OpenAI to Mass - Produce Custom AI Chips with Broadcom

Multiple sources familiar with the matter have revealed that the jointly - designed chips will be delivered next year. Broadcom's CEO, Hock Tan, disclosed last Thursday that a new, unnamed client had placed an order worth a staggering $10 billion. Although Broadcom did not officially identify the client, those with knowledge of the situation have confirmed it to be OpenAI. Following this announcement, Broadcom's stock price soared nearly 11% last Friday.

The need for custom chips has become increasingly apparent as AI companies grapple with the soaring demand for training complex models. OpenAI, which heavily depends on NVIDIA's off - the - shelf GPUs for its flagship chatbot, ChatGPT, has faced setbacks due to GPU shortages. These shortages have reportedly slowed down the development of new versions of ChatGPT. Broadcom, with its expertise in "XPU" semiconductors designed for specific applications, such as training ChatGPT, is well - positioned to assist OpenAI in fulfilling its hardware requirements.

This development follows in the footsteps of tech giants like Google, Amazon, and Meta, all of which have developed their own proprietary chips for running AI. As the demand for the computational power required to train and operate AI models continues to skyrocket, custom chips offer the potential for enhanced performance and the ability to overcome bottlenecks. Analysts at HSBC predict that Broadcom's custom chip business will experience much faster growth than NVIDIA's chip business in 2026. However, sources quoted by the Wall Street Journal suggest that Broadcom's chips are not intended to compete directly with NVIDIA but rather to fill the gap in OpenAI's hardware needs.