OpenAI Teams Up with Broadcom to Design Custom AI Chips: Full Details Inside

OpenAI Announces Partnership with Broadcom to Develop Its Own AI Chips by 2026, Aiming to Reduce Reliance on Nvidia

OpenAI Elevates Its Game with Broadcom

As previously reported, OpenAI has formalized its collaboration with Broadcom to create its dedicated AI chips. The goal is to roll out 10 gigawatts of capacity by 2029, which is equivalent to the power usage of millions of households.

This partnership emerges at a time when OpenAI is looking to secure its computational resources amidst rising demands for its models. This alliance could also allow Sam Altman’s company to gain more control over its infrastructure and potentially reduce costs.

Less Dependent on Nvidia

Developing its own chips is part of an ongoing effort by OpenAI to minimize dependence on traditional suppliers such as Nvidia, which currently holds a dominant position in the market with its high-end GPUs.

Despite recent extensive deals with Nvidia and AMD (10 GW and 6 GW respectively), the approach with Broadcom is distinct: OpenAI is directly designing its own accelerators. This move could also give the company more leverage in future negotiations by diversifying its technology partner portfolio.

Gradual Deployment Starting in 2026

The initial racks equipped with these custom chips are expected to be deployed in the second half of 2026, with full deployment planned by the end of 2029. Broadcom will provide not only the components but also the network infrastructure (including Ethernet), as an alternative to Nvidia’s InfiniBand technology.

The total cost of the project hasn’t been disclosed, but it is estimated to be in the tens, possibly hundreds of billions of dollars over several years. The financing model remains unclear, although it is known that OpenAI relies on funding rounds, pre-orders, and strategic support, particularly from Microsoft.

A Tense Sector

Similar to Google, Amazon, or Meta, OpenAI is looking to internalize part of its hardware to better control its value chain. However, like other companies, the question of short-term profitability remains unresolved. The market is questioning the viability of such a power race in a context where investments far exceed the revenues generated.

What’s the Verdict?

With this Broadcom partnership, OpenAI is betting big on achieving greater independence and efficiency. It remains to be seen whether these custom chips will live up to their technical and economic promises. There’s a real risk that this strategy could contribute to the current speculative bubble surrounding AI. However, if successful, OpenAI could finally take control of its hardware destiny.