Meta locks in Broadcom for MTIA chips through 2029

Meta and Broadcom just made their AI chip relationship official — and long-term. The two companies announced a multi-year, multi-generation partnership to co-develop and deploy MTIA (Meta Training and Inference Accelerator) silicon through 2029, starting with the industry's first 2nm AI compute accelerator.

The initial rollout commits to over 1 gigawatt of compute — just the first phase of a planned multi-gigawatt infrastructure buildout. Broadcom isn't just fabbing chips here; they're deep in the stack, supplying XPU design, advanced packaging, and high-bandwidth Ethernet networking to eliminate cluster bottlenecks at Meta's scale. Meta is also planning four new MTIA chip generations on a 6-month release cadence, targeting AI inference across its apps serving billions.

Notably, Broadcom CEO Hock Tan is stepping off Meta's board to take an advisor role focused specifically on the custom silicon roadmap.

Why it matters: Meta is going all-in on custom silicon to reduce Nvidia dependence — this deal gives it a credible multi-year hardware roadmap at a scale few companies can match.