Meta is testing its in-house AI chip

Meta is testing its first in-house chip for training AI systems, per Reuters in collaboration with TSMC.

The deets: Meta is now making its own AI chips instead of depending on companies like Nvidia.

The project has been executed via small deployment of a chip. The company plans to ramp up production for wide-scale use if the test goes well. 

The new AI chip is built only for AI tasks, making it more efficient than regular computer graphics chips usually used for such work.

Context: the development is Meta’s long-term plan. The move is aimed at cutting infrastructure costs as it bets big on AI for growth.

Meta has started testing its new AI chip after completing an important step called "tape-out." This means they finalized the chip’s design and sent it to a factory for production. It’s a key milestone in making new computer chips.

The company already has an Meta Training and Inference Accelerator (MTIA) chip in use. This provides AI-driven recommendation systems for Facebook and Instagram feeds. 

The latest chip in the Meta Training and Inference Accelerator (MTIA) series is undergoing testing, aiming to enhance AI processing efficiency.

The social media giant previously scrapped its own AI chip after it failed in early testing. Instead, the company switched gears and bought billions of dollars worth of Nvidia chips in 2022. Now the company is hoping for better results.

Big picture: Meta’s update comes a few days after Foxconn launched its first large language model, called ‘FoxBrain’. 

Every major tech giant including Google, OpenAI, Apple  Amazon, is aggressively building AI models, investing billions to stay ahead in the race.