Meta is moving its custom AI silicon from the drawing board to the assembly line. According to MLQ.ai, the company plans to begin manufacturing its in-house 'Iris' AI chip in September. A report from eciks.org adds that the effort is tied to Meta's goal of doubling its computing capacity.
The push is part of a broader strategy to control its own hardware destiny. TechRepublic reports that Iris would be Meta's fourth custom AI chip, and that the company is looking to cut AI costs while reducing its reliance on outside suppliers Nvidia and AMD. Cheddar frames the move bluntly, describing it as Meta taking aim at Nvidia with a major AI chip push.
Investors appear to like the direction. MarketWatch reports that Meta's stock rebounded as a combination of agentic AI coding tools and custom chips helped ease fears about the company's heavy spending on artificial intelligence.
Meta's appetite for squeezing more value from its hardware isn't limited to processors. According to The Register, the company has also designed a custom CXL bridge chip that lets it reuse memory from older servers in new ones — a cost-saving engineering trick that drew heavy discussion on Hacker News, where the story gathered hundreds of points and comments.
Taken together, these moves sketch a company trying to build more of its AI stack itself rather than buying it off the shelf. Designing chips in-house is expensive and technically demanding, but it can lower long-term costs and loosen the grip that dominant suppliers like Nvidia hold over the industry.
Why it matters: If Meta can manufacture competitive AI chips at scale, it signals that Big Tech's biggest buyers are increasingly willing to become builders — a shift that could reshape who profits from the AI boom.