The agreement reflects a trend observed in recent years, where major tech corporations offer substantial sums in partnerships with innovative startups to acquire their technology and talent without fully acquiring the company.
Groq specializes in inference, where pre-trained artificial intelligence models respond to user requests. While Nvidia leads the market for training AI models, it faces increased competition in inference from traditional adversaries like Advanced Micro Devices, along with startups such as Groq and Cerebras Systems.
Also Read: Tokyo inflation eases better than estimated; BoJ likely to continue with interest rate hikes
Nvidia has established a “non-exclusive” license for Groq’s technology, according to Groq. The startup announced that its founder, Jonathan Ross—who played a crucial role in launching Google’s AI chip program—as well as Groq President Sunny Madra and other engineering team members will transition to Nvidia.
An individual close to Nvidia confirmed the licensing arrangement.
Groq has not revealed the financial specifics of the deal. CNBC reported that Nvidia was set to acquire Groq for $20 billion in cash, but neither Nvidia nor Groq have commented on this report. Groq mentioned in its blog that it will continue to operate independently with Simon Edwards as CEO and that its cloud business will remain functional.
In similar recent agreements, Microsoft’s top AI executive emerged from a $650 million deal with a startup branded as a licensing fee, while Meta allocated $15 billion to attract Scale AI’s CEO without purchasing the entire company. Amazon similarly recruited founders from Adept AI, and Nvidia executed a comparable deal this year. These arrangements have undergone regulatory scrutiny, though none have been reversed yet.
“Antitrust appears to be the primary concern here, although structuring the deal as a non-exclusive license may help create an illusion of competition (even as Groq’s leadership and, we assume, technological talent transitions to Nvidia),” stated Bernstein analyst Stacy Rasgon in a note to clients following Groq’s announcement on Wednesday. Moreover, Nvidia CEO Jensen Huang maintains a notably strong relationship with the Trump administration compared to other major US technology firms.
Groq more than doubled its valuation, escalating from $2.8 billion to $6.9 billion in August last year, following a $750 million funding round in September.
Groq is among several startups that do not depend on external high-bandwidth memory chips, allowing them to circumvent the memory constraints affecting the global chip market. This strategy, utilizing a type of on-chip memory known as SRAM, enhances the speed of interactions with chatbots and other AI models, although it also limits the size of models that can be processed.
Groq’s main competitor using this method is Cerebras Systems, which Reuters reported intends to go public as early as next year. Both Groq and Cerebras have secured significant contracts in the Middle East.
Nvidia’s Huang dedicated much of his keynote address in 2025 to arguing that Nvidia would sustain its lead as AI markets transition from training to inference.
