OpenAI finalising first custom chip design, says Reuters
Make us a preferred source
Make us a preferred source on Google
The news: OpenAI is moving forward with its plan to develop its first in-house AI chip and reduce its reliance on Nvidia, with the design set to be finalised in the coming months, Reuters reported, citing unnamed sources.
The design will be sent for fabrication at Taiwan Semiconductor Manufacturing Co (TSMC) using 3-nanometer technology, with mass production targeted for 2026.
The numbers: The process, known as “taping out,” costs tens of millions of dollars and typically takes about six months unless OpenAI pays for expedited manufacturing.
The chip may not work on the first attempt, requiring troubleshooting and another tape-out, the news agency noted.
The chip is being developed by OpenAI’s in-house team, led by Richard Ho, formerly of Google, in collaboration with Broadcom. The team has doubled in size to 40 engineers in recent months.
If the tape-out succeeds, OpenAI could begin testing an alternative to Nvidia’s chips later this year, the report said.
The chip will initially be deployed on a limited scale to run AI models. It features a systolic array architecture, high-bandwidth memory, and extensive networking capabilities, according to the report.
What they said: OpenAI and TSMC declined to comment to Reuters.
The source: Reuters
