Meta to Begin Production of Its Iris AI Chip in September, Targeting 14 Gigawatts of Computing Capacity
Meta plans to start manufacturing its in-house Iris AI chip in September, aiming to double computing capacity to 14 gigawatts in 2027, an internal memo reviewed by Reuters shows.
Overview
Meta Platforms plans to begin manufacturing an in-house artificial intelligence chip code-named “Iris” starting in September, part of a plan to boost its total computing power to 14 gigawatts in 2027, according to an internal memo reviewed by Reuters. TechCrunch, citing the same memo, reported that Meta is on track to begin manufacturing its latest AI chips in September and that at least one chip has already completed testing.
What We Know
Iris is manufactured by Taiwan Semiconductor Manufacturing Company (TSMC) and designed in partnership with Broadcom, according to both Reuters and TechCrunch. The chip is part of a four-generation project under Meta’s Meta Training and Inference Accelerators (MTIA) program, Reuters reported.
Testing the chip took only six weeks and found no major issues, the memo showed, according to Reuters. Meta this year plans to deploy seven gigawatts of computing infrastructure, and the company plans to double that capacity again next year to reach a total of 14 gigawatts in 2027, the memo said, per Reuters. TechCrunch similarly reported a plan to deploy seven gigawatts in 2026 with capacity doubling the following year.
Meta has lined up supply agreements with Samsung Electronics for memory chips, SanDisk for flash storage, and Sumitomo Electric for fiber-optic equipment, according to Reuters.
The plan is to use the custom-built silicon to improve the AI that powers Meta’s Facebook and Instagram platforms, and the chip is meant to augment the large quantities of GPUs Meta purchases from Nvidia and AMD for its AI applications, Reuters reported. TechCrunch reported the chips are intended for training models for ranking and recommendation algorithms as well as broader AI workloads and inference.
Meta’s 2026 capital expenditure guidance stands at $125 billion to $145 billion, with the bulk of that spending directed toward data centers, GPUs, and custom silicon, according to TechCrunch.
Meta declined to comment on the memo, Reuters reported.
Mike Gualtieri, a vice president and principal analyst at research firm Forrester, told Reuters: “You can’t become an AI titan if you are dependent on another company for chips.” He added, “The hyperscalers and even SpaceX all plan chips because it will be the only way to compete on price for model usage.”
What We Don’t Know
Meta has not publicly confirmed the Iris name, the September production timeline, or the 14-gigawatt target; the reporting rests on an internal memo reviewed by Reuters that has not been made public, and the company declined to comment. Neither Reuters nor TechCrunch disclosed the financial terms of Meta’s manufacturing arrangement with TSMC or its design partnership with Broadcom, nor did either report specify what share of Meta’s total AI compute Iris and its successors are expected to handle relative to the Nvidia and AMD GPUs the company continues to buy.
Background
In April, Meta disclosed a four-generation MTIA roadmap built with Broadcom, committing to an initial deployment of more than one gigawatt of MTIA-based computing capacity, with plans to scale to multiple gigawatts in 2027 and beyond, as previously reported by The Machine Herald. The September production start for Iris and the updated 14-gigawatt target for 2027 mark the program’s move from roadmap to manufacturing, and a further acceleration of Meta’s compute ambitions since that April disclosure.