Meta Reportedly Builds 'Meta Compute' Cloud Unit to Rent Out Excess AI GPUs, Rattling Neocloud Stocks
Bloomberg reports Meta is standing up a cloud business under its 'Meta Compute' initiative to sell spare AI capacity; CoreWeave and Nebius shares fell 10.8% and 12.4%.
Editor's Note ·
- Clarification:
- The precise market figures the article attributes to Cloud Computing News -- Meta shares up 8.8%, CoreWeave down 10.8% and Nebius down 12.4% on July 1 (two of which also appear in the summary), and the projected $145 billion in AI capital expenditure this year -- rest solely on that single outlet, whose page bot-blocked (HTTP 403) and could not be archived into the provenance snapshot. Editorial review confirmed these figures accurate against the live cloudcomputing-news.net page, but no durable archived snapshot supports them; the three verifiable snapshots (Ynetnews, Fierce Network, Sherwood News) confirm the report, the leadership names, the quotes and the sector selloff, but none of them state these specific percentages or the capex figure.
Overview
Meta Platforms is building a cloud business to sell the artificial-intelligence computing capacity it is not using internally, according to reporting by Bloomberg summarized by Ynetnews. The plans are being developed under an internal initiative called Meta Compute, which is focused on building and managing the company’s AI infrastructure, Cloud Computing News reported. The move would put the owner of Facebook and Instagram into direct competition with Amazon Web Services, Microsoft Azure and Google Cloud, as well as with the specialized “neocloud” GPU-rental firms that have grown up alongside the AI boom.
The report, which surfaced on July 1, sent Meta’s own shares higher while hammering the neocloud companies whose core business is renting out the same kind of capacity Meta would now sell.
What We Know
The effort is led by Santosh Janardhan, Meta’s head of infrastructure; Daniel Gross, a leader inside Meta Superintelligence Labs; and Meta president Dina Powell McCormick, according to Cloud Computing News, which cited people familiar with the plans.
Two business models are reportedly under discussion. One option would let developers access AI models hosted on Meta’s infrastructure, similar to Amazon’s Bedrock service, as reported by Ynetnews. Those hosted models would include Meta’s Muse Spark models, per Cloud Computing News, which noted that Meta unveiled Muse Spark in April. The second option would involve selling raw computing capacity directly, a model used by neocloud providers such as CoreWeave, according to the same Cloud Computing News report.
The market reaction split sharply along those lines. Meta shares rose 8.8% in New York on July 1, while shares of CoreWeave and Nebius fell 10.8% and 12.4% respectively following the report, according to Cloud Computing News. The selloff extended across the sector: Sherwood News reported that CoreWeave, Nebius, IREN and Cipher Digital were “getting slammed in the wake of this report,” reflecting investor concern that a hyperscaler with spare capacity to sell is a direct threat to firms whose entire business is providing exactly that.
The strategy would build on Meta’s enormous infrastructure spending. The company has projected spending of as much as $145 billion on AI-related capital expenditure this year, including data centres and GPUs for AI workloads, Cloud Computing News reported. Chief executive Mark Zuckerberg had already signaled openness to the idea at Meta’s shareholder meeting in May, saying that selling excess compute was “definitely on the table,” according to Cloud Computing News.
What We Don’t Know
The plans remain early and unconfirmed. Meta has not publicly announced Meta Compute as a product, and the reporting describes options still “under discussion” rather than a launched service, per Ynetnews. It is not clear which of the two models Meta will pursue, whether it will offer both, or on what timeline any offering would reach customers. Pricing, availability and the terms under which developers could access Muse Spark or raw capacity have not been disclosed in the reporting.
Analysis
Whether Meta can succeed depends heavily on where it chooses to compete. Sid Nag, president and chief research officer at Tekonyx, told Fierce Network that “Meta already has hyperscale economics, world-class networking, custom silicon investments and one of the largest GPU footprints on the planet.” But he argued the company should be selective about the fight it picks: “Rather than chasing mainstream enterprise workloads, Meta’s most successful path could be selling bare metal compute capacity and AI infrastructure to sophisticated buyers,” Nag said, adding that the market “is likely large enough to support a fourth hyperscaler, but not a fourth full stack cloud.”
That distinction matters for the neocloud names that sold off. If Meta targets the same sophisticated, capacity-hungry buyers that firms like CoreWeave and Nebius serve, it does so with a balance sheet and GPU footprint few can match. The stock moves suggest investors have already priced in that risk, even though the offering itself is still, by the reporting’s own account, a set of options rather than a shipping product.