OpenAI to finalise first custom AI chip by 2026

Reuters

OpenAI plans to finalise its first custom AI chip, aiming for production by 2026, reducing reliance on Nvidia.

OpenAI is set to finalise the design of its first custom artificial intelligence (AI) chip within the next few months. The company plans to send the design for fabrication to Taiwan Semiconductor Manufacturing Co (TSMC), aiming for mass production by 2026.

This move is intended to reduce OpenAI's dependence on external suppliers, such as Nvidia, and strengthen its position in negotiations with other chip manufacturers.

The in-house chip development team, led by Richard Ho, has grown to 40 members and is working in collaboration with Broadcom. The upcoming chip is expected to feature advanced 3-nanometre technology, a systolic array architecture, high-bandwidth memory, and extensive networking capabilities, similar to Nvidia’s existing designs. Initially, it will be deployed on a limited scale for running AI models, with more advanced versions anticipated in the future.

This decision aligns with a broader trend among major technology firms, including Microsoft and Meta, who are investing in AI infrastructure and seeking alternative chip solutions due to increasing costs and reliance on suppliers such as Nvidia.

Tags

Comments (0)

What is your opinion on this topic?

Leave the first comment