“AI investment is a waste”: Is Arthur Hayes’s warning being borne out now? [Crypto Briefing]

[Financial News] The AI clock has stopped. The KOSPI, driven by AI chip stocks such as Samsung Electronics and SK hynix, fell for seven consecutive trading days, shedding 1,500 points. It relinquished the 6,000 level it had held for more than two months. In the United States, AI chip stocks such as Micron and SanDisk also slumped, with share prices reeling after falling about 18% over two weeks. The declines reignited debate over the profitability of AI investments. Bitcoin, by contrast, has held the 110 million won level so far this month, showing resilience even amid high interest rates. The warning from Arthur Hayes, who said, “Investment in AI infrastructure is a waste of capital, and Bitcoin lies at the end of it,” (pictured) is attracting attention.
Arthur Hayes, CIO of Maelstrom, said bluntly in an interview with Financial News on the 16th, “The current build-out of AI infrastructure is the most serious inefficiency in the allocation of capital.” A co-founder of crypto exchange BitMEX, he also serves as CEO of blockchain project Flop Labs.

Hayes cited profitability metrics to support his argument. “No matter how many data centers are built, the computing infrastructure they provide is not generating the returns people expect,” he said. Citing Anthropic’s results as an example, he noted, “For every $2 spent, the revenue generated was $1. That suggests profitability is low across the ecosystem.”
The problem comes next. He said, “If things continue this way, it could trigger a government bailout,” adding, “A bailout on a larger scale than the one during the 2008 Global Financial Crisis may be needed.” At that point, he turned to Bitcoin. “That is precisely why Bitcoin was created. It is also why it has been pushed up to its current price,” he said.
The $1 million figure he put forward (about 1.35 billion won) also fits this context. He stopped short of stating it as a certainty. “$1 million is an inspiring figure I’d like to set as a target. I don’t know whether it will actually happen,” he said, while adding, “Given that AI capital is being wasted, I think it could reach that price level in 10 years.” The logic is that the more indebted an economy is, the more it will ultimately have no choice but to pump money into the system, while those managing hundreds of billions will use that money to buy financial assets.

Asked what would become money when AI agents start making payments to one another, he pointed to decentralization as the answer. “The most robust networks are decentralized. Human currencies are not,” he said, adding, “If I were an AI, I wouldn’t want to depend on humans or the government.”
He also pointed to the limitations of dollar-based stablecoins. “Because stablecoins are based on the dollar, governments can influence the decision-making process,” he said. He added that “when you try to pay for computing with the stablecoins you hold, you may not get the terms you want,” explaining that a means of payment directly exchangeable for computing resources is needed.
In response to the observation that major players are entering the same market, he cited value distribution as a differentiator. “In centralized networks today, the benefits don’t accrue to users,” he said, adding, “If AI agents use our network, we can return that value to token holders.”
He said he was also watching what AI agents were actually doing. A trading game predicting Nvidia’s share price is underway, he said, adding, “AI was very good at exploiting flaws in the logic.”

Asked how he viewed the Korean market’s connectivity to overseas markets, he gave an unexpected answer. “From an exchange perspective, Korea is already quite globally connected,” he said. “If you use Upbit or Bithumb, you can withdraw crypto assets anywhere in the world and use them for trading, and you can also use decentralized exchanges such as Hyperliquid.”
He said Maelstrom’s investments fall into four categories: early-stage investments in promising projects, putting in several hundred thousand dollars in exchange for tokens; tokenomics advisory work; long positions in projects with established liquidity; and buyouts of crypto businesses. He cited Ethena Protocol, which created the fourth-largest stablecoin, as its most successful advisory case. Regarding investment in Korea, he said, “We don’t invest directly, but a lot of crypto-related investment is taking place, so we’re watching with interest.”
He cited Ethereum (ETH), Uniswap, and real-world asset (RWA) tokenization as areas he favors. The criteria were clear: projects that generate actual revenue through product-market fit and return that revenue to token holders, for example in the form of dividends.
Finally, he set out for himself the conditions under which his outlook would be wrong. “If computers become so expensive that AI agents can no longer be run, then the outlook that AI and crypto will form a single economic sphere would have to be reconsidered.”
[email protected] Young-jun Han, Hyung-min Kim Reporter