Meta is building its own brain. The company plans to begin production of a proprietary AI chip codenamed “Iris” in September 2026, a cornerstone of its strategy to bring what CEO Mark Zuckerberg calls “personal super intelligence” to billions of users across its apps and devices.
The chip falls under Meta’s MTIA program, short for Meta Training and Inference Accelerators. It’s the company’s clearest signal yet that it doesn’t want to keep renting compute power from Nvidia forever. Or at least, not exclusively.
What Meta is actually building
Iris is part of a four-generation chip roadmap spanning the MTIA 300, 400, 450, and 500 series. Each generation is designed to handle increasingly complex AI workloads, from the ranking and recommendation systems that decide what shows up in your Instagram feed to full-scale AI inference and training.
The compute targets are staggering. Meta is aiming for over 14 gigawatts of computing power by 2027, roughly double the 7 gigawatts it projects for 2026. For context, a single gigawatt can power approximately 750,000 homes. Meta wants 14 of those just to run AI models.
