Meta Platforms is drawing up plans for a cloud business to sell spare AI computing power and hosted AI models to outside customers, according to Bloomberg on July 1, 2026. The move would compete directly with Amazon Web Services, Microsoft Azure and Google Cloud.
Quick Summary – TLDR:
- Meta Compute, a new cloud infrastructure business, will sell AI computing power the company does not use internally, per Bloomberg.
- Three executives lead the effort: infrastructure chief Santosh Janardhan, Daniel Gross of Meta Superintelligence Labs, and Meta President Dina Powell McCormick.
- Meta shares rose as much as 8.6% in premarket trading on the report before paring gains.
- Meta’s 2026 capex guidance is $125 billion to $145 billion, raised on higher AI infrastructure costs, per its Meta earnings release.
- “Definitely on the table” was how CEO Mark Zuckerberg described selling excess compute in May, if Meta overbuilds data center capacity.
What Happened?
Meta Compute, the new cloud unit, would resell excess AI computing power and hosted models to outside customers, sitting alongside 2026 capex guidance of $125 billion to $145 billion per Meta’s Q1 earnings release, and shares climbed as much as 8.6% in premarket trading on the news.
Meta is considering two distinct paths into the cloud market, according to people familiar with the matter. One option would let outside developers access AI models hosted on Meta’s own infrastructure, similar to Amazon’s Bedrock service, with Meta operating the underlying data centers and chips and charging for access.
That hosted-model tier would reportedly include Meta’s Muse Spark models. The second option is selling raw computing capacity directly, the model used by “neocloud” providers such as CoreWeave. The plans remain in development and a Meta spokesperson declined to comment.
Meta has been rushing to secure expensive data centers and chips to fuel its own AI ambitions, and the cloud business is designed to monetize any resulting surplus rather than let it sit idle. The buildout sits alongside broader hiring pushes, per Meta employee count data on the company’s infrastructure and AI teams.
$META is reportedly developing a cloud business to sell access to excess AI compute, per Bloomberg.
— Wall St Engine (@wallstengine) July 1, 2026
The internal initiative is called Meta Compute.
The plans being considered:
AI model access hosted on Meta infrastructure, similar to AWS Bedrock
Raw AI compute capacity,… pic.twitter.com/X2LzlT1DdQ
Zuckerberg Flagged the Option Months Ago
At Meta’s annual shareholder meeting in May, Zuckerberg said “almost every week” other companies approached Meta asking to buy access to its AI models or spare computing capacity at a premium. The Meta Compute initiative is not a sudden pivot but a formalization of that hedge, moving a shareholder-meeting contingency into an operating project with named leadership within roughly two months of the CEO’s public remarks.
Zuckerberg said:
That framing matters now because Meta Compute reportedly moves the contingency from a hypothetical into an active internal initiative with named leadership spanning infrastructure, frontier AI research and corporate strategy. A cloud arm would also give Meta a second revenue line beyond advertising.
Any hosted-model tier would put Meta up against the workforce and infrastructure scale of Google Cloud’s parent company.
Why Investors Reacted?
Meta shares climbed on the report, with gains ranging from nearly 6% to as much as 8.6% in premarket trading Wednesday, before paring back some of those gains as the session opened and traders weighed the reported cloud pivot against Meta’s advertising-heavy revenue mix.
The reaction reflects how AI infrastructure spend has become a scrutinized line item for Big Tech investors: outlays of more than $700 billion across the industry this year raise the question of what happens if demand for a company’s own AI products does not keep pace with the capacity it bought. Meta itself raised 2026 capex guidance to $125 billion to $145 billion on higher AI infrastructure costs, per its Q1 2026 earnings release.
SQ Magazine’s Takeaway
Against an industry spending more than $700 billion this year, the dual-track approach reported here reads as deliberate hedging rather than a single strategic bet. A Bedrock-style hosted API keeps Meta closer to the AI-product layer it already competes in with OpenAI and Google, while the neocloud, CoreWeave-style raw-capacity resale is a more commoditized fallback if hosted-model demand disappoints. Running both tracks in parallel signals that Meta is unsure which cloud customers actually want and would rather test both than choose prematurely.
The more consequential detail is timing. Zuckerberg described this as a contingency this past May; within a couple of months it already has a name, three named executives, and active internal planning.
That gap between “on the table” and an operating initiative is short, and it suggests Meta’s own capacity build has outpaced its ad-driven AI product needs faster than the company signaled publicly. Whether that becomes a durable cloud business or a face-saving way to monetize a data center bet that ran ahead of demand will depend on whether hosted-model customers materialize before the neocloud fallback becomes the default.
