Meta begins testing its first self-developed AI training chip to reduce dependence on Nvidia

According to Reuters, according to people familiar with the matter, Facebook parent company Meta is testing its first independently developed chip for training artificial intelligence systems.This is an important milestone in the company's move towards designing more customized chips in-house and reducing its reliance on external suppliers such as Nvidia..

Meta has begun deploying the chip on a small scale and plans to expand production for large-scale application if testing goes well, two people familiar with the matter said. Meta’s new training chip is a dedicated accelerator, which means it is only used to handle AI-related tasks. Compared to integrated graphics processing units (GPUs) typically used for AI workloads,This chip is more energy efficient.

Meta is working with TSMC to produce the chip.People familiar with the matter said that Meta began test deployment after completing the first "tapeout" of the chip. Tape-out is an important milestone in chip development and refers to sending the preliminary design to the fab for production. Typically, the cost of a tape-out isup to tens of millions of dollars,And it takes about three to six months to complete,And the test results are not guaranteed to be successful. If the test fails, Meta needs to identify the problem and tapeout again.

As of press time, Meta and TSMC declined to comment.