Meta is set to begin manufacturing its custom Meta Iris AI chip in September. The move marks a major milestone in the company’s artificial intelligence strategy. It also highlights Meta’s push to build more of its AI infrastructure in-house and reduce its dependence on third-party chipmakers.
The chip, code-named Iris, is part of Meta’s Meta Training and Inference Accelerators (MTIA) program. It is designed to handle AI training and inference across Facebook, Instagram, and Meta’s expanding portfolio of AI products. According to an internal memo reviewed by Reuters, the company plans to begin production as it rapidly expands its computing capacity to support larger AI models. (Reuters)
Why the Meta Iris AI chip matters
Unlike the GPUs Meta buys from Nvidia and AMD, the Meta Iris AI chip is built specifically for the company’s AI workloads. That gives Meta more control over performance while helping lower long-term computing costs.
Meta partnered with Broadcom to design the chip. Meanwhile, Taiwan Semiconductor Manufacturing Company (TSMC) will manufacture it. This approach allows Meta to combine its own chip architecture with the expertise of established semiconductor companies.
More importantly, Iris supports Meta’s long-term AI strategy. The company wants to own more of the technology behind its AI services. That strategy goes beyond chips. It also includes AI models, networking equipment, and data centers.
In addition, the project appears to have reached an important milestone. Reuters reported that Iris completed testing in about six weeks without major issues. The successful test suggests Meta’s custom silicon program is becoming more mature after several years of development.
Meta also plans to expand its AI infrastructure at an aggressive pace. The company expects to deploy about 7 gigawatts of computing capacity this year. It then aims to double that figure to 14 gigawatts in 2027. To support the expansion, Meta expects to spend as much as $145 billion on AI infrastructure this year.
However, Meta is not replacing Nvidia and AMD overnight. Instead, the Meta Iris AI chip will work alongside existing GPUs. Over time, Meta expects the chip to improve efficiency, lower infrastructure costs, and reduce reliance on external suppliers.
The announcement also reflects a broader industry trend. Google, Amazon, Microsoft, and Apple are all investing heavily in custom AI chips. As AI becomes more important to their businesses, owning the hardware that powers those systems is becoming a key competitive advantage.
