Meta is moving to make more of its own artificial-intelligence hardware. According to reporting from Yahoo Finance and The Motley Fool, the company run by Mark Zuckerberg is putting its in-house AI chip, called Iris, into production in September.
The chip is part of a larger goal: doubling Meta's computing capacity to 14 gigawatts. In simple terms, gigawatts here are a stand-in for how much raw computing power Meta can bring to bear on training and running AI systems. Doubling that capacity signals just how much the company expects its AI ambitions to grow.
The move also reflects how quickly the ground has shifted. According to The Motley Fool, Meta CFO Susan Li acknowledged earlier this year that data center capacity planned 12 to 36 months ago is no longer sufficient. The problem is timing: new data center construction requires a multiyear lead time, so decisions made a few years back can't keep pace with today's surging demand for AI computing.
Building chips in-house is significant because most AI companies rely heavily on outside suppliers, most notably Nvidia, for the specialized processors that power AI. By developing Iris internally, Meta could reduce its dependence on those suppliers, gain more control over its hardware roadmap, and potentially lower costs as it scales up.
The sources do not detail Iris's technical specifications, its manufacturing partner, or how it compares to chips from rivals, so those questions remain open.
Why it matters: Meta's push to make its own chips and sharply expand computing capacity shows that the biggest tech companies increasingly see control over AI hardware, not just software, as central to competing in the AI race.