OpenAI has gone wild with compute deals this year, committing to spend far more than its balance sheet can currently sustain. So who takes the fall if it can’t pay? It won’t be Altman.
By Rashi Shrivastava, Phoebe Liu, and Richard Nieva
Over the last few months, OpenAI CEO Sam Altman has been on a tear of dealmaking, announcing multibillion dollar agreements with the biggest tech companies in the world. There’s Oracle, Nvidia, Microsoft, AMD, Broadcom, and most recently, Amazon. He’s committed to spending a grand sum total of $1.4 trillion on datacenters in the coming years — an eyebrow-raising figure for a company which claims its annual revenue is projected to reach $20 billion this year, begging an all-important question for the entire tech industry, whose fate is now tied to OpenAI: What happens if he can’t pay?
At an event this week, OpenAI CFO Sarah Friar seemed to suggest that the government could act as a “backstop” for the company’s commitments — comments she later walked back. And in a long-winded post on X, Altman addressed the question of what happens to OpenAI if its web of deals falls apart:
“If we screw up and can’t fix it, we should fail, and other companies will continue on doing good work and servicing customers,” Altman said. “…We of course could be wrong, and the market—not the government—will deal with it if we are.”
The odds don’t look great right now. In order to come through on its compute commitments, OpenAI’s revenue would have to grow to an estimated $577 billion by 2029, roughly the size of Google’s revenue that same year, Tomasz Tunguz, a general partner at Theory Ventures, wrote in a recent blog post. That’s a roughly 2900% jump from its current projections for 2025. “These are crazy numbers,” he told Forbes.
But OpenAI has options. One likely scenario is that the AI company pays for and utilizes only a portion of the compute it has booked, said D.A. Davidson analyst Gil Luria. In that case, companies like Oracle, Amazon, Microsoft, CoreWeave and others will most likely renegotiate the contracts and ensure they get at least some amount of business from OpenAI, especially if the alternative is getting none at all. “They don’t want OpenAI to go bankrupt, so their incentive is to renegotiate,” he told Forbes.
Renegotiating contracts isn’t uncommon in the data center world. These contracts are extremely complex and often spread out over a span of years; some parties can even further extend timelines in case companies aren’t able to meet commitments. Clients like OpenAI are typically billed on the basis of usage. The “big numbers” being announced are often larger than what’s actually committed under contract, largely due to variables like share price, data center construction cost and GPU price, data center expert Daniel Golding said. For instance, OpenAI has committed to buy up to 6 GW of AMD’s chips (estimated to be worth around $90 billion) in exchange for about 10% of AMD shares—no cash on either side. But that hinges on performance milestones for OpenAI’s technology and commercial business as well as AMD’s share price.
The contracts often have some capital I ifs. Thanks to constraints on power supply and chip availability, there’s a possibility that some of these infrastructure providers aren’t able to deliver in time, another opportunity for OpenAI to weasel out of paying some of the top line number. OpenAI’s $22.4 billion in total contracts with CoreWeave, for instance, can be terminated by either party at any time “for cause” (legal speak for things like delays).
But even with about a trillion dollars on the line, the bigger risk in Altman’s mind is not having access to enough cheap compute to train and run better AI models when the time comes— a move that is crucial for revenue growth. “We believe the risk to OpenAI of not having enough computing power is more significant and more likely than the risk of having too much,” he said, adding that the company is also exploring ways to sell compute to other companies directly, much like CoreWeave does. When reached for comment, OpenAI spokesperson Steve Sharpe said the company has “nothing to add.”
Without A Stake, Altman Doesn’t Have Anything To Lose
Altman’s fixation on grabbing as much gargantuan compute as he can is not surprising. He has always knelt at the altar of scaling laws. Earlier this year, Altman mused on the coming of AGI, or artificial general intelligence, the company’s overarching goal of creating AI that matches or surpasses human capabilities. “The intelligence of an AI model roughly equals the log of the resources used to train and run it,” he wrote, noting you could get “continuous and predictable gains” from pumping money into those resources. “The scaling laws that predict this are accurate over many orders of magnitude,” he wrote.
Even before ChatGPT was released, he told employees at his crypto-cum-identity company Worldcoin (now called World), that one of his personal operating principles is “scale it up and see what happens,” something he’s found effective with everything from large neural networks to fusion reactors, Forbes reported. And the faster the better. Scaling up “earlier than it makes sense to … is super, super valuable,” he told Worldcoin employees.
Experts note chief dealmaker Altman doesn’t have anything to lose. He has repeatedly claimed he does not have a stake in the company, and won’t have a stake even after OpenAI has restructured to become a public benefit corporation. “He has the upside, in a sense, in terms of influence, if it all succeeds,” said Ofer Eldar, a corporate governance professor at the UC Berkeley School of Law. “He’s taking all this commitment knowing that he’s not going to actually face any consequences because he doesn’t have a financial stake.”
