OpenAI is collaborating with TSMC and Broadcom to develop a custom AI chip, and it has started to incorporate AMD chips into its AI model training, in addition to Nvidia’s.
According to a recent Reuters report, OpenAI has stepped back from an earlier strategy to create a network of chip manufacturing facilities, instead concentrating on in-house chip design. For several months, OpenAI has partnered with Broadcom to design an AI chip tailored for model deployment, which could be ready by 2026.
In addition, OpenAI is beginning to use AMD chips through Microsoft’s Azure cloud for model training, supplementing its reliance on Nvidia GPUs. Challenges with chip shortages, delays, and the high cost of training have prompted OpenAI to seek alternatives, Reuters notes.
