Meta is done waiting in line for someone else’s chips. The company plans to begin manufacturing its custom AI chip, codenamed Iris, starting in September, as part of a broader push to bring its total computing capacity to 14 gigawatts by 2027.

The company is targeting 7 gigawatts of computing capacity in 2026, then doubling it the following year.

The Iris chip and Meta’s silicon roadmap

Iris isn’t a one-off experiment. It’s one of four planned generations under Meta’s Training and Inference Accelerators program, known internally as MTIA. The company laid out the full four-generation roadmap back in March and expanded its partnership with Broadcom in April to help bring these designs to life.

TSMC, the Taiwanese foundry that manufactures chips for practically every major tech company on earth, will handle production. Broadcom is the design partner.

The chips are designed to handle the AI workloads that keep Meta’s platforms running: ranking and recommendation systems that decide what shows up in your Instagram feed, and the generative AI tasks powering everything from chatbots to content creation tools.

Meta’s capex guidance for 2026 is between $125 billion and $145 billion, mostly on data centers, GPUs, and custom silicon.

Why Meta is building its own chips

Meta isn’t the first tech giant to decide that relying on external GPU suppliers is a vulnerability rather than a convenience. Google has been building its own TPUs for years. Amazon has its Trainium and Inferentia chips. Microsoft has Maia.

Building custom chips doesn’t mean Meta will stop buying from Nvidia overnight. But each generation of MTIA chips that successfully handles production workloads reduces Meta’s marginal dependence on external suppliers.

What this means for investors and the broader market

Broadcom stands to benefit directly. The company has positioned itself as the go-to design partner for hyperscalers building custom AI chips. Its expanded collaboration with Meta, announced in April, adds another major revenue stream to a portfolio that already includes similar work with Google and others. The September production milestone will be a key test of whether the partnership can deliver at scale.

TSMC also wins here. The foundry’s position as the only manufacturer capable of producing cutting-edge chips at scale means it captures value regardless of whether the chip says Nvidia, Meta, or anyone else on it.

The September production date gives investors a concrete milestone to watch. If Iris chips start rolling off TSMC’s lines on schedule and perform well in Meta’s data centers, it validates the entire MTIA roadmap and the billions being spent to execute it.