Meta AI chips are on track to enter production in September as it works to control infrastructure costs and reduce dependence on scarce general-purpose processors. TechCrunch reported the timetable, citing a Reuters report based on an internal memo.
The planned chips are part of Meta’s wider custom-silicon programme. Meta says its MTIA family is designed for workloads including ranking, recommendations and generative-AI inference, and that the company is developing four new generations over a two-year period.
Meta’s official account describes MTIA as an inference-first strategy built around industry standards. The company said it already deploys hundreds of thousands of MTIA chips for inference workloads across its applications and advertising systems.
TechCrunch said the new production push comes during an unusually tight component market. Developing chips for Meta’s own workloads can give the company more control over capacity and cost, although the company will continue sourcing silicon from multiple industry suppliers.
Meta’s roadmap includes MTIA 300, which it said is already in production, followed by MTIA 400, 450 and 500 for broader workloads and future generative-AI inference. The official announcement does not say that every chip in the roadmap will begin production in September.
The verified news is therefore the reported September start for the latest versions and Meta’s stated multi-generation plan. It does not establish final performance results or a complete replacement for external GPUs, so claims about market leadership should wait for independent testing and later company disclosures.
