OpenAI has partnered with Broadcom to design its first in-house AI chips, aiming to minimise reliance on Nvidia and meet increasing demand for computing power.
Under the deal, the ChatGPT manufacturer will design the chips, while Broadcom will develop them starting in late 2026.
Both companies plan to launch custom chips with 10 gigawatts of capacity, roughly the power used by eight million US homes.
According to a few credible analysts, this would not pose a significant threat to Nvidia's dominance in AI accelerators due to the complexity of chip development and manufacturing.
The Broadcom collaboration comes after OpenAI’s recent 6-gigawatt AI chip supply deal with AMD and Nvidia’s plan to invest up to $100 billion in OpenAI infrastructure.
However, financial details regarding the Broadcom deal remain under wraps, but analysts expect funding from a mix of Microsoft support, investor rounds, and future revenue.
custom chip efforts are rapidly growing across tech, with companies such as Google, Amazon, and Meta, developing in-house hardware to meet AI demand.
Broadcom’s stock price has exponentially increased nearly sixfold since 2022, partly driven by AI chip demand.
OpenAI and Broadcom aim to complete deployment by 2029, using Broadcom’s Ethernet and networking systems—challenging Nvidia’s InfiniBand dominance in AI data center infrastructure.
With this partnership, OpenAI and Broadcom aim to complete deployment by 2029, utilising Broadcom’s Ethernet and networking systems in a bid to outdo Nvidia’s InfiniBand dominance in AI data center infrastructure.
