OpenAI already wishes for a longer time to develop AI processors in-house to meet the required GPU capacity for its various GenAI solutions. Broadcom is one of the manufacturers with whom negotiations are now said to be serious.
According to the AI developer, the continued growing demand for OpenAI’s GenAI solutions and models makes it necessary to ramp up processor capacity. This is difficult because of global shortages of these types of processors, so OpenAI now wants to develop them itself. The AI developer is in talks about this with several chip designers, including Broadcom, reports The Information, based on sources. The negotiations would involve developing an AI server processor or proprietary GPU.
Hiring Google experts
Detailed information about these talks is not available, but an OpenAI spokesperson indicated that it is in talks with the chip industry and other stakeholders to get more access to the necessary infrastructure. “This is to ensure that all AI benefits are and remain widely accessible.”
OpenAI is already taking the lead in designing its own AI processors. To this end, it employs former Google employees who worked on its own Tensor AI processor, among others.
Earlier plans
The rumors about OpenAI wanting to develop its own AI processors are not new. Back in October last year, there were reports that CEO Sam Altman wanted to set aside billions of dollars for the production of these processors. This is in cooperation with chip manufacturers such as Intel, TSMC and Samsung Electronics.
Also read: OpenAI wants to develop its own AI processors to combat GPU shortage