CoreWeave's $35B Spending Spree Sends Stock Tumbling 20%

$35 billion. That's how much CoreWeave plans to spend next year building AI infrastructure. Wall Street's reaction? A brutal 20% stock selloff that wiped out billions in market value.

The Debt-Fueled AI Gold Rush

CoreWeave's business model sounds simple: buy Nvidia chips with borrowed money, pack them into data centers, then rent computing power to Microsoft, OpenAI, and other AI giants. But simple doesn't mean safe.

The New Jersey-based company's $30-35 billion capex guidance for 2026 shattered analyst expectations of $26.9 billion. CEO Mike Intrator defended the spending blitz, calling it a response to "once in a generation" demand. Yet investors are asking a uncomfortable question: when does aggressive growth become reckless gambling?

A House of Cards Built on Few Customers

Here's what makes CoreWeave's bet particularly risky: most of its revenue comes from a handful of hyperscale customers. If Microsoft decides to build its own infrastructure, or OpenAI finds a cheaper alternative, CoreWeave's entire model crumbles.

Intrator insists the company's "backlog is enormous," but Barclays analysts warn that shares could "pause near these levels" as investors digest the spending implications. The math is stark: more spending means squeezed margins, higher debt loads, and greater vulnerability to economic downturns.

The Credit Crunch Wildcard

CoreWeave claims its cost of capital has dropped 300 basis points over the past year, saving $700 million. But credit markets are notoriously fickle. What happens when rates rise or lenders get nervous about AI infrastructure investments?

JPMorgan analysts aren't optimistic: "If we end up with heightened economic volatility, CRWV shares would probably suffer disproportionately due to risk-off positioning." Translation: CoreWeave could be the first casualty if the AI bubble deflates.

The Bigger AI Infrastructure Question

CoreWeave's struggles highlight a broader tension in the AI economy. Companies are spending unprecedented amounts on infrastructure, betting that demand will justify the costs. But what if it doesn't? What if the AI revolution takes longer than expected, or follows a different path entirely?