OpenAI, the company that created ChatGPT and started the current AI revolution, is mulling a plan to make its own chips, people familiar told Reuters.
The company has been thinking about it since at least 2022 and is now evaluating a chip maker that OpenAI might buy, the sources said.
Frustrated by the ongoing shortage of chip supplies dating from the COVID War, ChatGPT also has considered creating its own chip, deepening relationships with existing chip maker Nvidia, or contracting with additional companies to make chips for it.
CEO Sam Altan has complained about the scarcity of the graphics processing chips needed to run ChatGPT, a market in which Nvidia controls 80 percent.
He also is unhappy with the staggering costs of running conventional computers needed to train new versions of AI. Currently, OpenAI uses a Microsoft-made supercomputer that runs on 10,000 Nvidia graphics chips.
Either designing and producing a custom chip or buying an existing company would likely take at least two years, leaving the company dependent on existing suppliers such as Nvidia and Advanced Micro Devices.
OpenAI declined Reuters’ request for comment.
TRENDPOST: Meta Platforms has tried building its own chips but ran into several problems that set the effort back considerably. Microsoft, OpenAI’s chief funder, also is developing a custom AI chip that OpenAI is trying out.
Most likely, OpenAI will form a new relationship, including a possible takeover, with an outside supplier instead of diverting its own resources and focus away from the increasingly competitive market for new and better AIs.