Meta Platforms is set to begin production of its custom AI chip, Iris, this September as part of a strategic plan to double its computing power to 14 gigawatts by 2027, signaling increased independence in its AI hardware capabilities.
- Iris chip production starts September, with rapid bug testing completed
- Computing capacity to expand from 7 GW in 2026 to 14 GW by 2027
- Long-term supply deals secure memory and fiber-optic components amid chip shortages
What happened
Meta Platforms plans to start manufacturing its proprietary AI chip, codenamed Iris, in September. This chip is part of a four-generation series of Meta Training and Inference Accelerators (MTIA) developed internally to enhance AI performance for its family of apps, including Facebook and Instagram. The Iris chip successfully passed testing in six weeks without major issues, marking progress for a project that has faced delays over more than five years.
The chip is custom-designed specifically for Meta’s needs, created in collaboration with Broadcom for design assistance and manufactured by Taiwan Semiconductor Manufacturing Company (TSMC). Iris is intended to work alongside commonly used GPUs from Nvidia and AMD to accelerate AI workloads. Meta is also investing heavily in expanding its data center computing resources, targeting a deployment of seven gigawatts in 2026 and doubling that to 14 gigawatts in 2027.
Why it matters
By creating and producing its own AI chips, Meta aims to reduce dependence on external suppliers like Nvidia, which has been a bottleneck affecting the timely adoption of new GPU hardware. This strategic move can help Meta control costs, improve supply chain stability, and gain a competitive edge in a market where AI demands are rapidly increasing.
Meta’s substantial planned outlay on AI infrastructure, estimated at up to $145 billion for the year, reflects Big Tech’s broader push to build powerful computing ecosystems. Long-term agreements with vendors like Samsung, Sandisk, and Sumitomo ensure critical components such as memory, flash storage, and fiber optics remain available despite ongoing industry-wide shortages, helping Meta keep pace with AI’s resource-hungry expansion.
What to watch next
Observers will focus on the commercial performance and scalability of the Iris chip as Meta rolls it out, especially its impact on reducing latency and costs for AI workloads on Facebook, Instagram, and other services. The company’s plan to release a new AI processor approximately every six months through 2027 is unusually aggressive compared to typical yearly cycles and may influence competitors’ chip development timelines.
The broader AI infrastructure race will also highlight supply chain dynamics as demands for memory and semiconductor components surge. Meta’s success in securing crucial long-term supply contracts and managing rising prices—amid a global 'chipflation' trend—will be key indicators of how well the company can sustain rapid growth and maintain technological independence in AI computing.