How AI eating crypto liquidity: Inside the $300B Oracle hit and Bitcoin miner pivots

Oracle did what every legacy tech giant dreams of. In September, it announced a $300 billion cloud deal wrapped around OpenAI, the hottest name in software, and watched its stock rip higher.

Two months later, the market gave its verdict. Oracle has shed more than $300 billion in market value, trading below its pre-AI announcement levels, while reports began calling it a “ChatGPT curse.”

Analysts are now treating the mega deal as a case study in what happens when AI promises outrun the cash flows that are supposed to support them.

At the same time, Cursor just raised $2.3 billion at a $29.3 billion valuation. The company crossed $1 billion in annualized revenue this year and more than tripled its valuation since June.

The coding tool vacuumed up venture capital on the promise that engineers would live inside an AI pair programmer that would write most of the code for them.

A private devtool 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 three-year-old startup a $29.3 billion price tag, does money still need crypto at all, or does crypto just get pulled into the same trade under a different ticker?

The AI money hose

A nice close look at the insane funding numbers explains this mood.

Global AI startup funding reached around $100 billion in 2024, roughly 80% more than in 2023 and close to a third of all venture capital that year. S&P Global puts generative AI funding at more than $56 billion in 2024, nearly double the prior year.

The Stanford AI Index tracks private investment in generative AI at $33.9 billion for 2024, more than eight times 2022. EY estimates that in just the first half of 2025, generative AI startups raised another $49.2 billion.

Crypto remembers what that looks like. In 2021, the hot trades were token issuance, DeFi yield, and metaverse equity. In 2024 and 2025, the center of gravity moved. The big checks went into training runs, data centers, and a small circle of foundation model labs. Barron’s counts roughly a third of global VC going into AI names like xAI, Databricks, Anthropic, and OpenAI.

On the public side, companies are raising giant debt piles to chase GPU capacity. Oracle is reportedly lining up around $38 billion of bonds to fund its cloud buildout. Nvidia’s data center revenue has reshaped entire equity indices. If you want exposure to “future cash flows from compute,” the highest beta now lives in AI infra and foundation models.

That does not mean liquidity vanished from crypto. It means marginal dollars are priced against a new benchmark. If a mid-size AI startup commands a $30 billion valuation and OpenAI can talk about trillion-dollar capex plans without being laughed out of the room, the bar for a $10 billion token with thin real-world usage gets higher.

AI tokens and the ASI experiment

Crypto did the logical thing: it tried to package AI inside tokens. The flagship effort was the Artificial Superintelligence Alliance, a plan to merge SingularityNET, Fetch.ai, and Ocean Protocol into a single ASI token and brand the whole stack as decentralized AI. Fetch.ai’s merger blog set out a simple sales pitch in 2024. One treasury, one token, three projects that claimed to cover agents, data, and models.

This worked for a while. Billions of dollars worth of AGIX, FET, and OCEAN liquidity were pointed at the same narrative. Exchanges lined up spot and perpetual pairs for ASI. Retail holders got migration bridges and one token that mapped cleanly to “AI” on a watchlist. It looked like crypto had found a way to compress a messy sector into something that could live in a single line of a derivatives blotter.

Then Ocean walked.

In October, the Ocean Protocol Foundation announced its withdrawal from the alliance, asking to depeg OCEAN from ASI and relist it as a separate asset.

Ocean framed the exit as a matter of “voluntary association.” Fetch.ai has since launched legal action, with court filings tracing conversions of more than 660 million OCEAN to FET and alleging broken promises around the merger.

This little governance drama tells you something about the AI token trade. It’s chasing the same story as the private AI boom, just with more volatility and basically no revenue. When ASI traded well, everyone wanted in. When valuations cooled and community politics reemerged, the “alliance” reverted to being three cap tables with different agendas.

From a liquidity point of view, AI tokens feel less like a separate asset class and more like a way for existing money in crypto to shadow what is happening in private AI. Cursor’s latest round or Anthropic’s new funding from Amazon do not move ASI on a strict basis, but they set the emotional tone. Crypto traders watch equity deals and price their AI baskets accordingly.

