Meta Is Quietly Building a Cloud Business to Sell Its Excess AI Compute, Bloomberg Reports
Meta Platforms is developing plans for a cloud infrastructure business that would sell access to AI computing power and Meta's own AI models, according to a Bloomberg report published Wednesday citing people familiar with the matter. The move would put Meta in direct competition with Amazon Web Services, Microsoft Azure, and Google Cloud, three companies it has spent the past two years leaning on for its own AI buildout.
Investors liked what they heard. Meta shares jumped as much as 8.6% in premarket trading before paring some of the gains, according to Reuters.
What the Plan Actually Looks Like
One option under consideration would let outside developers access AI models hosted on Meta's own infrastructure, an approach similar to Amazon's Bedrock service. Under that model, Meta would run the data centers and chips powering the models, including its Muse Spark models, and charge developers for access. A second option involves selling raw computing capacity directly, the model used by neocloud providers like CoreWeave.
The effort reportedly sits inside Meta Compute, an internal initiative focused on managing the company's AI infrastructure, and is being led by infrastructure chief Santosh Janardhan, Daniel Gross of Meta Superintelligence Labs, and Meta President Dina Powell McCormick.
Why Meta Suddenly Has Compute to Spare
Meta has poured hundreds of billions of dollars into data centers and chips as CEO Mark Zuckerberg has made AI superintelligence a defining company priority, signing major capacity agreements with CoreWeave, Google, and Oracle along the way. Big Tech firms overall are expected to spend more than $700 billion on AI infrastructure in 2026, up from roughly $400 billion in 2025, a buildout pace we track in our AI industry statistics guide.
That spending has drawn investor scrutiny over when and how it turns into revenue. A cloud business gives Meta a second path to monetizing infrastructure it built primarily for its own products. Zuckerberg flagged this possibility back in May at Meta's annual shareholder meeting, saying selling excess compute was "definitely on the table" and that other companies approach Meta "almost every week" asking to buy access to its AI models or spare capacity.
Notably, Meta's own compute position hasn't always been comfortable. Earlier this year, Google restricted Meta's access to its Gemini models after being unable to meet Meta's demand, a shortfall that delayed some internal AI work and prompted Meta to ask employees to limit their AI automation token usage. That Meta is now positioning itself to sell compute to others, after recently running short itself, says something about how fast the capacity picture is shifting across the industry, a shift reflected in our own Meta AI statistics breakdown.
What This Means for Businesses Evaluating Cloud AI
From my time advising companies on AI infrastructure decisions, the number of viable providers matters as much as the pricing on any single one. A fourth major hyperscaler entering the AI compute market, even in early form, is good news for buyers who currently negotiate against three entrenched players. Elon Musk's xAI has already shown what this playbook can generate, with SpaceX renting access to its Memphis data center to Anthropic and striking a deal with Google. Bloomberg Intelligence estimates that strategy alone could help xAI generate more than $50 billion in revenue by 2028.
For now, Meta's plans remain in development and unconfirmed by the company. Businesses shouldn't restructure vendor strategy around a report, but it's worth watching as a signal that the AI compute market is about to get more competitive, not less.
Cut Through the Noise
What is Meta planning according to Bloomberg?
Meta Platforms is developing plans for a cloud infrastructure business that would sell outside customers access to AI computing power and Meta's own AI models, competing directly with AWS, Azure, and Google Cloud. One option would offer developer access to Meta-hosted models similar to Amazon's Bedrock service; another would sell raw compute capacity.
Why is Meta doing this now?
Meta has committed hundreds of billions of dollars to AI infrastructure as part of CEO Mark Zuckerberg's superintelligence push, and investors have questioned how that spending will generate returns. Selling excess compute and model access gives Meta a second revenue stream from infrastructure it already built. Zuckerberg said in May that the option was "definitely on the table."
Who is leading the effort at Meta?
The initiative reportedly sits inside Meta Compute and is being led by infrastructure chief Santosh Janardhan, Daniel Gross of Meta Superintelligence Labs, and Meta President Dina Powell McCormick, according to Bloomberg's sources.
