"OpenAI is teaming up with Broadcom to produce its own computer chips to power its AI data centers. The deal is the latest in a series of partnerships designed to reduce the company's reliance on Nvidia and secure enough computing power to fuel apps like ChatGPT and Sora and realize its mission to develop superintelligent AI. OpenAI said designing its own chips allows it to "embed what it's learned from developing frontier models and products directly into the hardware, unlocking new levels of capability and intelligence.""
"The partnership, announced Monday, will enable OpenAI to develop and deploy "10 gigawatts of custom AI accelerators" using its own chips and systems. To put that number in context, the output of a typical nuclear reactor is around one gigawatt. Broadcom is expected to start deploying racks of equipment in the second half of 2026 and the deal should finish by the end of 2029, the announcement said."
OpenAI is partnering with Broadcom to design and manufacture custom AI accelerator chips and systems for its data centers. The partnership aims to develop and deploy 10 gigawatts of custom AI accelerators, a scale compared to the output of multiple nuclear reactors. Broadcom plans rack deployments beginning in the second half of 2026 with completion by the end of 2029. The move supplements prior 6 GW and 10 GW deals with AMD and Nvidia and follows changes to OpenAI's exclusive compute arrangement with Microsoft. Major tech firms are increasingly building custom chips to diversify supply and capacity, benefiting companies like Broadcom while not yet displacing Nvidia.
Read at The Verge
Unable to calculate read time
Collection
[
|
...
]