Meta's custom 'Iris' AI chip targets September 2026 launch to cut Nvidia dependence

Meta Platforms plans to begin manufacturing its first in-house AI chip, code-named 'Iris,' in September 2026. The custom-designed data center chip is part of the company's ambitious strategy to achieve 14 gigawatts of total computing power by 2027, while simultaneously reducing its heavy dependency on external suppliers like Nvidia. Iris will be purpose-built to accelerate the AI models that power Meta's core social platforms, including Facebook and Instagram, handling tasks such as content recommendation, moderation, and generative AI features.

This vertical integration move places Meta alongside other tech giants that have developed their own silicon — Google's TPU, Amazon's Trainium and Inferentia, and Microsoft's Azure Maia. By designing and manufacturing Iris in-house, Meta gains more control over performance, power efficiency, and supply chain resilience. The chip's development aligns with Meta's broader push into the AI infrastructure race, where the company has already committed billions to expanding data centers and purchasing Nvidia GPUs. Iris represents a strategic long-term bet, but it will take time before it meaningfully offsets Meta's reliance on external chipmakers.