Oracle has achieved what any legacy technology giant could have dreamed of. In September, the company announced a $300 billion cloud deal around OpenAI, the most popular name in the software industry, and watched its stock price soar.
Two months later, the market has given its verdict. Oracle has lost more than $300 billion in market capitalization and is trading below pre-AI announcement levels, which the press has begun to refer to as the “ChatGPT curse.”
Analysts are now treating the mega-deal as a case study in what happens when the promise of AI outstrips the cash flow meant to support it.
At the same time, Cursor just raised $2.3 billion at a valuation of $29.3 billion. The company’s annual sales exceeded $1 billion this year, and its valuation has more than tripled since June.
This coding tool has siphoned venture capital with the promise that engineers will live inside an AI pair of programmers and write most of the code for them.
A private development tools startup and a public software incumbent are suddenly part of the same mental spreadsheet as most L1 tokens, and investors are now asking a slightly rude question:
When AI can hand a $29.3 billion price tag to a three-year-old startup, does money still need cryptocurrencies, or will cryptocurrencies simply be pulled into the same trade under a different ticker?
AI Money Horse
A closer look at the insane fundraising numbers explains this atmosphere.
Global AI startup funding will reach approximately $100 billion in 2024, an increase of approximately 80% from 2023 and accounted for nearly one-third of all venture capital in that year. S&P Global predicts that AI funding generated in 2024 will be more than $56 billion, nearly double the previous year.
According to the Stanford AI Index, private investment in generative AI in 2024 will be $33.9 billion, more than eight times the amount in 2022. EY estimates that generated AI startups raised an additional $49.2 billion in the first half of 2025 alone.
Crypto remembers what it’s like. Trending deals in 2021 were token issuance, DeFi yield, and Metaverse equities. In 2024 and 2025, the center of gravity has shifted. Big checks were done on a small circle of training runs, data centers, and basic model labs. Barron’s counts that about a third of the world’s VCs are into AI names such as xAI, Databricks, Anthropic, and OpenAI.
On the public side, companies are racking up massive debt in pursuit of GPU power. Oracle is reportedly preparing about $38 billion in bonds to fund its cloud buildout. Nvidia’s data center revenue reshaped the entire stock index. If you want to take advantage of “future cash flow from compute,” the highest beta currently exists for AI infrastructure and underlying models.
That does not mean that liquidity has disappeared from cryptocurrencies. This means that marginal dollars are being priced against the new benchmark. If a mid-sized AI startup can hit a $30 billion valuation and OpenAI can talk about multi-trillion dollar capital investment plans without getting laughed out of the room, the bar for a $10 billion token with little real-world usage becomes even higher.
AI token and ASI experiment
Cryptocurrency did the logical thing. I tried to package the AI inside a token. The main effort is the Artificial Superintelligence Alliance, which plans to combine SingularityNET, Fetch.ai, and Ocean Protocol into one ASI token and brand the entire stack as decentralized AI. Fetch.ai’s merger blog laid out a simple sales pitch for 2024. Three projects that claim to cover one finance, one token, agent, data, and model.
This worked for a while. Billions of dollars worth of AGIX, FET, and OCEAN liquidity were also directed toward the same story. The exchange has lined up ASI spot pairs and permanent pairs. Retail holders received a migration bridge and one token that specifically mapped to “AI” on the watchlist. Cryptocurrencies seemed to have found a way to compress a messy sector into something that would fit on a single line on a derivatives blotter.
Then Ocean walked.
In October, the Ocean Protocol Foundation announced its withdrawal from the partnership, calling for OCEAN to be separated from ASI and relisted as a separate asset.
Ocean framed secession as a matter of “voluntary association.” Fetch.ai has since launched legal action, with court filings tracking more than 660 million OCEAN to FET conversions and alleging broken promises regarding the merger.
This little governance drama tells us something about AI token trading. It is following the same story as the private AI boom, but the volatility is high and there is basically no profit. When ASI was trading well, everyone wanted to be a part of it. As valuations cooled and regional politics re-emerged, the “alliance” returned to three cap tables with different agendas.
From a liquidity perspective, AI tokens feel less like a separate asset class and more like a way for existing funds in cryptocurrencies to mirror what’s happening with private AI in the shadows. Cursor’s latest round and Anthropic’s new funding from Amazon don’t move ASI on a strict basis, but they set an emotional tone. Cryptocurrency traders monitor stock trades and price the AI basket accordingly.
