Meta Platforms is placing what might be the largest corporate infrastructure bet in history: up to $145 billion in capital expenditures for 2026 alone, almost entirely directed at AI. The company currently sits at a market cap of roughly $1.51 trillion.
The scale of the commitment is staggering even by Big Tech standards. Meta plans to deploy 7 gigawatts of AI computing capacity in 2026, then double that to 14 GW by the end of 2027.
The Iris chip and Meta’s silicon ambitions
Central to Meta’s infrastructure push is its custom-built Iris chip, the company’s fourth-generation MTIA (Meta Training and Inference Accelerator). Manufacturing begins in September 2026, with Meta planning to release new chip iterations every six months through 2027.
Nvidia still dominates the AI accelerator market, and its GPUs remain the industry default. Meta is joining a growing club of hyperscalers designing their own silicon alongside Google’s TPUs and Amazon’s Trainium.
