Meta AI chips: production starts in September to challenge Nvidia

Meta has decided it's time to cut ties, or at least spending, with GPU giants like Nvidia. Zuckerberg's company wants less external reliance. To achieve this, Meta will begin production of its latest generation of AI chips in September 2026, aiming for greater autonomy.

It's no secret that Nvidia's graphics cards cost an arm and a leg, especially those optimized for artificial intelligence. Meta, with all its AI models and services, consumes them by the truckload. Creating its own chips is a way to save a lot of money and, let's be honest, to have tighter control over how its technological arsenal operates.

The idea of making hardware in-house isn't new for Meta. Meta's first generation of AI chips, the “Meta Training and Inference Accelerator” (MTIA), launched in 2023. These new chips, however, are a step up, designed to be even more efficient and specific to the company's internal needs. Will they measure up to Nvidia's established titans?

The challenge is huge. Nvidia isn't just hardware; it's also a robust software ecosystem that AI developers adore. Meta will need to prove that its chips, besides saving money, can offer comparable performance and flexibility. Or at least, good enough to justify the fuss. Otherwise, what's the point of all this hullabaloo?

For Meta, the change is big: it means less money going into Nvidia's pockets and more resources invested internally. This could lead to an acceleration in the development of its AI products, perhaps making them faster or more capable. It's a bit like growing your own vegetables instead of buying them at the supermarket: more control, fewer costs (if you do it right).

If Meta succeeds, we might see improvements in the services we use daily, from Instagram to Facebook, maybe with more responsive AI or brand-new features. It's not like your smartphone will turn into a supercomputer tomorrow, but behind the scenes, the chip war is getting more interesting. And a bit more homegrown, it seems.