ChipPub

ChipPub

What Kind of Chip Will OpenAI Develop?

OpenAI partners with Broadcom to develop custom AI chips, reducing reliance on Nvidia and cutting costs for GPT models. Set to debut in 2025.

Meng Li's avatar
Meng Li
Sep 08, 2025
∙ Paid
1
1
Share

"ChipPub" Publication: 20% Discount Offer Link.


OpenAI partners with Broadcom to launch first AI chip

According to reports, OpenAI is developing custom AI accelerators with the help of Broadcom, apparently to reduce its dependence on Nvidia and lower the costs of its GPT series models.

The UK's Financial Times, citing sources familiar with the matter, reported that Broadcom CEO Hock Tan revealed during Thursday's earnings call that Broadcom's mysterious $10 billion customer is none other than Sam Altman's AI hype factory, OpenAI.

Although Broadcom does not disclose its clients by habit, the company's intellectual property forms the foundation for most custom cloud silicon, which is an open secret.

Tan told analysts on Thursday's call that Broadcom is currently serving three XPU customers, with a fourth one on the way.

He said: "Last quarter, one of those potential customers issued a production order to Broadcom, so we are qualifying them as an XPU customer, and in fact, they have secured over $10 billion in AI rack orders based on our XPUs. As a result, we now expect our AI revenue outlook for fiscal 2026 to improve significantly from our previous quarter's expectations."

For some time, there have been rumors that OpenAI is internally developing a chip as an alternative to Nvidia and AMD GPUs. Sources told the Financial Times that the chip is expected to debut sometime next year, but primarily for internal use and not open to external customers.

This post is for paid subscribers

Already a paid subscriber? Sign in
© 2025 Meng Li
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture