Meta moves new in-house AI chip into September production
Meta’s next in-house AI chip is moving toward production as hyperscalers seek more control over the cost and architecture of their compute stacks.
Hyperscalers are increasingly trying to control more of their own silicon stack as AI infrastructure costs rise.
What happened
Meta plans to begin manufacturing a new in-house AI chip in September 2026.
The chip is part of Meta’s MTIA programme. Meta is working with Broadcom on design and TSMC on manufacturing.
The company is also aiming to expand overall computing capacity substantially as AI workloads grow.
Why it matters
Custom chips can optimise specific workloads and reduce dependence on general-purpose accelerators for every task.
Meta will still rely heavily on third-party hardware, but moving more silicon in-house can improve cost control and system-level optimisation.
The September production timeline turns the programme into a concrete manufacturing milestone rather than a research plan.
The bigger picture
AI infrastructure is becoming vertically integrated.
Cloud and platform companies increasingly design chips, software and data-centre systems together. Meta’s move shows that custom silicon is becoming a core strategic capability for the largest AI buyers.