OpenAI has joined forces with Broadcom to create its first custom artificial intelligence processors. The partnership marks a major step for the ChatGPT maker as it works to secure computing power for growing demand. According to Reuters, the companies announced the deal on Monday.
Massive Power and Timeline
OpenAI will design the chips while Broadcom handles development and deployment. The rollout starts in the second half of 2026. The companies plan to deploy 10 gigawatts worth of custom chips. That power consumption equals the needs of more than 8 million U.S. households or five times the electricity from the Hoover Dam.
The deal adds to OpenAI’s recent chip investments. Last week, the company unveiled a 6-gigawatt AI chip supply deal with AMD. That agreement includes an option to buy a stake in AMD. Days earlier, OpenAI said Nvidia plans to invest up to $100 billion in the startup. Nvidia will also provide data center systems with at least 10 gigawatts of capacity.
Financial Terms Not Disclosed
The companies did not reveal financial details. It remains unclear how OpenAI will fund the deal. OpenAI CEO Sam Altman called the partnership a critical step in building the infrastructure needed to unlock AI’s potential. Broadcom shares rose more than 7 percent in early trading after the announcement.
Custom Chip Competition Heats Up
The Broadcom partnership places OpenAI alongside cloud computing giants like Google and Amazon. These companies are developing custom chips to meet AI demand and reduce dependence on Nvidia’s processors. Nvidia chips are costly and limited in supply.
The approach carries risks. Similar efforts by Microsoft and Meta have faced delays or failed to match Nvidia chip performance, according to media reports. Analysts believe custom chips do not threaten Nvidia’s dominance in the short term.
Broadcom has become a major winner in the generative AI boom. Its stock price has risen nearly six-fold since the end of 2022. The company announced a $10 billion custom AI chip order in September from an unnamed customer. Some analysts speculated that customer was OpenAI.
Broadcom and OpenAI said the new chip deployment will finish by the end of 2029. The systems will use Broadcom’s Ethernet and networking gear. This gives Broadcom an edge over rivals like Marvell Technology and challenges Nvidia’s InfiniBand networking solution.
