Meta’s latest move in artificial intelligence is not a new model or app, but a giant, largely hidden bet on cloud infrastructure that could reposition the social media giant as a key utility provider to rival tech firms.

Early signals of a mega-deal

On July 17, reports emerged that Meta and Anthropic were in advanced discussions for "up to $10bn" in data centre capacity, a potential long-term arrangement that would see Meta provide the startup with massive computing power.¹ The talks come as Meta ramps up infrastructure spending, with about $145 billion reportedly earmarked for data centers and related hardware to power its AI ambitions.¹

Details of the proposed arrangement

Further reporting the same day indicated the structure of the deal: Meta is "reportedly in talks to lease computing power" to Anthropic, in an agreement that could be worth $10 billion over two years, paid in monthly installments.² Rather than a one-off investment, this would make Anthropic a recurring, large-scale customer of Meta’s infrastructure.

Anthropic’s multi-cloud strategy

For Anthropic, the Meta deal would be one piece of a broader plan to secure vast computing resources. The AI company "plans to invest $50 billion in building out its own data centers" while already having struck multibillion-dollar agreements for capacity from SpaceX and TeraWulf.² Adding Meta to that roster would deepen its diversified, multi-provider approach to AI infrastructure.

How it could reshape the AI cloud race

From Meta’s perspective, leasing capacity to Anthropic would effectively launch it into the cloud business alongside established hyperscalers, turning surplus or future data center buildout into a revenue-generating platform. For Anthropic, the deal would lock in short- to medium-term compute at scale, reducing dependence on any single provider.

Together, the talks highlight a shifting AI landscape in which social platforms morph into cloud utilities and frontier AI labs become anchor tenants in an increasingly crowded infrastructure race.