Meta To Start Production Of New AI Chips In September
Meta Platforms is set to begin production of a new in-house artificial intelligence chip in September, a move that would expand the company’s ability to run AI systems on its own hardware.
The chip, referred to as “Iris” in published reports, is part of Meta’s broader effort to increase its internal computing capacity. Reuters, citing a memo, reported the company is looking to double that capacity, with a longer-term target tied to reaching 14 gigawatts of computing power by 2027.
The planned production start in September is a key milestone for Meta as it works to scale the infrastructure behind its AI products and services. Bringing a proprietary chip into production would give the company an additional option alongside the third-party processors widely used across the tech industry for AI workloads.
Meta has made AI a central focus across its product lineup, from tools that generate and edit content to systems that rank and recommend posts, video, and ads. Those features depend on large-scale computing resources, and the ability to secure sufficient capacity has become a major operational issue for companies competing to build and deploy more advanced AI models.
The reported 14-gigawatt target underscores the size of the buildout Meta is planning. In practical terms, that level of computing power would represent a substantial expansion of the electricity and data-center capacity needed to train and run modern AI systems, placing infrastructure planning on the same footing as product development.
For Meta, moving to produce an in-house chip also highlights a broader industry shift toward custom silicon. Companies pursuing large-scale AI are increasingly looking for ways to optimize performance and manage costs through specialized hardware designed for their own workloads, rather than relying exclusively on general-purpose solutions.
What happens next is the transition from plans to execution. Production beginning in September will be closely watched as an indicator of whether Meta can bring Iris into the manufacturing pipeline on schedule and integrate it into its computing fleet as it pushes to expand capacity.
Additional details about the chip’s rollout and how quickly it will be deployed across Meta’s infrastructure were not included in the context provided, and the company has not publicly outlined a timeline beyond the reported production start.
Still, the start of manufacturing would mark a concrete step in Meta’s push to build more of the underlying AI stack itself, as the competition to secure computing power becomes as consequential as the race to develop new models.
