OpenAI Partners with Broadcom to Develop Custom AI Chips in $10 Billion Deal

OpenAI has officially announced a groundbreaking partnership with semiconductor giant Broadcom to design and manufacture custom AI accelerators, marking the company’s ambitious push toward hardware independence. The collaboration, unveiled today, centers on deploying 10 gigawatts of OpenAI-designed AI accelerators in a multi-year agreement valued at approximately $10 billion[3][4].

The strategic move represents OpenAI’s effort to reduce its reliance on Nvidia hardware while addressing the surging computational demands of its AI models and products. With over 800 million weekly active users, OpenAI has experienced unprecedented growth that requires massive computing infrastructure to support[3].

Custom Chip Development and Timeline

Under the partnership, OpenAI will design custom AI accelerators while Broadcom handles development and deployment. The companies have signed agreements to begin deploying racks of AI accelerator and network systems starting in the second half of 2026, with full deployment expected by the end of 2029[3].

“Partnering with Broadcom is a critical step in building the infrastructure needed to unlock AI’s potential and deliver real benefits for people and businesses,” said Sam Altman, OpenAI’s co-founder and CEO[3]. The custom chips will allow OpenAI to embed insights gained from developing frontier models directly into the hardware, potentially unlocking new levels of capability and intelligence.

AI chip manufacturing

Following Industry Trends

OpenAI joins a growing list of tech giants developing proprietary AI chips. Google, Amazon, and Meta have all created specialized processors for AI workloads, recognizing the strategic advantages of custom hardware[1][2]. These custom AI chips, often called “XPUs,” are expected to capture significant market share in the AI industry, currently dominated by Nvidia[2].

The partnership with Broadcom comes after CEO Hock Tan recently mentioned a new customer committing $10 billion in chip orders, which sources confirmed was OpenAI[1][4]. The announcement sent Broadcom shares up 15% in early trading, lifting its market value to approximately $1.7 trillion[4].

Meeting Growing Computational Demands

The custom chip initiative addresses OpenAI’s escalating need for computing power. Following the release of GPT-5, Altman announced plans to double the company’s compute fleet over the next five months[2]. The internal-use chips will help OpenAI avoid potential GPU shortages and could result in significant cost savings.

OpenAI initially began exploring custom chip development in 2023 after Altman blamed GPU shortages for API speed and reliability issues[2]. The company has been working with Broadcom since last year, though production timelines remained uncertain until this announcement[4].

Technical Infrastructure

The collaboration will leverage Broadcom’s comprehensive portfolio of Ethernet, PCIe, and optical connectivity solutions. All racks will be scaled entirely with Ethernet and other connectivity solutions from Broadcom, deployed across OpenAI’s facilities and partner data centers[3].

“Custom accelerators combine remarkably well with standards-based Ethernet scale-up and scale-out networking solutions to provide cost and performance optimized next generation AI infrastructure,” said Charlie Kawwas, President of Broadcom’s Semiconductor Solutions Group[3].

The partnership reinforces the importance of custom accelerators and Ethernet as the preferred technology for scale-up and scale-out networking in AI datacenters, marking what Broadcom’s CEO called “a pivotal moment in the pursuit of artificial general intelligence”[3].

This strategic collaboration positions OpenAI to maintain its competitive edge in the rapidly evolving AI landscape while building the infrastructure necessary to advance its mission of ensuring artificial general intelligence benefits all of humanity.

Sources:

Photo by manseok_Kim on Pixabay