The AI Boom Runs Through One Private Company. Its Books Are the Story This Week

On Thursday the Nasdaq fell 2.6%, its steepest one-day drop since early June, and the trigger was a company that had just beaten expectations. Meta reported its first cash outflow in more than a decade, driven by spending on artificial intelligence, and its shares fell 7%. The Philadelphia Semiconductor Index lost 6.7% in a day. “Investors are starting to realize that there are some holes in the AI story,” said Paul Nolte, market strategist at Murphy & Sylvest.¹

Wednesday brings two more tests. The Federal Reserve announces its decision at 2 p.m. Eastern, with rates at 3.75% to 4% after a September hike; futures put the odds of a hold at about 90%, and a December hike is seen as more likely.² After the close, Alphabet and Microsoft report, and the line investors will read first is capital spending.³

But the most important company this week is not on the earnings calendar. It is private. Its name is OpenAI, and a remarkable share of the spending that moves Nvidia, Oracle, Microsoft and the rest is tied to its promises to pay.

But the most important company this week is not on the earnings calendar.

A trillion dollars of promises

Last November, Sam Altman wrote on X that OpenAI expected to end 2025 “above $20 billion in annualized revenue run rate” and added: “We are looking at commitments of about $1.4 trillion over the next 8 years.”⁴

The other side of those commitments is a who’s who of the AI trade. Oracle signed a cloud contract worth about $300 billion over roughly five years starting in 2027, and its shares rose 36% in a day, their best since 1992. Nvidia agreed to invest $100 billion in OpenAI in $10 billion tranches. The loop is simple: Nvidia invests in OpenAI, OpenAI pays Oracle, Oracle buys Nvidia chips. As one analysis put it at the time, “these are press releases and letters of intent.”⁵ A tally by venture investor Tomasz Tunguz put OpenAI’s hardware and cloud commitments at about $1.15 trillion for 2025 to 2035: $350 billion with Broadcom, $300 billion with Oracle, $250 billion with Microsoft, $100 billion with Nvidia, $90 billion with AMD, $38 billion with Amazon and $22 billion with CoreWeave.⁶

✓ Trusted partner One veteran analyst says he has named the single company he believes could set off the next crash, and he has laid out the steps he says investors should take before it does →

What the books say

Against that, here is what has leaked out about the company’s finances. In the first quarter of 2026, OpenAI took in $5.7 billion of revenue and burned $3.7 billion of cash, both roughly three times the year before, and finished the quarter with $73 billion in cash and securities after a record funding round. It had forecast cash burn of $25 billion for 2026 and $57 billion for 2027.⁷ Its 2025 revenue was $13 billion, its inference costs rose fourfold last year, its adjusted gross margin fell to 33% from 40%, and it is targeting about $600 billion of compute spending through 2030.⁸

The burn is covered the way it has always been covered: by the next round. On Thursday, the same day the Nasdaq sold off, OpenAI was reported to be raising up to $30 billion at a $1.25 trillion pre-money valuation, led by Coatue and investors from the UAE. Its last round, in March, valued it at $852 billion. It filed confidentially for an IPO in June but is pursuing another private round first.⁹ Revenue is growing fast. So is the bill.

The backstop that was not a backstop

The market has already seen what happens when someone at the center of this says the quiet part out loud. Last November, OpenAI’s chief financial officer, Sarah Friar, told a Wall Street Journal event that she wanted government help in the form of “the backstop, the guarantee that allows the financing to happen.” Within hours she had posted on LinkedIn that “OpenAI is not seeking a government backstop for our infrastructure commitments.” Altman followed on X: “We do not have or want government guarantees for OpenAI datacenters.” The White House AI czar, David Sacks, answered as well: “There will be no federal bailout for AI.”¹⁰

Three denials inside two days is a strange way to settle a question nobody was supposed to be asking. The reason it was asked is that the financing is the issue. Every one of those commitments depends on a private company being able to raise, borrow or earn the money to honor it.

What a code red actually meant

Then came the memo. In early December, an internal note seen by The Information and The Wall Street Journal showed Altman declaring a “code red” at OpenAI. Privately, he acknowledged that ChatGPT had fallen behind Google. “We know we have some work to do, but we are catching up fast,” he wrote, and the memo said the company was holding back the introduction of ads to ChatGPT.¹¹ He called it “a critical time for ChatGPT,” and the advertising plans were delayed so the company could focus on the core product. Chief operating officer Brad Lightcap said later: “It’s a way of forcing company focus.”¹²

✓ Trusted partner See the research behind this story →

Read plainly, the code red was about product, not money. But at this company the two cannot be separated. The ads that were pushed back are revenue, and revenue is what pays for data centers, and the rival that forced the memo is one of the companies reporting on Wednesday.

The warnings, and the loop

This month the warnings moved from bloggers to the people who run money. On October 7, Ray Dalio told a conference in Singapore that AI is a “classic bubble” nearing the point where it bursts, with debt-funded investment meeting rising rates. “I think we’re close to that,” he said.¹³ Two days later Joachim Klement of Panmure Liberum said: “My core conviction is that the AI bubble will either burst in 2027 or in 2028, so sometime in the next two years.” His year-end 2027 target for the S&P 500 is 5,000, about 36% below where it stood.¹⁴ On October 13, the IMF’s Global Financial Stability Report flagged the arrangements in which AI firms act as one another’s investors, financiers and customers, and warned that trouble at one company could spread quickly to the others.¹⁵

That is a description of the structure above, with one private company in the middle of it.

There is a fair case on the other side. OpenAI’s revenue is tripling, it has $73 billion in the bank, and sophisticated investors are still lining up at a $1.25 trillion price. Many of the commitments are letters of intent that can be stretched or shrunk. Meta’s results beat expectations even as its stock fell. Nobody can say when a boom like this ends, and one strategist above puts the date in 2027 or 2028, not this week, while Dalio says only that it is close.

But consider what Thursday revealed. “Every index on the planet is basically an AI index,” said Matt Miskin of Manulife John Hancock. IBM’s chief executive, Arvind Krishna, said customers “are also getting worried if some of the AI spending is completely out of control with no real return.”¹⁶ When Alphabet and Microsoft report on Wednesday, the number beneath their numbers is a private company’s ability to keep paying. An investor does not have to predict the crash to want a plan for it, and the plan starts with knowing where the weight sits.