AI market jitters: CoreWeave down 45%, Nvidia sliding, and the AI boom fragility debate

Summary

CoreWeave dropped 45% as market sentiment shifted against AI stocks. Nvidia fell 4% in a single day, down 7% last week and another 3% on Tuesday. Oracle took the biggest hit from its three-month high. The selloff followed a broad recovery on Monday after news of a possible end to the government shutdown, but AI stocks fell again Tuesday.

Companies have built exceptional products and are executing at hyperscaler-level infrastructure quality. The market ran away with that narrative and is now pulling back. Companies that doubled revenue are trading down.

Revenue to spending

OpenAI plans to spend $1.4 trillion over the next eight years but pulls in only $20 billion in annual revenue today. The company is projecting losses will swell to $74 billion in 2028. OpenAI lacks a clear business model to reach the hundreds of billions it needs within a few years to sustain spending growth.

Sam Altman felt compelled to defend the company on X last week, acknowledging the spending causes concern. He pointed to new consumer devices, robotics, and an AI cloud computing service as potential revenue sources. None of these currently exist.

All three proposed revenue streams are historically cash-incineration engines at early scale. Meta's hardware efforts, early hyperscaler spending, and robotics companies all absorb capital rather than generate it until they reach massive scale.

Talent exodus and liquidity

OpenAI employees who stayed two-plus years sold $6.6 billion of equity last month, with many hitting a $20 million cap. Morale is high, but turnover is rising. New hires report shock at how many Slack accounts get deactivated each day. The exodus is driven by one dominant force: liquidity. Anthropic's option liquidity is the worst among frontier labs, which explains why fewer people have left.

Dozens or perhaps a couple hundred researchers from OpenAI, XAI, Google, and DeepMind are founding companies in the current climate. When engineers can cash out tens of millions, they tend to leave. How to retain talent without overpaying them into independence remains unsolved.

The market is pricing for a long timeline

WSJ reporting notes the AI boom looks increasingly fragile. Investors accustomed to 12-month return horizons now face companies talking about plural years of major infrastructure investments still ahead. Generative AI requires massive data centers, state-of-the-art chips, and server racks that don't come together quickly. Early-stage AI investing demands patience on a timeline that conflicts with how venture and growth capital typically operates.

The contrarian case

Apple, which has sidestepped major AI infrastructure bets and remained cautious on the technology, is up 21% over the trailing 12 months. Microsoft, which owns a third of OpenAI and has invested heavily in AI, is only up 18%. The company that invested least in infrastructure and most in brand optionality outpaced the one that went all-in.