The Trillion-Dollar AI Data Center Rush: A Bubble Waiting to Burst?
At a recent high-profile investment summit in Singapore, Arthur Hayes, the co-founder of BitMEX, presented a provocative thesis: humanity might be engaged in a massively expensive "waste" with its current spending on artificial intelligence data centers.
The Inevitable Glut of Computing Power
Hayes argues that the frenzied race for computing power among tech giants is fueling a global data center construction boom. He believes this超前 investment, based on optimistic forecasts of future demand, is setting the stage for a severe overcapacity of computational resources.
"When all the currently planned facilities come online, computing power will become extremely cheap and abundant," Hayes stated. He highlighted that major off-takers like SpaceX, OpenAI, and Anthropic are not yet profitable in their core operations. The real stress test will come when infrastructure providers demand payment for the massive computing capacity these companies have committed to.
Crash and Bailout: A Classic Financial Cycle
Hayes predicts the AI investment mania will follow a familiar financial script: overinvestment, followed by a market crash, culminating in government or central bank intervention and bailouts. He is betting on this sequence of events playing out.
"I think the crucial test could arrive between late 2027 and 2028," he added. At that point, high fixed costs may collide with potentially underwhelming demand growth, creating significant financial strain for many projects.
Winners and Losers in the AI Gold Rush
It's important to note that Hayes isn't dismissing the entire AI ecosystem. He acknowledges that upstream players like memory chip makers and core hardware suppliers such as Nvidia have already booked substantial profits. The risk, in his view, is concentrated in the overbuilding of downstream application layers and infrastructure.
An Unexpected Opportunity for Crypto
Within Hayes's framework, crisis breeds opportunity. A key corollary of his prediction is that the massive liquidity inevitably unleashed to stave off an economic downturn following a crash will need a destination. This "excess liquidity" could very well flood into assets like Bitcoin, potentially driving their prices significantly higher.
Simultaneously, Hayes is actively exploring the intersection of AI and crypto. He revealed details of a new initiative called "Flop," designed to create a decentralized spot market for computing power. The project plans to use a token to incentivize participants who contribute GPU resources for AI inference tasks, with a launch targeted for the first quarter of 2027.
Hayes's warning serves as a sobering counter-narrative to the prevailing AI euphoria. It underscores the need to balance enthusiasm for technological innovation with vigilance against the risks of irrational exuberance and capital misallocation. History suggests that when a bubble in one sector deflates, the capital that flows out often finds its way into creating the next one.