OpenAI is said to be in talks with Broadcom about developing custom AI inference chips to run its models. On an earnings call last week, Broadcom disclosed that an AI developer had placed a $10 billion order for AI server racks using its chips. That new customer was reported to be OpenAI, which has relied primarily on hotly sought-after Nvidia GPUs for model training and deployment. Broadcom specializes in XPUs — accelerator chips designed for specific uses, like inference for ChatGPT. OpenAI CEO Sam Altman has publicly complained that a shortage of chips has impeded the company’s ability to get new models and products to market.

“To solve this problem, OpenAI has been working with Broadcom for over a year to develop a custom chip for use in model training,” reports The Wall Street Journal.

The Broadcom deal marks the first time OpenAI has moved to customize an AI chip of its own, according to the Financial Times.

WSJ adds that “the Broadcom chip isn’t designed to challenge Nvidia, but rather to plug the gaps in OpenAI’s hardware needs.” Nonetheless, rumors that OpenAI signed on as a client came with a Broadcom stock surge of up to 16 percent on Friday.

The $10 billion order — placed by a fourth AI developer “with immediate and fairly substantial demand really changes our thinking of what 2026 would be starting to look like,” Broadcom CEO Hock Tan said in WSJ as the company announced record fiscal Q3 revenue of $5.2 billion, 63 percent year-over-year growth.

Financial Times reports HSBC analysts expect Broadcom’s custom AI chip business will propel it to greater growth next year than Nvidia, whose AI chip business some feel has hit maturity.

“Nvidia continues to dominate the AI hardware space, with the Big Tech ‘hyperscalers’ still representing a significant share of its customer base,” according to Financial Times. “But its growth has slowed relative to the astronomical figures it saw at the start of the AI boom.”

Global AI momentum “has catapulted Broadcom into the upper echelons of tech companies,” writes WSJ, which notes the Palo Alto-based company has seen its share price increase “sixfold since the start of 2023 as it moved further into data centers,” making it “the world’s seventh-largest publicly listed company.”

In August, the company launched its next generation Jericho4 networking chip, “designed to help speed up AI computing by connecting data centers as far as 60 miles apart,” according to Reuters.

Related:
How Could an OpenAI Partnership with Broadcom Shake up Silicon Valley’s Chip Hierarchy?, Los Angeles Times, 9/25