The artificial intelligence gold rush consumes capital at an unprecedented rate. However, former BitMEX CEO Arthur Hayes believes the market faces an inevitable crash. In his view, massive spending on AI infrastructure will trigger severe overcapacity.
This bubble will burst, leading to market panic and eventual government bailouts. Consequently, Hayes predicts this cycle will send cryptocurrency valuations soaring.
Hayes delivered this blunt warning at the Gamma Prime Investing Conference in Singapore. He argued that the tech sector is wasting trillions of dollars on data centers. These facilities will soon outstrip actual economic demand.
Tech giants still rush to secure computing power for next-generation models. Yet this furious building spree will soon make compute power cheap and abundant.
Hayes’ view clashes directly with Silicon Valley’s relentless optimism. His argument follows the historical patterns of major technological revolutions. Every major rollout from 19th-century railroads to the dot-com bubble shares the same cycle.
First comes irrational exuberance and extreme overbuilding. A painful market crash quickly follows, ending with a government rescue operation.
The Inevitable Tech Overbuild
The fundamental flaw in the current AI expansion, according to Hayes, lies in the severe disconnect between infrastructure costs and end-user profitability.
Leading developers like OpenAI, Anthropic, and xAI drive computing demand. However, these firms still operate without sustainable profits.
Right now, AI companies aggressively pre-book server space to win the tech race. But data center operators will soon demand payment for these computing loads.
According to a recent report by CNBC, Hayes expects this financial reckoning to materialize in late 2027 or 2028, which is when the bulk of this newly constructed data center capacity is scheduled for delivery.
The only viable scenario that avoids a crash is if artificial intelligence applications become so indispensably useful over the next twelve months that end-user demand explodes, allowing AI firms to suddenly turn massive profits and comfortably cover their infrastructure bills.
While companies supplying the actual hardware, such as Nvidia and various memory chipmakers, are currently realizing enormous revenue, Hayes cautions investors to carefully evaluate whether they are paying the right multiples for these companies’ future earnings.
Despite his bearish outlook on the overarching AI bubble, he actively discourages betting against the market through short-selling, noting that shorting major technological rollouts rarely presents a sound investment strategy given the sheer unpredictability of market timing.
The Crypto Upside and a Spot Market for Compute
Instead of betting on the collapse itself, Hayes is positioning himself for the aftermath. Whenever financial history produces a severe overbuild and a subsequent crash, governments and central banks predictably step in with bailouts and massive liquidity injections.
For Hayes, this represents a massive macroeconomic trigger for digital assets. He argues that Bitcoin and other cryptocurrencies are perfectly positioned to soak up the excess fiat liquidity that will inevitably flood the system when the financial rescue operations begin.
Investors who patiently position themselves for this exact scenario stand to reap the same kind of monumental gains seen in the wake of the 2008 financial crisis.
Beyond just waiting for a liquidity event, Hayes is actively building infrastructure to capitalize on the resulting glut of cheap processing power. His latest venture, a crypto project called Flop, is designed to launch in early 2027 as a dedicated spot market for computing power.
Because the overbuild will likely make compute extraordinarily cheap and plentiful, Hayes anticipates an explosion in the deployment of autonomous AI agents.
These agents will require constant processing power to function, essentially consuming compute as their primary resource. Flop intends to provide the underlying payment network, rewarding participants with tokens for supplying GPU power and performing AI inference.
By creating a native currency that AI agents can directly convert into the computing power they need to survive, Hayes is betting that the wreckage of the AI infrastructure bubble will inadvertently lay the groundwork for a thriving, machine-driven economy.
