According to a recent report from the Financial Times, artificial intelligence giant OpenAI has reached a large-scale partnership with semiconductor giant Broadcom to bring its proprietary AI chip to mass production. This move is seen as a key strategy for OpenAI to reduce its reliance on Nvidia and expand its hardware business.
Last October, Reuters revealed that OpenAI was collaborating with Broadcom and TSMC to develop its first proprietary AI chip, and planned to integrate AMD chips alongside Nvidia GPUs to meet surging computing power demands. Further news in February of this year revealed that OpenAI was using its proprietary chips as a bargaining chip in negotiations with suppliers. Its first product will focus on AI model training, with plans to develop more powerful processors in the future. Nvidia currently holds approximately 80% of the global AI chip market, and its GPUs are widely used by companies like OpenAI, Google, and Meta.
However, the road to developing its own chips is fraught with challenges. Despite years of research and development, tech giants like Microsoft and Meta have failed to launch competitive products. OpenAI's chip team, led by former Google executive Richard Ho, has grown to 40 people, but remains significantly smaller than those at Google or Amazon. Industry insiders estimate that a single chip design could cost as much as $500 million (approximately 3.57 billion RMB). This cost could double if supporting software and hardware are included.
If this collaboration is successful, it will not only enhance OpenAI's supply chain autonomy but also potentially open up new revenue streams, further reshaping the AI hardware industry landscape.
