AI Chip Infrastructure OpenAI Commissions Broadcom to Manufacture Its Own AI Chips

Related Vendors

Broadcom Corporation

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

OpenAI is developing its own AI processors for the first time and is relying on Broadcom for manufacturing and rollout. The long-term cooperation includes up to 10 GW of accelerators; initial deployments are planned for the second half of 2026.

As early as the beginning of September, rumors circulated that OpenAI might commission semiconductor provider Broadcom to manufacture custom AI chips. On Monday, October 13, this became a certainty: OpenAI has agreed with the US chip manufacturer Broadcom on the joint development of a proprietary AI chip. The measure is part of a comprehensive strategy to meet the growing demand for computing power in the long term.

The chips are expected to be operational from the second half of 2026 and deliver up to ten gigawatts of computing power. The components will be used in OpenAI's own facilities as well as in data centers of partner companies.

Manufacturing based on own design, utilization of Broadcom's peripheral portfolio

OpenAI takes over chip design, while Broadcom is responsible for the development and deployment of the hardware. Broadcom's complete portfolio of Ethernet, PCIe, and optical connectivity solutions is utilized to enable scalable and energy-efficient systems.

"Our partnership with OpenAI continues to set new industry standards for the design and deployment of open, scalable, and energy-efficient AI clusters," said Dr. Charlie Kawwas, President of the Semiconductor Solutions Group at Broadcom Inc. "Custom accelerators can be excellently combined with standards-based Ethernet scale-up and scale-out network solutions to deliver a cost- and performance-optimized next-generation AI infrastructure. The racks include Broadcom's end-to-end portfolio of Ethernet, PCIe, and optical connectivity solutions, once again reaffirming our leadership in AI infrastructure."

With this move, OpenAI joins a growing list of technology companies investing in their own AI chips – including Google, Amazon, and Microsoft. The goal is to reduce costs, avoid supply bottlenecks, and strengthen control over their own infrastructure. Currently, many leading providers rely on Nvidia as a supplier of GPUs for AI acceleration.

According to OpenAI CEO Sam Altman, the partnership is a key element in building the infrastructure needed to fully realize the potential of AI. Custom chips allow experiences from the development of advanced models to be directly integrated into the hardware.

The project is considered ambitious by analysts. Financing is likely to come from multiple sources, including investment rounds, pre-orders, support from Microsoft, as well as potential partnerships and credit lines.

In parallel, OpenAI recently finalized a deal with AMD for the purchase of chips with a capacity of six gigawatts. Nvidia announced plans to invest up to 100 billion US dollars in a joint infrastructure project with OpenAI.

Despite the ambitions, the move is not seen as a direct threat to Nvidia's market position. The development and scaling of custom chips are considered highly complex. For Broadcom, however, the collaboration is a strategic gain in the booming AI infrastructure market. (sg)