Arthur Hayes Warns of AI Overbuilding and Bets on Crypto

Arthur Hayes Sees Overbuilt AI Infrastructure

Former BitMEX chief executive officer Arthur Hayes argues that the current artificial intelligence infrastructure boom is on a collision course with overcapacity. Speaking on Tuesday, Oct. 6, 2026, at the Gamma Prime Investing Conference in Singapore, Hayes stated that humanity is wasting multi-trillion dollars on building artificial intelligence data centers. According to CNBC Technology, the co-founder and chief investment officer of crypto investment firm Maelstrom contends that this massive buildout will ultimately make computing power extremely cheap and extremely plentiful. A photo illustration by Budrul Chukrut for SOPA Images via Getty Images captured the intersection of bitcoin and artificial intelligence themes that frequently appear in discussions surrounding these digital asset markets.

The thesis runs directly counter to the capital expenditure cycles driving global technology markets, where major corporations race to secure advanced silicon and grid capacity. Hayes views the current trajectory through a historical lens of industrial overextension. If you study financial history and every major technological rollout, it always is overbuilt, followed by a crash and a subsequent government bailout, according to his remarks at the Singapore conference. Investors who position for those bailouts stand to benefit, much like those who observed the aftermath of the 2008 financial crisis and other episodes over the past two decades.

Projected Compute Glut and Market Timeline

The financial pressure point could arrive in late 2027 or 2028, when much of the new data center capacity currently under construction is delivered to the market. Hayes noted that major end users driving demand for computing power, including SpaceX, OpenAI, and Anthropic, currently do not make money. Once infrastructure providers demand payment for the compute capacity those entities have committed to, the economic strain will intensify. While analyzing these tech market cycles, some observers might also consider Jim Cramer says 'frozen' conditions are holding back U.S. stocks when evaluating broader equities.

While some suppliers in the semiconductor supply chain are already making money, including memory chipmakers and Nvidia, the question for investors is whether they are paying the right multiple for those companies forward earnings. The central bull case relies on artificial intelligence becoming so useful over the next 12 months that natural end-user demand grows enough for AI companies to become profitable. Despite anticipating a downturn, Hayes stated he does not like betting on falling prices or shorting artificial intelligence companies, calling it not really a great investment opportunity.

Flop Venture and Crypto Liquidity Bets

Rather than shorting equities, Hayes is positioning capital for the macroeconomic policy response he expects will follow any sector correction. Thankfully, we have bitcoin and other crypto to soak up that excess liquidity, Hayes said, adding that investors just have to be patient to capture the upside when the bailout arrives. Portfolio planners often balance such high-stakes volatility with personal stability measures, much like Adam J. Sagot-Well-Being in the Boardroom, From Initiative to Strategic Advantage highlights for corporate leadership.

This anticipated abundance of cheap computing power underpins his latest venture, an artificial intelligence agent payments project named Flop, which is slated to launch in the first quarter of 2027. Because current financial architecture lacks a dedicated payments network for autonomous software agents, Flop intends to establish a spot market for compute. Participants will be rewarded with Flop tokens for providing graphics processing units and performing artificial intelligence inference. If agents can convert a currency directly into compute, which is what they eat and consume, then they will use this currency, Hayes said of the project's core mechanism.