Meta aims to lease its excess AI computing capacity, and Wall Street is supportive of this concept.

      A reported strategy to sell excess computing power would position Meta in competition with AWS, Google Cloud, and Azure, providing investors with a reason to overlook its expenditures. Meta has spent the last two years acquiring AI computing resources, and it now seems to be exploring the option of selling some of that capacity.

      On Wednesday, Bloomberg reported, referencing sources familiar with the situation, that the company is developing a cloud business to monetize its surplus AI capacity. Investors responded positively to the news amid ongoing concerns regarding Meta's spending and its potential returns.

      According to the report, the plans are still in the early stages and may evolve. Meta is considering the specifics of the offering rather than deciding whether to move forward. One possibility is to sell access to AI models hosted on its own infrastructure, similar to Amazon’s Bedrock. Another option is to offer raw computing capacity, a model that new cloud providers like CoreWeave have built successful businesses around. Either approach would pit Meta against Amazon Web Services, Google Cloud, and Microsoft Azure, the current leaders in the market it would be entering.

      This initiative is reportedly organized under a new division called Meta Compute, which is headed by Santosh Janardhan, the company’s infrastructure chief, alongside Daniel Gross from Meta Superintelligence Labs and Meta president Dina Powell McCormick. The rationale is quite clear.

      Meta has projected capital expenditures of $115 billion to $135 billion in 2026, an enormous investment in chips, land, and power, making a cloud business one of the few avenues to generate revenue from idle resources instead of incurring a loss.

      The market reacted positively right away. Meta's shares rose by over 10% following the report, a notable increase for a stock that had been struggling, down nearly 15% as of the previous day and trailing the S&P 500 amid concerns over its AI investments.

      A credible path to recouping investments in infrastructure, even if still hypothetical, seemed to boost investor sentiment. Meta is not the first to recognize this opportunity. SpaceX has been leasing extra capacity from xAI's Memphis data center to Anthropic, a partnership that Bloomberg Intelligence predicts could generate over $50 billion by 2028 and $100 billion by 2030.

      This trend is becoming commonplace in the industry: create far more computing power than currently needed, betting on future demand, and rent out the surplus to offset costs in the interim.

      For Meta, the surplus is significant and increasing. The company boasts a 2,250-acre hyperscale campus in Louisiana, a data center under construction in the American Midwest, and additional external agreements, including new capacity from Crusoe amounting to about 1.6 gigawatts across two locations.

      However, this demand has faced challenges elsewhere, with Google recently limiting Meta’s access to its Gemini models due to its own capacity constraints.

      There is an irony in a company that has been striving for computing resources now preparing to sell them, highlighting the uneven nature of this expansion. Capacity is delivered in large, indivisible quantities, scheduled according to projections rather than current needs, often leaving even the most resource-hungry buyers with more than they can immediately utilize.

      Selling the surplus is how new cloud providers, and now seemingly their clients, plan to balance the economics. Contracts like Jane Street’s $6 billion deal with CoreWeave illustrate the significant funds flowing through this emerging layer.

      Currently, this remains a report rather than an official product. Meta has not confirmed the plan, and no details regarding pricing or launch timing have been announced. Those discussing the initiative emphasize that the strategy could still change. However, the incentive is clear.

      With over $100 billion committed annually to infrastructure, finding buyers for unused capacity evolves from being a side endeavor to becoming a critical necessity.