Monday, August 3, 2026

Chips & Hardware

Meta to start producing custom AI chips in September

Meta is reportedly on track to begin production of its latest custom AI-specific chips in September, aiming to reduce its reliance on external chipmakers.

Server rack filled with 72 Meta MTIA 400 chips connected by purple cables in a data center
Photo: Meta

Meta is on track to start making the latest versions of its AI-specific chips in September, Reuters reported, citing an internal memo. At least one of the chips completed its testing phase in about six weeks, according to the memo.

Meta is working with Broadcom on the design of the chips, though manufacturing will run through Taiwan Semiconductor Manufacturing Company (TSMC, a major semiconductor foundry). The company is also sourcing components from several other suppliers:

  • Samsung for RAM
  • Sandisk for storage
  • Sumitomo Electric for fiber-optic equipment

These chips are among the four new chips Meta detailed in March under its Meta Training and Inference Accelerator (MTIA) program — the custom AI chip effort Meta has run since 2023. The design takes a modular, chiplet-based approach: “Each MTIA generation builds on the last, using modular chiplets, incorporating the latest AI workload insights and hardware technologies, and deploying on a shorter cadence,” Meta wrote when it detailed the chips. Meta said in April that it expects capital expenditures of between $125 billion and $145 billion this year, much of it tied to its AI buildout.

Meta is also striking data center and power deals across the world, and plans to deploy 7 gigawatts of compute this year, doubling that amount next year, according to Reuters, which cited the memo. Beyond MTIA, Meta signed a deal with ARM last year to secure compute for its recommendation systems, alongside a multibillion-dollar deal with AMD for GPUs and a multibillion-dollar deal with Amazon for homegrown CPUs. Meta joins other companies moving away from dependence on Nvidia: OpenAI is building an inference processor with Broadcom, Anthropic is considering developing its own chips with Samsung, and Google develops its own chips for AI training and inference.

Why it matters

By manufacturing its own silicon, Meta aims to lower its GPU costs amid an unprecedented component shortage and shift the balance of power away from dominant chipmakers like Nvidia and AMD.