On December 24th, 2025, NVIDIA and AI chip startup Groq announced a non-exclusive AI inference technology licensing agreement, alongside a transaction reported to be valued at $20 billion. The announcement was disclosed through company statements and confirmed through media reporting on the same day.

The agreement centers on NVIDIA licensing Groq’s inference technology, while Groq remains an independent company. NVIDIA confirmed the reported transaction size publicly, while Groq outlined the licensing structure and leadership changes associated with the deal.

NVIDIA Groq Announce AI Inference Deal Image

Above: An image of NVIDIA President & CEO Jenson Huang giving a keynote speech at CES 2025.

Deal Structure

According to Groq’s announcement, the agreement grants NVIDIA a non-exclusive license to Groq’s inference technology. Groq stated that the arrangement is intended to accelerate AI inference deployment at a global scale, while allowing the company to continue operating independently and licensing its technology more broadly.

Groq said the licensing agreement covers inference technology rather than manufacturing assets, and that the company will continue to develop and support its own products and services independently.

As part of the transaction, Groq founder and Chief Executive Officer Jonathan Ross and President Sunny Madra, along with additional members of Groq’s technical team, will join NVIDIA. Groq said Simon Edwards has been appointed as its new Chief Executive Officer following Ross’s departure.

Groq did not indicate that the agreement changes its existing customer relationships or alters its current product roadmap.

NVIDIA described the transaction as a technology licensing deal rather than an acquisition. Groq retains ownership of its intellectual property and continues to operate its business, including its existing products and services.

Technology Focus

Groq develops specialized AI processors designed for inference workloads. Its architecture emphasizes deterministic execution and high throughput performance for running trained AI models, rather than for training. The licensed technology is intended to complement NVIDIA’s existing AI hardware and software offerings, which already support both training and inference workloads.

NVIDIA stated that the agreement is intended to support its AI inference efforts. The companies did not disclose specific product timelines, deployment plans, or integration details associated with the licensed technology.

Transaction Details

The transaction was reported to be valued at approximately $20 billion, a figure that NVIDIA later confirmed publicly. The companies did not disclose detailed financial terms beyond this amount.

Because the agreement is structured as a licensing arrangement rather than an acquisition, it does not involve regulatory review processes typically associated with large technology mergers. Groq said the non-exclusive nature of the license allows it to continue operating independently and engaging with other partners.

AI Hardware Market

NVIDIA and Groq both framed the agreement around inference-focused workloads, reflecting the role of inference hardware in deploying trained AI models at scale. The announcement did not include projections regarding revenue impact, deployment scale, or product alignment. Both companies characterized the agreement as an expansion of technical collaboration, rather than a shift in their broader corporate strategies.

Summary

The NVIDIA and Groq announcement establishes a non-exclusive licensing arrangement focused on AI inference technology, alongside the transition of key Groq executives and engineers to NVIDIA. Reported to be valued at roughly $20 billion, the deal expands NVIDIA’s inference capabilities while leaving Groq independent and operational.