Meta Plans to Sell AI Compute to Rivals Amid Investor Pressure

Meta Announces Plans to Sell AI Compute to Rivals

Meta is preparing to sell its massive AI computing capacity to other firms, marking a dramatic strategic pivot for the social media giant. Facing growing Wall Street skepticism over its soaring capital expenditures, Mark Zuckerberg revealed on the company’s late-July earnings call that Meta would start “selling compute directly” to large AI players, effectively launching a cloud business aimed at the booming neocloud market.

The shift follows a difficult second quarter in which Meta’s revenues missed expectations and free cash flow plunged 91%, largely because of the company’s own AI infrastructure build-out. Unlike cloud rivals Amazon, Microsoft, and Google, which can point to customer contracts that turn data centre spending into visible revenue, Meta has until now used its gigawatt-scale campuses exclusively to power internal services like Facebook and Instagram. With annual capex projected at $145 billion, investors are demanding a quicker path to returns, and the company acknowledged it does not yet monetize AI applications meaningfully.

Zuckerberg described a two-pronged strategy: first, providing cloud access to Meta’s own AI models, pitting it against labs like OpenAI and Anthropic; second, renting out raw computing power to other AI-intensive companies. Although Meta has not confirmed details, Reuters reported that a deal with Anthropic could be worth $10 billion over two years. Analysts say the move targets the hottest segment of the data centre market—neoclouds such as CoreWeave, Core Scientific, and Lambda—which grew 223% year-over-year in the fourth quarter.

The entry is complicated by Meta’s existing relationships: it is one of the largest neocloud customers, with a $21 billion contract with CoreWeave itself. Some observers frame the pivot as a retreat from consumer AI ambitions that have yet to deliver blockbuster returns, while others see a necessary evolution to placate investors. Either way, Meta’s decision injects a powerful, capital-rich competitor into the AI cloud space.

How Meta’s AI Cloud Pivot Reshapes Its Business and the Neocloud Landscape

The Spending Disconnect and Wall Street’s Demand for Revenue

Meta’s capex bill has ballooned to $145 billion this year, dwarfing any demonstrable AI revenue stream. Because it lacks an external cloud division, every dollar of infrastructure spending leans on the advertising business for a return. After Q2 results missed estimates and free cash flow cratered, investors sent the stock down more than 8% from the start of the year. Analyst Josh Gilbert of eToro captured the core tension: “Meta is spending like a hyperscaler without a hyperscaler’s business model.” Selling compute directly offers investors a quantifiable near-term revenue line, a crucial circuit-breaker as the firm continues to invest heavily.

Why the Neocloud Market Is an Attractive Target

The neocloud sector posted 223% year-over-year revenue growth in Q4, according to Synergy Research Group, with more than 190 operators vying for business. Large players like CoreWeave have anchored themselves by selling blocks of GPU capacity to hyperscalers. Meta is positioning itself to do the same, initially targeting AI firms that need massive compute. Zuckerberg told analysts the company receives “a lot of offers for compute at a significant premium over what we paid for it,” suggesting the underlying demand could quickly convert into meaningful revenue. If the reported $10 billion Anthropic deal materializes, it would instantly become one of the largest neocloud contracts.

Meta’s Tightrope: Partnering with Providers It Now Competes Against

Meta’s $21 billion contract with CoreWeave—expanded from $14.2 billion in April—places it in the unusual position of being both a top customer and a direct competitor to the neoclouds. The company also buys capacity from AMD and Crusoe. While the two-pronged strategy may allow Meta to leverage existing relationships for mutual benefit, it risks unsettling partners who may see the pivot as a threat to their own revenue. Evercore ISI’s Mark Mahaney noted the move is “not going after the cloud businesses” like AWS or Azure, but squarely into the neocloud niche where Meta’s excess capacity fits.

Strategic Coup or Capitulation? Analysts Weigh In

Opinions diverge. Some see a pragmatic monetization of sunk costs, especially after SpaceX’s AI division (formerly xAI) began selling surplus compute and reportedly generates $20 billion annually from similar deals. Others, like Primary Venture Partners’ Brian Schechter, argue the shift reflects a retreat: “Being able to monetize their compute after a missed training run shows how compute can function more like a commodity.” While Meta insists opportunity, not necessity, drives the decision, the timing—directly after disappointing earnings and falling share prices—strongly suggests investor anxiety was the catalyst.

What Investors, Competitors, and AI Firms Should Watch For

For Meta’s investors: Focus on the next earnings call for specifics on cloud revenue segmentation and progress on the unconfirmed Anthropic deal. A confirmed $10 billion two-year contract could materially offset the $145 billion annual capex burden and would likely boost sentiment.

For neocloud operators (CoreWeave, Core Scientific, etc.): Meta’s entry introduces a competitor with immense balance-sheet strength and existing infrastructure. Prepare for potential margin compression on large-scale compute contracts, especially given Zuckerberg’s statement that received offers carry a premium over cost.

For AI firms seeking compute: Meta may offer an attractive alternative to existing neocloud suppliers, potentially lowering rates. However, due diligence on service reliability and sales support is warranted, as Meta must build a third-party sales organization from scratch.

For Meta’s management: Executing the two-pronged model demands building a cloud sales and support team rapidly. Balancing its role as both a leading customer and a competitor to the neocloud ecosystem will require careful relationship management to avoid losing access to critical capacity from partners like CoreWeave.