The AI boom has a lot riding on a handful of companies spending billions on chips, cloud computing, and data centers. OpenAI is one of the biggest customers keeping that machine running. So when its reported annualized revenue came in nearly $20B below an earlier figure, investors started questioning just how much the AI industry can afford to spend.
Where did the missing $20B go?
OpenAI reported roughly $50B in annualized revenue at the end of September, well below the $68B figure that had circulated just weeks earlier.
The difference came down to accounting, not disappearing customers. The earlier figure included gross revenue from partners to allow comparisons with Anthropic, while OpenAI typically reports revenue on a net basis.
And business is hardly standing still. OpenAI's Q3 investor presentation showed annualized revenue growing 77%, with enterprise revenue climbing 107%. Still, investors have plenty of questions about the $852B valuation ahead of its planned 2027 IPO.
Oracle has the most riding on OpenAI
Oracle has made one of the biggest bets on OpenAI's future, signing a $300B cloud agreement spanning five years. That single deal represents roughly 45% of Oracle's $664B backlog.
The company is spending heavily to deliver on those commitments, with quarterly capital expenditures reaching $28.5B and free cash flow running $5.4B in the red. Investors are already demanding more protection against its debt, pushing Oracle's credit default swaps from 70 to 200 basis points.
CoreWeave has plenty at stake, too. The cloud provider rents powerful graphics chips to AI companies and has $22.4B in OpenAI contracts. But with $5.13B in 2025 revenue and continuing losses, it needs those contracts to translate into profitable business.
Microsoft, Amazon, Alphabet, and Meta also took hits, although their declines were smaller.
hipmakers aren't getting a free pass
Nvidia may not depend on OpenAI directly, but the cloud providers buying its chips certainly depend on AI spending. If those companies slow their data center expansion, Nvidia could eventually feel the squeeze.
Broadcom is developing custom chips for OpenAI, while Advanced Micro Devices is another name exposed to the industry's spending plans.
For now, the chip business is still delivering. Samsung and TSMC reported record profits and sales, suggesting demand for AI hardware hasn't lost its momentum.
The bigger problem is who's paying
OpenAI and Anthropic account for more than half of roughly $2T in cloud commitments. Neither company is profitable, yet both are central to the spending plans of some of the world's biggest technology companies.
There's another wrinkle. Around 70% of US data center capacity was built for training AI models, and executives at both labs have discussed slowing model development for safety reasons. Any slowdown could leave cloud providers with expensive infrastructure and less demand than expected.
SoftBank also ranks among the stocks exposed to concerns about OpenAI's revenue. Oracle faces particularly concentrated risk given the size of its cloud commitment.
The initial selloff didn't last long, with US tech futures rebounding the following day. But the questions surrounding AI spending aren't going away.
OpenAI confidentially filed its IPO prospectus in June and is reportedly seeking another $30B in funding. A public listing could finally give investors a clearer look at its finances.
Until then, the companies building the AI boom are making enormous bets on how much two private, unprofitable businesses will spend. And investors are still waiting to see whether those bets will pay off.
