Meta tests its first in-house chip designed for AI applications

Photo: Reuters

Meta, the company that owns Facebook, Instagram, and WhatsApp, has started testing its very first in-house chip designed specifically for training artificial intelligence (AI). This chip is created to help Meta reduce its dependence on external suppliers like Nvidia. According to Reuters, the company is initially testing a small number of these chips, with plans to increase production if the testing goes well. The goal is for this chip to help lower Meta's infrastructure costs, which are rising due to the company's big investments in AI to boost its growth.

Meta has predicted that its total expenses for 2025 will range from $114 billion to $119 billion. A large portion of this will be spent on AI infrastructure, with up to $65 billion allocated for capital expenses.

Stay informed with the latest news. Follow DXB News Network on WhatsApp Channel  

The new chip that Meta is testing is a dedicated AI accelerator, which means it's designed to perform only AI-related tasks. This makes it more power-efficient compared to the general-purpose graphics processing units (GPUs) that are commonly used for AI work. Meta has partnered with TSMC, a chip manufacturer based in Taiwan, to produce this chip. TSMC recently announced a $100 billion investment in the U.S. to support such technologies.

This new chip is part of Meta’s broader project called the Meta Training and Inference Accelerator (MTIA) series. Although the MTIA program faced some challenges at first, it gained momentum last year when Meta used one of the MTIA chips for inference, which is the process of running AI systems while users interact with them. For example, it helps decide what content appears on Facebook and Instagram feeds. Meta’s executives have said they hope to use their own chips for training AI by 2026.

Chris Cox, Meta's Chief Product Officer, shared at a conference that the company is focused on improving how they train their systems, especially for recommendations, and eventually aims to use these chips for more advanced AI, including generative AI.

Even though Meta is moving towards creating its own AI chips, many tech companies, including Nvidia, still rely heavily on external AI chips. Nvidia recently became one of the most valuable companies in the world, but many companies are now exploring alternatives.