OpenAI Becomes Nvidia’s Biggest Customer for Next-Generation AI Chips With Groq Technology
OpenAI has positioned itself as the main customer for Nvidia’s next generation of AI accelerators, which will integrate advanced technology from Groq.
This partnership follows a $20 billion licensing agreement under which Nvidia will incorporate Groq’s LPU architecture to strengthen its inference capabilities.
OpenAI Secures 3 Gigawatts of Dedicated Inference Capacity
A key part of the deal is OpenAI’s reservation of 3 gigawatts (GW) of dedicated inference capacity. For OpenAI, this is not just a performance upgrade, but a necessary step to keep up with the growing demand for its language models.
While traditional GPUs have been the standard for training AI models, Groq’s technology stands out for its extremely low latency during inference — the phase where AI systems generate responses to users. This allows near-instant text generation and helps remove current bottlenecks. In the future, this could mean much shorter waiting times for responses from services like ChatGPT.
Before Nvidia finalized this agreement, OpenAI had explored other hardware options, holding direct talks with startups such as Groq and Cerebras. However, Nvidia’s move to license Groq’s technology and bring in key talent — including CEO Jonathan Ross, a former Google TPU chip designer — was decisive. This allows Nvidia to use Groq’s architecture directly and become the exclusive provider of this hybrid solution.
The new accelerator is expected to be officially presented at GTC 2026. By combining Nvidia’s CUDA ecosystem with Groq’s fast execution speed, OpenAI could improve service performance while also reducing power consumption. Lower energy use would also help reduce operating costs.
The alliance between OpenAI and Nvidia appears stronger than ever, and it could give the company an edge over competitors such as Google.
