OpenAI has partnered with Broadcom to develop custom AI processors, marking the ChatGPT maker’s first foray into proprietary chip design. According to OpenAI, the collaboration aims to deploy 10 gigawatts of custom accelerators beginning in the second half of 2026, with completion by the end of 2029.
The deal positions OpenAI to reduce its dependence on third-party chip suppliers while controlling infrastructure costs.
OpenAI will design the custom AI processors, which will be developed and deployed in partnership with Broadcom.
The new systems will use Broadcom Ethernet and networking gear, offering an alternative to Nvidia’s InfiniBand.
The move comes as OpenAI assembles a diversified chip strategy. Last week, OpenAI also announced a multi-year supply agreement with AMD for 6 gigawatts of GPU capacity, including warrants that could let OpenAI acquire up to 10% of AMD if milestones are met.
In late September, Nvidia said it plans to invest up to $100 billion in OpenAI and supply systems totaling at least 10 gigawatts, under a letter of intent, according to Reuters.
The timing aligns with OpenAI’s infrastructure buildout to support increasingly compute-intensive AI models. Custom chips could offer performance optimizations tailored specifically to OpenAI’s architectures.
The partnership extends existing co-development agreements between the companies. Both firms confirmed deployment would build on their current supply relationship, though neither disclosed financial terms of the arrangement.
“Our collaboration with Broadcom will power breakthroughs in AI and bring the technology’s full potential closer to reality,” said OpenAI co-founder and president, Greg Brockman.
“By building our own chip, we can embed what we’ve learned from creating frontier models and products directly into the hardware, unlocking new levels of capability and intelligence.”
