Meta Platforms Inc. is in early negotiations to lease AI computing power to Anthropic in a deal valued at roughly $10 billion [1].
This potential agreement signals a shift in how AI infrastructure is distributed among competitors. As the cost of training next-generation models climbs, companies are increasingly looking toward infrastructure sharing to manage capital expenditures and operational scale.
The reported deal centers on Meta's extensive data-center capacity. According to reports on July 17, Meta is exploring ways to monetize its hardware assets by providing the necessary processing power for Anthropic to develop and refine its AI models [2].
Anthropic requires immense amounts of compute to maintain its competitive edge in the generative AI market. By leasing capacity from Meta, the company could potentially accelerate its development cycles without the immediate need to build out its own proprietary data centers at the same scale [3].
Rising AI compute costs have created a bottleneck for many developers. Meta's move to lease its resources suggests the company sees a viable revenue stream in acting as a compute provider for other AI labs [2]. This arrangement would allow Meta to offset the massive investments it has already made into GPU clusters, and power infrastructure.
Details regarding the duration of the lease or the specific hardware involved have not been disclosed. The talks remain in the early stages, and neither company has confirmed a final agreement [1].
“Meta is in early negotiations to lease AI computing power to Anthropic in a deal valued at roughly $10 billion.”
This deal reflects a growing trend of 'compute diplomacy' in the AI sector. By transitioning from a pure consumer of hardware to a provider, Meta can turn a cost center into a profit center. For Anthropic, the move reduces the barrier to entry for high-scale training, though it creates a strategic dependency on a primary competitor's infrastructure.
