According to Reuters, sources related to the news reveal that OpenAI is collaborating with Broadcom and TSMC to develop their own AI processing chips. Additionally, the company has begun purchasing AI chips from AMD in addition to NVIDIA to meet increasing demand.
Previously, reports indicated that OpenAI was exploring ways to reduce reliance on external AI processing chips to have chips that work on their own, similar to what major tech companies do. Initially, OpenAI aimed to have complete control from start to finish, including manufacturing their own chips. However, the company abandoned this plan due to high costs and lengthy timelines.
Sources say that OpenAI has been partnering with Broadcom for several months in designing AI processing chips focusing on Inference tasks, expecting higher future demand as more AI applications emerge. This team at OpenAI consists of approximately 20 individuals, some of whom have previously worked on designing TPUs at Google.
Regarding chip production, OpenAI has placed orders with TSMC through negotiations with Broadcom, with the first chip expected to enter production in 2026.
Furthermore, OpenAI plans to increase the usage of AMD’s AI processing chips through deployment on Microsoft’s Azure to accommodate growing demands, as the quantity of NVIDIA chips is insufficient.
OpenAI, Broadcom, and TSMC declined to comment on the aforementioned news source: Reuters.
TLDR: OpenAI collaborates with Broadcom and TSMC to develop AI processing chips, purchases AMD chips, plans for increased use of AMD chips via Microsoft’s Azure, and declines to comment on the news.
Leave a Comment