Skepticism is mounting about the viability of Oracle’s multiyear deal with OpenAI, a company that has yet to report a profit. Oracle’s stunning results on Tuesday inspired confidence from the market about the longer term viability of AI demand. Oracle shares ripped 36% higher on Wednesday for its best day since 1992 . But doubts are growing about the growth forecast given its customer concentration and its high leverage. Oracle said it has $455 billion in remaining performance obligations, or RPO, up 359% from a year earlier. However, the bulk of this tally is tied to just one customer: OpenAI . In other words, the market is betting on OpenAI’s ability to follow through on its commitment. CNBC confirmed a Wall Street Journal report that the AI company agreed to buy $300 billion in computing power over the next five years from Oracle beginning in 2027. OpenAI has already committed more than $100 billion to other AI and neocloud projects while having about $12 billion in annual recurring revenue. According to D.A. Davidson managing director Gil Luria, OpenAI would need more than $300 billion in revenue to “justify that level of spend.” Luria is one of the handful of analysts on Wall Street that remain neutral-rated on Oracle shares after its quarterly report. “Our enthusiasm for Oracle’s backlog announcements is significantly tempered by the report that it came almost entirely from OpenAI,” Luria said. He noted that OpenAI’s nonprofit status is limiting its ability to take in the $40 billion it raised in March, although he expects this issue to be resolved by next year given Microsoft’s stake in the company. Luria’s $300 price target suggests that Oracle stock — which is up 75% for the year — could gain less than 3% from Friday’s close. Shares slid 5% on Friday as investors raked in profits after this week’s more than 25% pop. Murky details JPMorgan analyst Mark Murphy similarly remains neutral-rated on Oracle. He said the risk-reward looks balanced after its gains this week. He lifted his year-end price target by $60 to $270, but that implies shares could slide more than 7%. Murphy likes that more than 70% of Oracle’s total revenue is recurring. Also, he expects the company could benefit from a broader market rotation into value as it is a “relative safe haven” in the software sector. However, the murky profiles of Oracle’s contracts remains top-of-mind for the analyst. “Oracle is also closing business outside of OpenAI, including at least 2 other contracts that are worth ‘multi-billions’, and yet we are in the dark as to whether the multi-billions add up to $4B out of the $455B RPO balance, or $10B, $50B, $100B, etc.— a huge swing factor when trying to assess customer concentration risk and likelihood of customers being able to fund $455B worth of future payments — plenty of tech companies carry net debt, and/or minimal cash, and/or will be burning cash for the next several years — that list is not limited to Oracle itself,” Murphy wrote Wednesday in a note to clients. “Short-term trading responses tend to be headline-driven and controlled by the instinctual mind, while deeper thinking takes longer to settle in, in some cases requiring quarters or years,” the analyst added. Although Oracle’s stock gains took the company to the tenth-largest spot in the S & P 500 , its cash holdings paint a different picture than its peers. The company had $11.2 billion in cash and short-term investments at the end of its May 31 quarter, which is significantly less than the “Magnificent Seven” cohort. Microsoft and Google parent Alphabet , for example, have a cash moat of about $95 billion each. To be sure, Oracle reported a free cash flow of $5.8 billion for its fiscal year 2025, which marked a big jump from the previous year’s $394 million. ORCL 1Y mountain Oracle stock performance over the past year. Talks of an AI bubble Oracle’s huge stock move this week could mark signs of a peak AI bubble, according to Gary Marcus, a New York University professor and Geometric Intelligence founder. OpenAI is more vulnerable than many people realize, said Marcus. The AI researcher cited the disappointing reception for its Chat GPT-5 release and its lack of a technical advantage to back up his claim. “There’s many reasons to wonder whether OpenAI is really going to be able to pull this off. … There’s lots of ways in which things could go wrong for them — having one customer that you have this agreement with that may not even be able to pay the bills, and, of course, Oracle has its own financial issues and doesn’t actually have the chips,” Marcus said in an interview. “It all just seems very speculative to me. … I would be astonished if they really did collect all of the $300 billion.” On social media, some have flagged the dangers of an interconnected AI ecosystem, which relies on close partnerships between AI model makers, hardware suppliers for GPUs and infrastructure and software makers. “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,” said Ophir Gottlieb, chief executive of boutique market research firm Capital Market Laboratories, in an X post . Lazard portfolio manager Shanu Mathew similarly questioned the Oracle-OpenAI deal in an X post , writing that “so much of AI feels levered to whether OpenAI can eventually generate meaningful revenues and keep financing massive amounts of infrastructure … separately, even on the token usage side (inference), AI continues to be AI’s biggest customer — agents, vibe coding, etc. It feels a little house of cards-y.” —CNBC’s MacKenzie Sigalos contributed to this report.