Meta and Anthropic are reportedly discussing a $10 billion compute lease that would launch Meta into cloud services
The talks, first reported by the New York Times, would give Anthropic badly needed GPU capacity while opening an entirely new revenue line for Meta.
What matters
- Meta is in early-stage talks to lease AI computing power to Anthropic in a deal worth up to $10 billion over two years, per the NYT.
- Anthropic proposed the deal in June 2026; terms remain in flux and either company could opt out early.
- The arrangement would launch Meta into the cloud infrastructure business, competing with AWS, Azure, and Google Cloud.
- Meta expects $125–145 billion in 2026 capex and hired former AWS EC2 chief Dave Brown to build out the cloud effort.
- Anthropic has been stacking compute leases — including SpaceX Colossus 1 (~$1.25B/month) and TeraWulf Kentucky (~$19B over 20 years) — as it races toward a reported October 2026 IPO.
What happened
Meta is in early-stage talks to lease computing power from its AI data centers to Anthropic in a deal that could be worth up to $10 billion over two years, the New York Times reported on July 17, 2026. CNBC confirmed the discussions.
Anthropic reportedly proposed the arrangement in June 2026. Under the terms as currently discussed, Anthropic would pay Meta in monthly increments over a two-year period, and either company could opt out of the agreement early. The specifics remain in flux, the people familiar with the discussions said, and the talks may not result in a final deal.
The report follows earlier Bloomberg reporting that Meta was eyeing an entry into the cloud services business. Meta has not publicly confirmed the discussions.
Why it matters
For Meta, leasing data center capacity to outside companies would be an entirely new business line. The company makes the vast majority of its revenue from advertising, but it has been pouring unprecedented sums into AI infrastructure — expecting capital expenditure of $125 billion to $145 billion in 2026 alone. Selling excess compute would help justify that spending.
Meta CEO Mark Zuckerberg has previously signaled openness to the idea. On an earnings call last October, he said companies "regularly ask if we have compute they could buy from us at some premium," and in May he confirmed Meta was considering entering the cloud market. Meta also hired former AWS EC2 chief Dave Brown earlier this year, a signal that the infrastructure-as-a-service ambition is organizational, not just opportunistic.
For Anthropic, the talks underscore a persistent problem: despite reportedly approaching a $1 trillion valuation, the company cannot build compute capacity fast enough to meet demand for its Claude models. Access to AI chips remains a bottleneck, and Anthropic has placed usage limits on its most advanced models. The Meta discussions are just the latest in a series of compute leases Anthropic has assembled, including a roughly $1.25 billion-per-month arrangement with SpaceX to use the Colossus 1 data center's Nvidia GPUs and a reported ~$19 billion, 20-year lease with TeraWulf for a Kentucky facility.
The deal would also put Meta in direct competition with established cloud providers like Amazon Web Services, Microsoft Azure, and Google Cloud — a notable strategic shift for a company primarily known as a social media and advertising platform.
What to watch
- Whether the talks progress beyond preliminary discussions or fall apart — both companies can opt out early.
- Meta's formal launch of a cloud computing business line, which Zuckerberg has been telegraphing for months.
- Anthropic's broader compute strategy ahead of a reported October 2026 IPO targeting a ~$1 trillion valuation and $60 billion-plus raise.
- Whether other AI labs seek similar lease arrangements with Meta as GPU scarcity persists.
What to do next
Developers
Monitor whether Meta opens a public compute or API offering, as a new cloud provider could eventually expand available GPU capacity for model training and inference.
If Meta formally launches a cloud business, developers may gain an alternative to AWS, Azure, and Google Cloud for AI workloads.
Founders
Track Anthropic's compute lease strategy as a case study in scaling infrastructure through partnerships rather than solely owned data centers.
Anthropic's approach of leasing from Meta, SpaceX, and TeraWulf shows how AI startups can secure capacity without building everything in-house.
PMs
Assess whether a Meta cloud offering would create new pricing pressure or bundling opportunities in the AI infrastructure market.
Meta entering cloud services could shift competitive dynamics for compute pricing and availability, affecting product roadmaps that depend on GPU access.
Investors
Evaluate Meta's capital expenditure trajectory ($125–145B in 2026) against the potential revenue from a cloud business, and watch for signs of a formal launch.
A $10B Anthropic deal would be a meaningful early data point for Meta's ability to monetize its massive AI infrastructure investment outside advertising.
Operators
Watch for Meta's cloud business structure — pricing, SLAs, and whether it targets enterprise or AI-lab customers specifically.
If Meta Compute launches, infrastructure teams may need to evaluate it alongside existing cloud providers for workload portability and cost optimization.
Testing notes
Caveats
- This story involves reported business deal talks that are not yet finalized and no product or service is available to test.