeave. Its IPO filing showed that before 2022, most of its revenue still came from crypto mining; after that, the crypto business was fully shut down. Today CoreWeave is one of the leading AI cloud companies backed by Nvidia.
PANews lists a string of large, long-term agreements signed by mining-related operators and AI customers:
- In 2026, TeraWulf signed a roughly 401-megawatt, 20-year data center lease with Anthropic, with an initial contract value of about $19 billion.
- Cipher Mining signed a 300-megawatt, roughly $5.5 billion, 15-year agreement with Amazon Web Services.
- Core Scientific committed a large block of data center capacity to CoreWeave on a long-term basis.
- Hut 8 signed two separate 15-year leases at its Beacon Point campus in Texas, each with a base contract value of about $9.8 billion.
- After reaching a $9.7 billion cloud services agreement with Microsoft, IREN disclosed another $2.8 billion in new contracts in July 2026.
CoinShares estimated that, as of the first quarter of 2026, publicly listed miners had announced more than $70 billion in AI and high-performance computing contracts. At the same time, Bitcoin mining revenue per unit of computing power had at one point dropped to about $30 to $35 per PH/s per day. PANews says a group of sites using older equipment or paying higher electricity prices had moved close to loss-making territory.
The shift is straightforward. Miners that once supplied computing infrastructure to the crypto economy are now supplying core infrastructure to the AI economy. The capabilities did not change much. The customer base did.
From OpenSea to OpenRouter
The flow from crypto into AI is not limited to mining sites. People trained in crypto are moving too. Alex Atallah, OpenSea’s co-founder and former CTO, is one of the examples in the article. At the peak of the NFT boom, OpenSea’s monthly trading volume once topped $4 billion. In July 2022, Atallah left the company and prepared to start again. In 2023, he launched OpenRouter.
OpenRouter addresses a practical problem. Model providers keep multiplying, and their prices, speed and capabilities vary. Developers do not want to integrate a separate API for each one. Through OpenRouter, they can connect once, access hundreds of models and route requests according to price, performance and availability.
PANews argues that OpenRouter and OpenSea are not the same business, but they share a similar commercial structure. OpenSea aggregated NFT buyers and sellers. OpenRouter aggregates models, compute suppliers and developers. One matched digital asset trades. The other matches inference demand. The product changed, but the skill of building a marketplace and organizing fragmented supply did not.
In 2025, OpenRouter completed a combined $40 million in funding at a valuation of about $500 million. In May 2026, it raised a $113 million Series B led by CapitalG, lifting the valuation to about $1.3 billion. Over the past six months, weekly token volume processed on the platform rose from 5 trillion to 25 trillion.
Some crypto traces remain visible in the product itself. The article notes that OpenRouter’s sign-up page still places MetaMask alongside Google and GitHub login options, and the platform accepts USDC for payment.
Fal.ai is another case. Founder Burkay Gur had worked on machine learning infrastructure at Coinbase. He started the company in 2021, initially building machine learning data pipelines and deployment tools. After Stable Diffusion was open-sourced, the team saw that image and video models were multiplying, but inference remained slow, deployment was cumbersome and GPU utilization was poor. Fal.ai then shifted its focus to generative media inference.
The change paid off quickly. By mid-2025, Fal.ai’s annualized revenue was close to $95 million. In December that year, the company completed a $140 million Series D led by Sequoia, reaching a $4.5 billion valuation. Adobe, Canva and Perplexity are among the companies using its generative media infrastructure.
Crypto capital also found its way into AI
If miners supplied power and data center capacity, capital built in the crypto cycle entered AI through investments and hardware purchases.
PANews points first to Jed McCaleb. He founded Mt.Gox and later co-founded Ripple and Stellar, making him one of the earliest crypto billionaires. In 2023, Navigation Fund, which he backed, spent about $500 million to buy 24,000 Nvidia H100 GPUs in one shot and formed Voltage Park to lease GPU capacity to AI companies and research institutions.
Instead of launching another blockchain, McCaleb converted crypto wealth into one of AI’s scarcest assets. In 2026, Voltage Park merged with AI development platform Lightning AI. The transaction implied a valuation of about $2.5 billion for the combined entity.