From Bitcoin mines to AI model farms

The clearest merger between AI and crypto sits in power contracts. Bitcoin miners spent a decade building data centers in cheap-energy regions, and AI hyperscalers are now paying up for the same megawatt base.

Bitfarms is the most explicit case. The company has announced plans to wind down Bitcoin mining entirely by 2027 and redeploy its infrastructure into AI and high-performance computing.

Its 18-megawatt facility in Washington state will be the first site converted, with racks designed for Nvidia GB300-class servers and liquid cooling capable of handling around 190 kilowatts per rack.

Bitfarms’ press release describes a fully funded $128 million agreement with a large US data center partner. Management claims that one AI facility could out-earn the company’s entire historical Bitcoin mining profits.

Bitfarms is not alone. Iris Energy rebranded as IREN and is shifting its hydro-powered sites into AI data centers, with Bernstein research pointing to billions in expected revenue from Microsoft-backed GPU deployments.

Hut 8 talks openly about being a power first platform that can point 1,530 megawatts of planned capacity to whatever workload pays best, with AI and HPC at the top of the list.

Core Scientific went far enough down this route that AI cloud provider CoreWeave agreed a $9 billion all-stock deal to buy it, aiming to lock up more than a gigawatt of data center power for Nvidia-heavy clusters, before shareholders pushed back.

The pattern is the same in each of these cases. Bitcoin mining gave these firms cheap power, grid connections, and sometimes hard-fought permits.

Then AI came along and offered a higher dollar per megawatt. For shareholders that have watched multiple halvings compress mining margins, routing energy into GPU stacks clearly looks like swapping a maturing carry trade for growth.

This is where the “AI is eating crypto liquidity” headline gets literal for Bitcoin. Every megawatt that moves from SHA-256 to GB300 or H200 is a unit of energy that no longer secures the network. Hash rate has continued to grow as new miners enter and older hardware is retired, but over time, a higher share of cheap power will be priced by AI’s willingness to pay.

When AI attacks the rails

There is one more junction between AI capital and crypto: security.

In November, Anthropic published a report on what it called the first large-scale espionage campaign orchestrated by an AI agent. A China-linked group jailbroke the company’s Claude Code product and used it to automate reconnaissance, exploit development, credential harvesting, and lateral movement across roughly 30 victim organizations.

Some of the attacks succeeded. Some failed because the model hallucinated fake credentials and stole documents that were already public. But the most alarming part was that most of the attack chain was driven by natural-language prompts rather than a room full of operators.

Crypto exchanges and custodians sit right in the middle of that blast radius. They already rely on AI inside trading surveillance, customer support, and fraud monitoring.

As more operations move into automated agents, the same tools that route orders or watch for money laundering will become targets. A dense concentration of keys and hot wallets makes them attractive to any group that can point a Claude-sized agent at a network map.

The regulatory response to that sort of event will not care whether the affected venue trades Nvidia equity, Bitcoin, or both. If a major AI-driven breach hits a big exchange, the policy conversation will treat AI and crypto as a single risk surface that sits on top of critical financial infrastructure.

So is AI really eating crypto liquidity?

The honest answer is that AI is doing something more interesting. It’s setting the price of risk for anything that touches compute.

Venture money that might once have chased L1s is now funding foundation models and AI infra. Public equity investors are weighing 30% drawdowns in Oracle against the chance that a $300 billion OpenAI cloud deal really does pay off.

Private markets are happy to value a devtool like Cursor on par with a mid-cap token network. Bitcoin miners are rebranding as data center operators and signing long-term contracts with hyperscalers. Token projects are trying to bolt “AI” onto their ticker because that is where the excitement sits.

Looking at this market from the depths of the crypto industry makes it look like a food chain where AI simply devours everything.

But alas, it’s always more nuanced and complicated than it looks. Over the past two years, AI has become the reference trade for future computing, and that trade drags Bitcoin infrastructure, AI tokens, and even exchange security into the same story.

So, liquidity is not leaving outright. It’s moving around, pricing everything else against the one sector that convinced markets to fund trillion-dollar capex plans on a promise and a demo.

The post How AI eating crypto liquidity: Inside the $300B Oracle hit and Bitcoin miner pivots appeared first on CryptoSlate.