Meta is set to begin manufacturing its latest AI-specific chips in September, with at least one chip having completed its testing phase, as the company aims to reduce GPU costs amid component shortages. The chips, developed under the Meta Training and Inference Accelerator (MTIA) program, are designed in collaboration with Broadcom and will be manufactured by Taiwan Semiconductor Manufacturing Company (TSMC), utilizing RAM from Samsung and storage from Sandisk. Meta plans to deploy 7 gigawatts of compute this year, doubling that next, as part of its $125 billion to $145 billion capital expenditure for AI efforts.
Hyperscalers are increasingly building custom AI silicon to reduce reliance on external GPU suppliers, shifting compute demand to foundry partners and memory providers.