Wedoany.com Report-Dec.25, Nvidia has entered into a non-exclusive licensing agreement with artificial intelligence startup Groq to use Groq's chip technology focused on AI inference, Groq announced in a blog post on Wednesday.

Under the agreement, Nvidia will license Groq's specialized architecture designed for the inference stage, where trained AI models handle user queries and generate responses. While Nvidia holds a leading position in the market for training large AI models, the inference segment features competition from established companies such as Advanced Micro Devices and newer entrants including Groq and Cerebras Systems.

Groq founder Jonathan Ross, who previously played a key role in launching Google's AI chip development program, together with President Sunny Madra and several members of the engineering team, will join Nvidia to help integrate and further develop the licensed technology.

Groq emphasized that it will continue to operate independently, with Simon Edwards serving as the new CEO, and its cloud platform will maintain normal operations without disruption.

The companies did not disclose specific financial details of the transaction. Although some media reports suggested a valuation around $20 billion, neither Nvidia nor Groq commented on those figures.

This arrangement follows a pattern seen in recent years among major technology companies, where valuable innovations and expertise are secured through licensing deals and talent transitions rather than complete acquisitions. Similar approaches have been used in deals involving Microsoft, Meta, Amazon, and Nvidia itself earlier this year.

Bernstein analyst Stacy Rasgon wrote in a note to clients on Wednesday: "Antitrust would seem to be the primary risk here, though structuring the deal as a non-exclusive license may keep the fiction of competition alive (even as Groq’s leadership and, we would presume, technical talent move over to Nvidia)."

Groq's design relies on on-chip SRAM memory rather than external high-bandwidth memory components, which helps avoid supply constraints affecting the broader chip industry. This approach enables faster processing speeds for interactive applications such as chatbots, although it places limits on the maximum size of models that can be supported.

Following a $750 million funding round in September, Groq's valuation rose to $6.9 billion, more than doubling from the previous year. The company has established significant customer agreements, including large-scale partnerships in the Middle East, similar to those secured by competitor Cerebras Systems, which has signaled intentions to pursue a public listing as early as next year.

During his main keynote presentation in 2025, Nvidia CEO Jensen Huang highlighted the company's ongoing efforts to maintain leadership as the artificial intelligence market increasingly shifts focus from model training toward efficient, large-scale inference deployment.

This collaboration reflects continued innovation and strategic alignment in the fast-evolving field of AI hardware, aimed at improving performance and expanding capabilities for real-time artificial intelligence applications.