Meta's Cloud Ambitions Send CoreWeave and Nebius Shares Tumbling

Shares of AI cloud upstarts CoreWeave and Nebius plunged Tuesday on a report that Meta is preparing to sell its own AI computing capacity — turning one of the industry's biggest buyers of compute into a potential competitor of the companies it pays to supply it.

Tyler Grant July 1, 2026 at 10:30 PM

The AI infrastructure trade wobbled Tuesday on a simple, unsettling idea for investors: what if the biggest buyers of computing power start selling it too?

The selloff

CoreWeave closed down about 14 percent and Nebius Group fell about 17 percent, MarketWatch reported, while Meta rose roughly 9 percent as investors reframed its enormous spending on AI hardware as a possible new revenue stream. The trigger was a Bloomberg report that Meta is preparing a service to rent out surplus AI computing capacity — and, potentially, access to its own AI models — to outside developers and companies. Stock moves are as of July 1.

What Meta is reportedly planning

Meta has guided to capital spending of well over $100 billion in 2026, building out data centers and AI chips at a scale that analysts say rivals the big cloud providers. Selling spare capacity to outside customers is "definitely on the table" if the company builds more than it needs, chief executive Mark Zuckerberg has said, as CNBC noted. Such a move would put Meta alongside Amazon Web Services, Microsoft Azure and Google Cloud in the market for enterprise AI infrastructure.

Why the neoclouds fell hardest

CoreWeave and Nebius belong to a class of companies Wall Street calls "neoclouds" — pure-play providers that rent out GPU computing tailored for AI. Both carry large contract backlogs, but both also depend heavily on a small number of anchor customers, a concentration that becomes a risk if a giant like Meta enters the market and eventually buys less from them or competes for the same clients.

Opportunity or threat?

Analysts split on the news. Some argued that monetizing spare compute could actually reduce Meta's earnings risk by turning idle capacity into near-term revenue, helping justify spending that markets had begun to question. For the neoclouds, the read was less comforting: if the largest consumers of AI compute start selling surplus capacity instead of only buying more, the market open to specialist providers narrows.

The selloff rippled beyond the two names, dragging chip-equipment and memory stocks lower on fears of a broader rethink of AI capital spending. For now, the contracts and backlogs remain in place — but Tuesday showed how fast confidence in a capital-hungry, customer-concentrated business can turn when the shape of the competition changes.