OpenAI has announced a partnership with Broadcom to manufacture custom computer chips intended for its AI data centers. This collaboration is part of a broader initiative to decrease OpenAI’s dependency on Nvidia and to ensure an adequate supply of computing power for applications such as ChatGPT and Sora, as well as to advance its goal of developing superintelligent AI.
OpenAI stated that by designing its own chips, it can incorporate insights gained from developing advanced models and products directly into the hardware, which may enhance its capabilities and intelligence.
The arrangement with Broadcom is projected to see the deployment of equipment racks beginning in the latter half of 2026, with the overall deal expected to conclude by the end of 2029. Sam Altman, co-founder and CEO of OpenAI, indicated that this partnership is vital for constructing the necessary infrastructure to unlock the potential of AI and deliver tangible benefits to users and businesses.
The initiative to create custom chips reflects a growing trend in the technology sector, with major companies like Meta, Google, and Microsoft also pursuing similar strategies to strengthen supply lines amid escalating demand, while reducing reliance on Nvidia’s AI chips. So far, these custom chip endeavors have not posed a significant challenge to Nvidia’s dominance in the market, although they have provided substantial opportunities for companies like Broadcom.
Source: https://www.theverge.com/news/798827/openai-broadcom-custom-ai-chips