Broadcom discussed with OpenAI the possibility of developing an accelerator for AI systems

The ambitions of OpenAI management in the development and production of its own accelerators for artificial intelligence systems are no secret, and the day before, The Information reported that the company was in negotiations to develop a corresponding chip with Broadcom. Against this background, the shares of the latter managed to increase in price by 3%.

Image Source: Shutterstock

As the source notes, Broadcom was just one of the developers with whom OpenAI was negotiating. Among those involved in the process were people from Google with experience in developing processors of the Tensor family. OpenAI managed to hire some of them to implement their own plans. The specialists were tasked with developing server-class accelerators for artificial intelligence systems.

OpenAI CEO Sam Altman has repeatedly complained about the lack of hardware infrastructure to support his plans for the development of artificial intelligence systems. His company could assist partners in building not only data centers, but also component manufacturing plants and power plants. Broadcom has a division that develops chips taking into account the wishes of specific customers.

If the OpenAI project had been implemented, the Broadcom processor would have started to be produced no earlier than 2026 at best. Along the way, OpenAI management had to negotiate with Samsung and SK hynix regarding equipping the corresponding accelerators with HBM type memory. OpenAI planned to produce the chip at the facilities of TSMC, and Altman was negotiating with the latter about increasing the production volume of Nvidia chips, which OpenAI also actively uses.