**Broadcom to Develop Custom AI Chips for OpenAI**
*By Dwaipayan Roy | Oct 13, 2025*
OpenAI has announced a strategic partnership with Broadcom to develop its first in-house artificial intelligence (AI) processors. This collaboration is a key part of OpenAI’s plan to address the rapidly growing demand for its AI services.
Under the agreement, OpenAI will take charge of designing the custom AI chips, while Broadcom will handle their development and deployment, starting from the second half of 2026.
**Significant Energy Impact**
The partnership aims to roll out custom processors with a combined power consumption of 10 gigawatts (GW). To put this into context, 10GW is enough energy to power over eight million U.S. households — approximately five times the electricity generated by the Hoover Dam. This highlights the immense scale of AI infrastructure required as the tech industry races to build systems capable of matching or surpassing human intelligence.
**Strategic Partnerships and Industry Moves**
This new deal follows closely on the heels of a separate 6GW AI chip supply agreement between OpenAI and AMD, which also includes an option for OpenAI to acquire a stake in the chipmaker. Just days prior, NVIDIA revealed its plans to invest up to $100 billion in OpenAI and provide data-center systems offering at least 10GW of capacity.
Sam Altman, CEO of OpenAI, emphasized the importance of the Broadcom partnership:
*”Partnering with Broadcom is a critical step in building the infrastructure needed to unlock AI’s potential.”*
**Aligning with Industry Giants**
By collaborating with Broadcom, OpenAI aligns itself with cloud-computing leaders like Google and Amazon, both of which are developing custom AI chips to meet skyrocketing demand while reducing dependency on NVIDIA’s expensive and limited processors.
However, it’s worth noting that similar chip development efforts by Microsoft and Meta have encountered delays or have yet to match NVIDIA’s performance.
**Deployment Timeline and Technical Edge**
Broadcom and OpenAI expect to complete the deployment of these new custom AI chips by the end of 2029. This initiative will build upon their current co-development and supply agreements.
Notably, the new AI systems will be fully scaled using Broadcom’s Ethernet and networking technologies. This may provide Broadcom an advantage over smaller competitors like Marvell Technology and directly challenge NVIDIA’s InfiniBand networking solutions.
—
This partnership marks a significant milestone in OpenAI’s hardware strategy as it seeks to power next-generation AI capabilities with custom-designed, high-performance chips.
https://www.newsbytesapp.com/news/science/openai-partners-with-broadcom-to-develop-custom-ai-chips/story
