Meta's Iris AI chip enters production in September 2026

Meta will begin production of its in-house AI chip, code-named Iris, in September 2026, according to an internal memo reviewed by Reuters. The chip represents the first in a planned series of four custom processors designed to power AI workloads across Facebook and Instagram.

Iris cleared its bug-testing phase in approximately six weeks without encountering major issues, the memo indicated. Broadcom is serving as Meta's design partner, while Taiwan Semiconductor Manufacturing Co. will handle fabrication. The chip falls under Meta's MTIA program—Meta Training and Inference Accelerators—which aims to reduce compute spending and decrease dependence on external GPU suppliers including Nvidia and Advanced Micro Devices.

Why it matters

Meta's aggressive custom silicon strategy reflects a broader industry shift as hyperscale companies seek greater control over AI infrastructure costs and supply chains. With AI infrastructure spending projected to reach as high as $145 billion this year, building proprietary chips allows Meta to optimize performance for specific workloads while mitigating vendor concentration risk. The six-month chip release cadence through 2027 significantly outpaces the industry's typical annual cycle, signaling the urgency Meta places on this capability.

Aggressive release schedule

Meta publicly introduced Iris in March 2026 as part of a four-chip roadmap extending through 2027. The company plans to release a new processor approximately every six months, a notably faster pace than standard industry timelines. An internal memo noted that adopting the latest GPUs at Meta's scale "has been a heavy lift, and it has cost us time."

The Iris chip is designed to supplement—not replace—the substantial volumes of GPUs Meta continues to purchase from Nvidia and AMD. This hybrid approach allows Meta to balance custom optimization with proven third-party hardware.

Infrastructure expansion

The chip production announcement accompanies a major infrastructure buildout. Meta plans to bring seven gigawatts of computing capacity online in 2026, expanding to 14 gigawatts by 2027. To support this growth, the company has secured extended supply contracts for memory chips from Samsung Electronics, flash storage from SanDisk, and fiber-optic equipment from Sumitomo Electric. None of the three suppliers provided comment to Reuters.

Broader silicon partnerships

Meta formalized its custom silicon strategy with Broadcom earlier in 2026, agreeing to expand their partnership on custom AI chips through 2029 across multiple MTIA generations. That agreement included a commitment to deploy more than one gigawatt of computing capacity as an initial phase of a multi-gigawatt buildout.

Separately, Meta has established a multiyear agreement with AMD to deploy up to six gigawatts of AMD Instinct GPUs, part of an effort to diversify compute supply beyond a single vendor. This multi-supplier approach aims to reduce supply chain risk while maintaining access to cutting-edge GPU technology.

Meta declined to comment on the production timeline and infrastructure plans. The details were first reported by Reuters based on an internal company memo.