Meta to Begin Production of In-House AI Chip in September, Memo Shows

Meta Platforms plans to begin manufacturing a custom-designed artificial intelligence chip in September, a key step in its strategy to significantly expand its computing infrastructure and lessen its dependence on third-party suppliers. The plan was detailed in an internal company memo reviewed by Reuters.

In-House Silicon Strategy

The chip, codenamed "Iris," is part of a four-generation project for Meta's own Training and Inference Accelerators (MTIA). The primary goal is to use custom silicon to complement the vast number of high-end graphics processing units (GPUs) the company purchases from market leaders like Nvidia and Advanced Micro Devices.

According to the memo, a six-week testing phase for the Iris chip concluded without any major issues, signaling positive momentum for the long-running internal project. Meta is reportedly collaborating with Broadcom on the chip's design and with Taiwan Semiconductor Manufacturing Co (TSMC) for its manufacturing.

Aggressive Infrastructure Expansion

The move into production supports an ambitious plan to boost Meta's total computing infrastructure to 7 gigawatts this year. The company aims to double that capacity again, targeting a total of 14 gigawatts in 2027, the memo stated.

Ad

To facilitate this rapid expansion amid high demand for components, Meta has secured long-term, multi-year supply agreements for other critical hardware. The memo cited deals with Samsung Electronics for memory chips, Sandisk for flash storage, and Sumitomo Electric for fiber-optic equipment.

Market Context and Implications

Developing custom chips is an increasingly common strategy for hyperscale tech companies seeking to optimize performance for specific AI workloads, control escalating costs, and gain more independence from a concentrated chip market. "You can’t become an AI titan if you are dependent on another company for chips," Mike Gualtieri, a vice president at research firm Forrester, told Reuters.

Meta's push comes as intense demand for AI hardware has created component shortages and rising prices, a trend some Morgan Stanley analysts have termed "chipflation." The memo also revealed Meta's plan for an accelerated release schedule, aiming to launch a new in-house AI processor roughly every six months through 2027.