From Bitcoin mines to AI model farms
The most obvious convergence of AI and cryptocurrencies is in power contracts. Bitcoin miners spent a decade building data centers in regions with cheap energy, and AI hyperscalers are now paying the same per-megawatt costs.
Bitfarms is the most obvious case. The company announced plans to completely phase out Bitcoin mining by 2027 and redeploy its infrastructure to AI and high-performance computing.
The company’s 18-megawatt facility in Washington state will be the first to feature racks designed for Nvidia GB300-class servers and liquid cooling capable of handling approximately 190 kilowatts per rack.
Bitfarms’ press release describes a fully-funded $128 million deal with a leading US data center partner. Management claims that a single AI facility could exceed the company’s entire historical Bitcoin mining profits.
Bitfarms is not alone. Iris Energy is rebranding to IREN and migrating hydropower plants to AI data centers. Bernstein’s research points to billions of dollars in revenue expected from Microsoft-backed GPU deployments.
Hut 8 puts AI and HPC at the top of the list, and speaks openly about being a power-first platform that allows you to direct 1,530 megawatts of planned capacity to what will most benefit your workloads.
Core Scientific fared far along this path, before shareholders revolted, until AI cloud provider CoreWeave agreed to a $9 billion all-stock deal aimed at securing more than 1 gigawatt of data center power for Nvidia-heavy clusters.
The pattern is the same in each case. Bitcoin mining has given these companies cheap power, grid connections, and sometimes a license to fight hard.
Then AI came along and made it more expensive per megawatt. For shareholders who have seen mining margins compressed over multiple halvings, powering the GPU stack clearly looks like trading a mature carry trade for growth.
Here, the headline “AI is eating up crypto liquidity” literally applies to Bitcoin. Each megawatt you move from SHA-256 to GB300 or H200 is no longer a unit of energy that protects your network. Hashrate continues to rise as new miners enter and old hardware is retired, but over time a larger proportion of cheap power will be priced in by AI’s willingness to pay.
When the AI attacks the rails
There is another intersection between AI capital and cryptocurrencies. It’s security.
In November, Anthropic published a report on what it called the first large-scale espionage operation organized by AI agents. A Chinese-linked group jailbroken the company’s Claude Code product and used it to automate reconnaissance, exploit development, credential collection, and lateral movement across approximately 30 victim organizations.
Some of the attacks were successful. Some failed because the models hallucinated fake credentials and stole documents that had already been published. But the most alarming part is that most of the attack chains were triggered by natural language prompts rather than a room full of operators.
Cryptocurrency exchanges and administrators are right in the middle of that explosion spectrum. They already rely on AI internally for transaction monitoring, customer support, and fraud monitoring.
As more operations move to automated agents, the same tools that route orders and monitor money laundering will become targets. The high concentration of keys and hot wallets makes them attractive to groups that can direct Claude-sized agents onto the network map.
The regulatory response to this type of event does not care whether the affected venue trades Nvidia stock, Bitcoin, or both. If a major AI breach were to occur at a major exchange, policy discussions would treat AI and cryptocurrencies as a single risk surface sitting on top of critical financial infrastructure.
So, is AI really eating up the liquidity of cryptocurrencies?
The honest answer is that AI is doing more interesting things. It sets the price of risk for everything involved in computing.
Venture money that may once have chased L1 is now funding underlying models and AI infrastructure. Public equity investors are weighing Oracle’s 30% drawdown against the likelihood that the $300 billion OpenAI cloud deal will actually be profitable.
The private market is willing to value a development tool like Cursor on par with a mid-cap token network. Bitcoin miners have rebranded themselves as data center operators and have long-term contracts with hyperscalers. Token projects are looking to bolt “AI” into their tickers. Because that’s where the excitement comes from.
If you look at this market from the depths of the crypto world, it looks like a food chain where AI devours everything.
But sadly, it’s always more nuanced and complex than it seems. Over the past two years, AI has become the standard trade for the future of computing, and that trade has dragged Bitcoin infrastructure, AI tokens, and even exchange security into the same story.
Therefore, liquidity is not completely drained. For one sector that the market has persuaded to fund a multi-trillion dollar capital investment program based on promises and demonstrations, it is moving around, setting prices for all others.
