The Cracks in the AI Spending Story

The Cracks in the AI Spending Story

For the past two years, Wall Street has had a straightforward thesis on AI: whoever builds the most compute wins. Pour money into chips, data centers, and cloud capacity, and the revenue will follow. Today, that thesis got its first real stress test.

Shares of companies tied to AI infrastructure tumbled Tuesday after The Wall Street Journal reported that OpenAI has been missing its own internal projections for user growth and revenue. The selloff was swift and broad — Oracle dropped more than 6%, CoreWeave fell 7%, Nvidia slid 5%, and SoftBank — one of OpenAI’s largest backers — sank roughly 10% in Asia trading.

The numbers tell a story the market hadn’t priced in: the gap between AI capability and AI adoption may be wider than anyone wanted to admit.

What the Report Actually Says

According to the WSJ, OpenAI has recently fallen short of its own targets, and the concern isn’t just a quarterly blip. CFO Sarah Friar has reportedly warned colleagues that if revenue growth doesn’t accelerate, the company could struggle to fund future compute agreements — the massive, long-term data center contracts that underpin the entire AI buildout.

That’s a significant admission. OpenAI has been signing deals at a scale that would make a Fortune 500 CFO sweat. Its partnership with Oracle alone is valued at $300 billion over five years. Commitments like that only make sense if the revenue flywheel spins fast enough to keep up.

OpenAI pushed back hard. “This is ridiculous,” the company told CNBC. “We are totally aligned on buying as much compute as we can and working hard on it together every day.” The company also just closed a record $122 billion funding round at an $852 billion post-money valuation — not exactly the posture of a company in distress.

Why the Market Reacted So Strongly

The selloff wasn’t really about OpenAI’s quarterly numbers. It was about what those numbers imply for the entire AI infrastructure stack.

The AI boom has been driven by a single, powerful assumption: demand for compute is essentially unlimited. Chip companies, cloud providers, and neoclouds have all been investing on that basis. If OpenAI — the company that started this era — is struggling to grow revenue fast enough to justify its spending, the assumption starts to wobble.

Investors didn’t wait for confirmation. They sold first.

The Skeptics Have a Point — and So Do the Bulls

Mizuho’s Jordan Klein offered a measured read: “You would assume any slowing was known by the investors, right? If not, shame on OAI. How new could the update be as the round closed end March when the quarter would have ended. And it’s not even May 1. I highly doubt OAI fundamentals slowed that fast in under 30 days.”

He’s right to be skeptical of the panic. A $122 billion funding round closed barely four weeks ago. The investors who wrote those checks — among the most sophisticated in the world — had access to financials. If the shortfall were catastrophic, the deal wouldn’t have closed at an $852 billion valuation.

But the bulls shouldn’t be too comfortable either. Investor confidence at close doesn’t mean the trajectory is fine. Revenue shortfalls compound. And the cost of being wrong, when you’ve pre-committed to hundreds of billions in compute capacity, is not small.

The Competitive Subplot Nobody Is Talking About Enough

Buried under the selloff headlines is a structural shift that matters more in the long run: OpenAI no longer has the enterprise AI market to itself.

Anthropic has been gaining serious traction with corporate customers. Google’s Gemini models are picking up momentum across the enterprise. And companies are increasingly adopting a multi-provider strategy rather than betting on a single AI stack.

This is how maturing markets work. The pioneer captures attention and mindshare; the competition captures margin. OpenAI built the category. Now it has to win in it — against well-capitalized rivals who learned from watching it operate.

That’s a harder game. And it’s one that requires consistent revenue growth, not just impressive funding rounds.

The Real Question

Today’s selloff will probably look like noise in six months, or it will look like the moment the market finally started asking the right questions. We don’t know yet.

What we do know is this: the AI infrastructure buildout has been running on faith that demand would catch up to supply. That faith is still largely intact — but it is no longer unquestioned. And markets, once they start asking questions, don’t stop until they get answers.

The next few earnings cycles just got a lot more interesting. Apple’s Leadership Shift Signals a Hardware-First AI Future | Maya