OpenAI has announced a partnership with Broadcom to design and produce custom computer chips aimed at powering its AI data centers. This initiative aims to reduce OpenAI’s dependence on Nvidia chips and secure sufficient computing power to support AI applications like ChatGPT and Sora. By developing its own chips, OpenAI plans to integrate its deep insights from AI model development directly into hardware, enhancing performance and intelligence. The collaboration involves creating and deploying 10 gigawatts of specialized AI accelerators, a scale comparable to the output of ten nuclear reactors. Broadcom is set to begin installing equipment in mid-2026, with the project expected to conclude by the end of 2029. This move follows recent deals between OpenAI and major chipmakers such as AMD and Nvidia and aligns with an industry trend where tech giants including Meta, Google, and Microsoft develop custom AI chip solutions to reduce reliance on Nvidia and strengthen supply chains. While custom chips have yet to challenge Nvidia’s dominance, they have notably benefited companies like Broadcom.
Back