The portfolio left behind by collapsed crypto exchange FTX and its founder Sam Bankman-Fried, or SBF, offers a more dramatic example in the article. In 2022, SBF invested $500 million in Anthropic when the company was still little known, taking roughly a 13.5% stake. After FTX went bankrupt, the estate sold those shares in stages in 2024 and recovered about $1.3 billion. Anthropic is now described as carrying a post-money valuation of $96.5 billion. The article estimates that if FTX had not sold, its stake would likely have been reduced to about 6.7% and would now be worth about $6.5 billion, around 130 times the original $500 million investment.
The Cursor story is more extreme. In April 2022, Alameda, another SBF-linked firm, invested $200,000 in an early round for Anysphere, the company that later launched the AI coding tool Cursor. After FTX entered bankruptcy, the estate sold that stake in April 2023 for the same $200,000, roughly flat. In June 2026, SpaceX announced an all-stock acquisition of Anysphere valued at $60 billion. Based on public reporting, Alameda had initially received about 5% of the company. Ignoring later dilution entirely, the paper value of that stake would have reached $3 billion, or 15,000 times the original investment.
PANews says this should not be reduced to a claim that SBF was an investment genius. Its reading is narrower: before ChatGPT was released, some of the most aggressive and risk-tolerant capital from the crypto bull market had already started searching for AI projects. That money was acting on two beliefs. Computing power would keep becoming more valuable, and software networks could scale globally in a short period. AI happened to fit both.
Crypto money is also funding new technical and organizational experiments
The article then turns to Nous Research. Its Hermes Agent is described as an open-source AI agent that can accumulate long-term memory and generate skills automatically. According to statistics cited from OpenRouter, Hermes Agent ranks first globally in token calls, ahead of Claude Code.
In 2025, crypto investment firm Paradigm led Nous Research’s $50 million Series A. Reporting at the time said the financing implied a roughly $1 billion valuation for its not-yet-issued token. Earlier backers included crypto VC Distributed Global and former Coinbase CTO Balaji Srinivasan.
Beyond Hermes, Nous is also developing Psyche, a distributed model training network built on Solana. Traditional AI labs usually gather large numbers of GPUs in a single data center. Psyche is trying to test a different route: connect GPUs spread across different regions and owned by different participants, train models jointly, and use smart contracts to coordinate training progress, verify participants and allocate rewards.
The article notes that Psyche remains an experiment for now. Its testnet token has also been explicitly labeled by the team as having no economic value. Even so, PANews presents it as another example of how crypto capital in AI is not limited to buying GPUs. It is also backing experiments in technology and organizational design.
OpenAI itself once seriously examined a token option
The piece closes by looping back to OpenAI. OpenAI was founded in 2015 as a nonprofit, but the funding needs of frontier models quickly outgrew what donations could support. By the end of 2017, Sam Altman and Greg Brockman had already begun discussing new funding structures, and one option on the table was a token issuance.
Internal emails released later showed that the team seriously studied issuing a token in early 2018. Elon Musk objected clearly, saying a token sale would badly damage OpenAI’s credibility. OpenAI later added that by the end of January that year, the team itself had also begun losing interest in the plan.
OpenAI ultimately chose to create a for-profit entity and later secured major investment from Microsoft. Sam Altman did not leave crypto behind, though. In 2019, he co-founded Worldcoin with Alex Blania and Max Novendstern. The project uses an iris-scanning device called Orb to verify that a user is a real and unique human, then builds an identity and payments network around World ID and the WLD token.
A transfer of resources, not a simple identity change
PANews ends with a caution. Stories involving Crusoe, CoreWeave, OpenRouter, Fal.ai and Nous Research carry obvious survivor bias, and they do not prove that crypto companies have a higher chance of succeeding if they pivot into AI.
What they do show, the article argues, is that crypto left behind assets that AI can use. Miners accumulated power, land and interconnection permits. Exchanges and Web3 companies trained engineers who understood distributed systems, GPU scheduling and global product rollouts. Wealth created by token appreciation became capital for buying graphics cards, investing in model companies and funding technical experiments. Crypto did not turn into AI out of nowhere. It passed the resources of one cycle into the next industry that wanted them more urgently.