Last week, Oracle surprised Wall Street with a massive $300 billion deal with OpenAI, a five-year deal that helped send Oracle’s stock soaring—and brought simmering fears of an ‘AI bubble’ back to the surface.

Oracle shocked analysts in its latest quarterly earnings call with revenue projections that cited $455 billion in contracts, up 359% from a year earlier. The optimistic forward-looking numbers caused the company’s stock to jump 36% on Wednesday, the company’s biggest one-day increase ever, and briefly made CEO Larry Ellison the richest man in the world.

Part of the reason Oracle was able to strike the deal with OpenAI at all is due to Ellison’s courting of Nvidia CEO Jensen Huang, which has allowed his company, despite previously trailing behind other cloud providers, to secure a large stockpile of top-of-the-line Nvidia GPUs and position itself as a significant player in the AI infrastructure space. The rally added to the 45% gain the company already notched up this year, and cemented Oracle’s AI-fueled comeback.

But while securing top-tier GPUs has bolstered Oracle’s infrastructure position, some analysts were quick to warn that the financial risk was heavily concentrated in a single, unproven customer. According to a Wall Street Journal report, the bulk of the company’s $455 billion remaining performance obligations, or RPO, will come from the $300 billion deal OpenAI. The AI firm announced it will tap Oracle’s computing infrastructure under the multi billion deal, one of the largest cloud contracts ever signed. It also far exceeds OpenAI’s current revenue, which recently hit $12 billion in annualized revenue, per The Information.

Because remaining performance obligations represent contracted but not yet delivered services, they are not guaranteed revenue; customers can delay, renegotiate, or even cancel portions of these commitments.

Cue fresh alarm bells over a potential AI bubble.

Fears that the AI sector might be in a bubble have intensified recently due to a combination of sky-high valuations, early signs of disappointing returns, and cautionary remarks from industry leaders. A recent study from MIT that found 95% of AI pilot programs fail to deliver meaningful returns, despite over $40 billion having been invested in generative AI projects, fueled fears that a gap was emerging between investment hype and real-world results. Days before the report was released, OpenAI CEO Sam Altman also said that he believed the AI sector might be experiencing a bubble in the private markets, expressing concern over the level of investor enthusiasm and the overvaluation of some startups.

Gary Marcus, an AI expert who has been warning of a potential bubble and problematic economics of AI since 2023, called the OpenAI-Oracle deal “peak bubble.”

“Oracle’s new market cap, near a trillion dollars, up nearly 50% this week, driven largely by this one apparently non-binding deal with a party that doesn’t have the money to pay for the services, seems more bonkers than most,” Marcus wrote in a Substack post.

He wasn’t the only one raising alarms about the deal’s credibility.

“This is a grotesque attempt by both Oracle and OpenAI to mislead investors and the markets at large with a contract that neither party can fulfill, and it virtually guarantees that OpenAI will run out of cash in the next few years,” Ed Zirtron, a technology writer and founder and CEO of EZPR who has also emerged as a vocal skeptic of the hype surrounding AI, said in a blog post. “OpenAI, while claiming it’ll make more revenue than NVIDIA by 2030, needs $250bn funding over the next four years to pay its $300bn compute contract with Oracle, who cannot physically build the data centers to service it in time.”

And it wasn’t just those who doubt the underlying potential of today’s AI models that questioned the economics of the deal.

“I’m not an AI bubble person, but it is very understandable for investors to be confused/concerned by the OpenAI-Oracle deal lol. OpenAI hasn’t even gotten the for-profit conversion approved and is promising people 300 billion dollars??” Miles Brundage, an AI researcher and former head of policy research at OpenAI, added in a post on X.

Investors also had questions. “How is this all going to work exactly? ORCL has to buy the chips, take on more debt, while OpenAI has $10B in revenue but will spend $60B/yr in CapEx for five years. What?” Ophir Gottlieb, CEO of Capital Market Laboratories, wrote on X .

OpenAI has made several other billion-dollar deals recently, including $10 billion to develop custom AI chips with Broadcom. The company is also still in the process of figuring out how exactly it’s going to restructure its corporate governance to allow its for-profit to raise more capital, although it’s made a significant step recently by getting lead investor Microsoft on board.

Taken together, OpenAI’s current revenue and capital commitments fall far short of what would be needed to fully fund the Oracle contract, with analysts estimating the company would need hundreds of billions in annual revenue to meet these obligations.

Representatives for OpenAI and Oracle did not immediately respond to a request for comment from Fortune.