OpenAI is reportedly working with Broadcom to develop new custom chips designed to handle its large inference AI workloads and secure production capacity with TSMC, according to sources the company has spoken with Reuters. OpenAI has reportedly built a chip development team of about 20 people, including senior engineers who previously worked on Google's Tensor processors for AI.
However, based on the current schedule, production of the bespoke hardware may not begin until 2026.
The information had reported in July that OpenAI was discussing developing its own AI chip with Broadcom and other semiconductor designers, and earlier this year Bloomberg reported that OpenAI is working on building its own network of foundries, but according to Reutersthese plans were put on hold for cost and time reasons.
The reported strategy puts OpenAI on a similar path to the other tech companies trying to manage costs and access to AI server hardware with tailored chip designs. But Google, Microsoft and Amazon are several generations further along in their efforts, and OpenAI may need significantly more funding to become a true competitor.