Key Points:
- Meta introduced four new in-house AI chips to its hardware family.
- The company aims to reduce its heavy reliance on outside vendors like Nvidia.
- Engineers designed the chips to fit existing server racks for easy future upgrades.
- Tech giants plan to spend a combined 650 billion dollars on AI this year.
Meta just revealed four brand new artificial intelligence chips. The social media giant added the MTIA 300, 400, 450, and 500 models to its internal processor family. This move helps the company meet massive computing demands while reducing its dependence on major suppliers like Nvidia and AMD.
These new chips tackle different parts of Meta’s daily business. Some models focus on ranking and recommending the content users see on Facebook and Instagram. The higher-end models power complex generative artificial intelligence tasks.
The MTIA 400 stands out as a powerful tool for generating content. Meta designed it so engineers can link 72 of these chips together inside a single server rack. Company leaders claim this specific model delivers raw performance that easily competes with the best commercial products available, while also saving them significant money.
Moving up the ladder, the MTIA 450 offers much faster high-bandwidth memory. The flagship MTIA 500 pushes the limits even further with maximum memory capacity and top-tier speeds. Meta already uses some of these chips today and plans to roll out the rest across 2026 and 2027. They all share the exact same physical infrastructure, allowing technicians to swap them out quickly whenever upgrades become necessary.
Meta is hardly alone in this hardware race. Google and Amazon have built their own processors for years, and Microsoft recently launched its Maia 200 chip. Interestingly, reports show Meta also recently signed a massive agreement to use Google processors alongside its own hardware.
This growing trend of building custom silicon creates potential headwinds for traditional chipmakers. Nvidia recently noted that major tech companies account for more than half of its data center revenue. However, these corporations show absolutely no signs of closing their wallets. Experts expect Amazon, Google, Meta, and Microsoft to spend a staggering 650 billion dollars collectively on artificial intelligence infrastructure this year alone.
