"Powering the Next Wave of Intelligence—OpenAI Teams Up with Broadcom"
OpenAI’s chip jump with Broadcom represents a daring shift from dependency to design—10 gigawatts of proprietary AI horsepower, designed to scale and designed to own.
The race to AI autonomy just got hardware-deep.
OpenAI and Broadcom announced a groundbreaking partnership with the aim of co-designing 10 gigawatts of bespoke AI accelerators in a significantAI infrastructure shift. OpenAI will, for the first time, design chips and systems in-house and integrate its frontier model insights directly into hardware. The move represents a move toward self-sufficiency while maintaining partnerships with vendors such as Nvidia and AMD.
The clusters are to be designed and delivered in close collaboration with Broadcom, utilizing its Ethernet, PCIe, and optical connect technologies to provide scalable, power-efficient AI clusters. Installation gets underway in the second half of 2026 and finishes up in the fourth quarter of 2029 and covers OpenAI’s sites and partner data centers.
The stock gained over $150 billion in market capitalization after the report, although the executives refuted OpenAI as the $10 billion mystery customer in the previous report. Both CEOs stressed the partnership in developing the artificial general intelligence (AGI) and filling the global AI demand.
Having more than 800 million weekly active users, OpenAI’s surge highlights the need for comprehensive infrastructure. The partnership not only reinforces the market-leading position in AI networking held by Broadcom but also enables OpenAI to innovate the limits of capability, intelligence, and access in AI systems in the interest of all humankind.