Meta is preparing one of its largest artificial intelligence infrastructure upgrades yet, and a central piece of the plan is building more of its own hardware.
According to an internal company memo reviewed by Reuters, Meta intends to begin manufacturing its own AI chips and to move to a rapid cadence of releases, with reports describing plans for a new AI chip roughly every six months. National Technology News reports that Meta is preparing in-house AI chip production specifically to expand its computing capacity.
The strategic goal behind the effort, according to the reporting, is to reduce Meta's dependence on Nvidia, the company that currently dominates the market for the high-end processors used to train and run large AI models. By designing and producing chips internally, Meta would gain more control over the supply, cost, and design of the hardware that powers its AI ambitions.
A twice-a-year release schedule would be an aggressive pace for custom silicon, which is typically slow and expensive to develop. If Meta can sustain it, the approach would let the company tune each generation of chips more closely to the demands of its own AI systems rather than relying entirely on general-purpose hardware bought from an outside supplier.
The details reported so far come from an internal memo rather than a public product announcement, so specifics such as performance, timing, and manufacturing partners remain limited to what the sources describe.
Why it matters: If a company the size of Meta can reliably build its own AI chips on a fast schedule, it signals that the biggest technology firms are moving to loosen Nvidia's grip on the AI hardware that increasingly underpins the industry.