Is Meta outsourcing its brain while keeping the wallet?
The question the AI market is asking since Engadget broke the news is simple: why would Meta — which operates one of the largest GPU fleets on Earth and just trained Llama 4 — need a data center deal with Anthropic?
What we know
According to Engadget, Meta is in advanced negotiations for a multibillion-dollar data center agreement with Anthropic, the AI startup founded by former OpenAI employees. The deal would give Anthropic access to Meta's computational infrastructure in exchange for profit-sharing or preferential access to Anthropic's technology. Exact figures haven't been disclosed, but sources close to the negotiations estimate the contract at several billion dollars.
What we don't know
It's unclear whether the deal involves only spare compute capacity (Meta has underutilized data centers during low-demand training periods) or a deeper strategic component — such as Meta gaining access to Anthropic's Claude model for internal products. It's also unknown whether other big techs, like Microsoft (Anthropic's largest investor), are involved.
Our take
The most underappreciated angle is what it reveals about industry dynamics: AI infrastructure is becoming a commodity that big techs are willing to monetize among themselves. If Meta — which poured tens of billions into its own GPUs — is renting capacity to an AI competitor, it's because the data center bill has gotten too steep even for them. For Anthropic, it's a lifeline: inference demand for Claude is growing faster than provisioning capacity. The risk? Over-reliance on a partner that could become a direct competitor tomorrow.
Source: Engadget
