Meta Platforms is reportedly building a cloud infrastructure business to sell its surplus AI computing resources to outside customers. If launched, it could become a "fourth hyperscaler" competing directly with Amazon Web Services (AWS), Microsoft Azure, and Google Cloud.
July 1, 2026 · Meta
Meta weighs a cloud business to sell its excess AI compute
Reports point to an early-stage internal effort dubbed "Meta Compute" — renting out spare capacity from a record-scale data center buildout, and potentially competing with AWS, Azure, and Google Cloud.
$125–145B
2026 capital expenditure — an expected record high
up to 5GW
Scale of the "Hyperion" AI data center cluster
+7–8.6%
Meta shares in premarket after the report
The lone hyperscaler-without-a-cloud
Of the four major U.S. AI infrastructure players, Meta is the only one without a cloud business today.
Offering 1
Model access sales
Host Meta's own models (incl. Muse Spark) and charge developers via API.
Reference model → AWS Bedrock
Offering 2
Raw compute rental
Direct access to data centers and chips — a GPU-specialized "neocloud."
Reference model → CoreWeave
Still on the table, not launched
Zuckerberg called cloud entry "definitely on the table," noting outside firms ask about buying compute "almost every week." No name, timing, pricing, or chip lineup is confirmed.
• Named clusters cited: Prometheus , Hyperion
Bull case
Turns a massive AI cost center into a revenue source, with strategic leverage while compute supply stays tight.
Caution case
Selling "excess" may be the flip side of overbuilding. Matching hyperscalers' reliability, breadth, and multi-year contracts will take time.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…