TLDR:
- AI Bubble risk grows as oil prices threaten to raise computation costs.
- US restrictions on Anthropic’s models expose foreign users to sudden access loss.
- Chinese open-source AI models cost far less than premium US alternatives.
- Cheaper alternatives threaten the high-margin pricing US AI firms depend on.
The AI bubble could face a sharp correction, according to Arthur Hayes, co-founder of BitMEX. In an interview with Bonnie Blockchain on June 26, 2026, Hayes outlined three factors that could trigger a downturn in artificial intelligence valuations.
His comments touched on energy costs, government policy, and the growing appeal of open-source alternatives. Hayes framed these issues as interconnected risks facing the sector.
Oil Prices Could Strain AI Computing Costs
Hayes pointed to rising oil prices as one risk to the AI bubble. He noted that geopolitical tensions, including a possible US-Iran conflict, could push oil prices markedly higher within four to six months. Such an increase would raise energy costs across the board.
AI computation relies heavily on energy-intensive data centers. Higher oil prices would translate into higher operating costs for companies running large-scale AI models. This dynamic could test whether current AI investments can withstand a sustained rise in energy expenses.
Hayes suggested that many portfolios built around AI growth assumptions may not have accounted for this variable.
If energy costs climb sharply, the profitability models underlying many AI firms could come under pressure. This remains one of the more immediate risks he described.
Government Policy Adds Uncertainty for Users
The politicization of AI development is another factor Hayes raised. He cited the US government’s restrictions on Anthropic’s Mythos and Fable models as an example, which limited access to American users and, at times, within the company itself. Hayes described this as evidence of how policy decisions can disrupt access without warning.
This kind of restriction creates uncertainty for non-US companies and individuals who depend on these models. A sudden policy shift could cut off service entirely, regardless of a user’s payment status or business needs. Hayes argued that this vulnerability is not fully priced into current AI valuations.
For foreign users, the risk of losing access to premium AI tools introduces an added layer of caution. Businesses built around continuous AI service may need contingency plans. Hayes suggested this uncertainty could shape how companies choose their AI providers going forward.
Open-Source Models Threaten Premium Pricing
Hayes also addressed the shift toward open-source AI models, many of which originate from Chinese developers. He said these models often cost roughly one-tenth of comparable US offerings while delivering similar performance. This price gap could draw users away from proprietary systems.
Open-source models also give users greater control over their own data, according to Hayes. This appeals to companies and individuals wary of relying on closed systems tied to a single government’s regulatory decisions. Cost and control together make these alternatives increasingly attractive.
Hayes concluded that this shift threatens the pricing power of US AI companies. Many of these firms depend on high margins to justify their valuations. A move toward cheaper, self-hosted alternatives could pressure that business model over time.
