Broadcom held discussions with OpenAI about producing AI chip: report

Broadcom (NASDAQ:AVGO) has had discussions with OpenAI, the generative artificial intelligence startup backed by Microsoft (MSFT), about building a new artificial intelligence server chip, The Information reported.

OpenAI has hired engineers from the unit of Google (GOOG) (GOOGL) that created its tensor processing unit in an effort to create its own AI server chip, the news outlet added, citing three people involved in the conversations.

The efforts to create its own AI server chip — which in turn could reduce its reliance on Nvidia (NVDA) — are a part of OpenAI Chief Executive Sam Altman’s vision to boost the global semiconductor infrastructure. Altman has publicly complained about the lack of access to AI infrastructure. “The world needs more AI infrastructure – fab capacity, energy, data centers, etc. – than people are currently planning to build,” Altman previously said. “OpenAI will try to help!”

Broadcom has a unit that works with companies like Google and others to make chips for them, known as application-specific integrated circuits.

If such a chip were built, it would not be produced until 2026 at the earliest, one of the people added, as details are still being worked out. These include chip packaging and memory components, among others. Altman has spoken with South Korea’s Samsung (OTCPK:SSNLF) and SK Hynix about the need for high-bandwidth memory and his chip plans, The Information added, citing two people with knowledge of the conversations.

Altman has also spoken with Taiwan Semiconductor (TSM) to see if the global foundry could increase production of Nvidia’s chips or if they could produce the proposed AI chip from OpenAI, the news outlet added.

OpenAI and Broadcom did not immediately respond to a request for comment from Seeking Alpha.

Broadcom shares ended Thursday with a gain of 2.9%, bucking the broader market decline.