That’s not good corporate governance, according to Jo-Ellen Pozner, a professor of management and entrepreneurship at Santa Clara University’s Leavey School of Business. “We allow leaders that we see as being super pioneering to behave idiosyncratically, and when things move in the opposite direction and somebody has to pay, it’s unclear that they’re the ones that are going to have to pay,” she said.
Luria adds: “He can commit to as much as he wants. He can commit to a trillion dollars, ten trillion, a hundred trillion dollars. It doesn’t matter. Either he uses it, he renegotiates it, or he walks away.” There are of course more indirect stakes for Altman, experts said, like the reputational blow he’d take if the deals fall apart. But on paper, he’d seemingly be off the hook, they said.
As OpenAI has grown in prominence, tech giants have clamored to strike deals with the AI behemoth. “More of the world wants to work with us, so deals are quicker to negotiate,” Altman said recently. And they’ve reaped the upsides: Oracle, Nvidia, AMD and Broadcom gained a collective $636 billion in market cap on the days their deals were announced. Most recently, when OpenAI announced a $38 billion AI infrastructure deal with Amazon on Monday, the company’s stock increased 4%, adding $10 billion to Jeff Bezos’ net worth. “The math that Mr. Altman is doing in his head is ‘they need me more than I need them,’” Luria said.
Companies involved in these circular deals have already indicated they’re willing to bail each other out. In September, Nvidia said it would buy CoreWeave’s unsold computing capacity through 2032, initially worth $6.3 billion. Presumably, that could expand to include anything OpenAI doesn’t use given that it’s CoreWeave’s largest customer.
“If you owe the bank a hundred thousand dollars, the bank owns you. If you owe the bank a hundred million dollars, you own the bank,” said Lloyd Walmsley, a Mizuho analyst who covers Meta, Google and Amazon. “ Everyone’s holding hands and, and having this leap of faith that the products are so powerful.”
The Best And Worst Case Scenarios
It’s also possible that OpenAI uses all the compute it has booked, and needs more. In that case, the AI titan will need to raise more funding—either through private or public markets—and increase its revenue exponentially. Part of why Altman has spoken about a potential OpenAI IPO is that it would make raising cheaper debt much easier. Altman said on Thursday that he’s confident that revenue will continue to grow, primarily from upcoming enterprise products and categories like robotics and consumer devices.
Then, take the extreme case. If OpenAI were to file for bankruptcy protection, who would get paid first? Who would, in Sam Altman’s words, “get burnt”? For one, it’s entirely possible that some cash-rich company, perhaps one on the other end of one of its big contracts like Microsoft or Oracle, buys it in a fire sale. In the very unlikely event OpenAI goes bankrupt and liquidates instead, debtholders would get their money back first. Then equity investors, and finally—if there’s anything left over—common shareholders.
OpenAI has only announced one debt deal so far, a $4 billion credit facility with nine banks including JPMorgan, Citi, Goldman Sachs and Morgan Stanley. The company announced the loan in October 2024 and said it is a revolving credit facility (akin to a giant corporate credit card). It’s unclear if OpenAI has additional debt, which it isn’t required to disclose as a private company.
King among OpenAI’s equity holders is Microsoft, which owns 27% of the company following OpenAI’s restructuring as a for-profit company last week. Microsoft has invested $11.6 billion of its $13 billion commitment to OpenAI, and OpenAI has committed to purchase $250 billion of Microsoft Azure’s compute services over the coming years. The companies have two-way plans to share revenue. “Of the dominoes that are going to fall when OpenAI can’t pay everybody [and the question arises of] who gets paid first, and I’d argue that Microsoft gets paid first,” Luria added. Other large shareholders include Thrive, SoftBank, Dragoneer and investors in the $6 billion merger between Jony Ive’s io and OpenAI.
Common shareholders—employees, cofounders and the nonprofit—would get what’s left, proportionally to how much they paid for their shares. Interestingly, OpenAI’s nonprofit holds a special “Class N” common share, which gives it majority voting and veto power in elections of board directors, though that share isn’t entitled to any financial return if the company were to go under. (The nonprofit’s 26% stake likely also includes ordinary common shares.)
Altman’s wheeling and dealing is nothing short of mind-boggling, though many of the biggest payments aren’t due for a few years. In the AI universe, that’s plenty of time to figure out how he’s going to pay for it all, whether it’s additional fundraising or explosive revenue growth. Seems like nobody knows the answer just yet, not even him.
Read the full article here
