Exclusive: Nvidia strikes record $20bn deal for Groq assets in major AI chip play
Nvidia has agreed to acquire key assets from AI chip startup Groq in a deal valued at about $20 billion in cash, according to Alex Davis, chief executive of Disruptive, the investment firm that led Groq’s most recent funding round.
If confirmed, the agreement would mark Nvidia’s largest transaction to date and underline its determination to tighten its grip on the rapidly evolving market for artificial intelligence hardware, particularly in the fiercely competitive area of inference chips.
Davis said the deal came together quickly. Groq raised $750 million in September at a valuation of roughly $6.9 billion, with investors including BlackRock, Neuberger Berman, Samsung, Cisco, Altimeter and 1789 Capital, where Donald Trump Jr is a partner. Disruptive has invested more than $500 million in Groq since the company was founded in 2016.
Groq said in a blog post that it has entered into a non exclusive licensing agreement with Nvidia for its inference technology, without disclosing financial terms. Under the agreement, Groq founder and chief executive Jonathan Ross, president Sunny Madra and other senior engineers will join Nvidia to help advance and scale the licensed technology. The company added that it will continue to operate as an independent business, with finance chief Simon Edwards stepping in as chief executive.
Nvidia declined to comment on the transaction. Chief financial officer Colette Kress did not respond to a request for comment.
According to Davis, Nvidia is effectively acquiring all of Groq’s assets, with the exception of its cloud business. Groq said GroqCloud will continue to operate without interruption.
The scale of the deal dwarfs Nvidia’s previous acquisitions. Its largest purchase until now was the $7 billion acquisition of Israeli chip designer Mellanox in 2019. By the end of October, Nvidia reported $60.6 billion in cash and short term investments, up sharply from $13.3 billion in early 2023.
In an internal email obtained by CNBC, Nvidia chief executive Jensen Huang said the agreement would significantly expand the company’s capabilities.
“We plan to integrate Groq’s low latency processors into the Nvidia AI factory architecture, extending the platform to serve an even broader range of AI inference and real time workloads,” Huang wrote. He added that while Nvidia is licensing Groq’s intellectual property and hiring its talent, it is not acquiring Groq as a company.
The structure mirrors a growing trend among big technology firms, which are increasingly paying large sums to secure cutting edge AI talent and technology without completing full takeovers. In September, Nvidia spent more than $900 million to hire the chief executive and staff of AI hardware startup Enfabrica and license its technology. Microsoft, Meta, Google and Amazon have all pursued similar arrangements in recent years.
Nvidia has also ramped up investments across the AI ecosystem as its cash reserves have grown. It has backed AI infrastructure company Crusoe, model developer Cohere and increased its stake in cloud provider CoreWeave ahead of its public listing. In September, Nvidia said it intended to invest up to $100 billion in OpenAI, with the startup committing to deploy at least 10 gigawatts of Nvidia products, though a formal agreement has yet to be announced. The same month, Nvidia also revealed plans to invest $5 billion in Intel as part of a partnership.
Groq has been targeting revenue of $500 million this year amid surging demand for AI accelerator chips used to speed up inference tasks for large language models. Davis said the company was not actively seeking a sale when Nvidia approached it.
Read More: AI Bubble at $2.1T Is 2025 the Breaking Point? What Happens Next
Founded in 2016 by former Google engineers including Ross, Groq has focused on inference performance. Ross was one of the creators of Google’s tensor processing unit, or TPU, which some companies use as an alternative to Nvidia’s graphics processing units. Early filings show Groq also counted Douglas Wightman, a former engineer at Google X, among its founders, though he left the company in 2019.
Groq is one of several chip startups to gain prominence during the AI boom. Cerebras Systems, a rival specialising in large scale processors, withdrew plans for an initial public offering in October after raising more than $1 billion, though it has said it still hopes to list in the future. Both Groq and Cerebras have secured major deals in the Middle East.
Reuters separately reported that Nvidia has agreed to license Groq’s technology and hire away its chief executive, describing the deal as part of a familiar pattern in which tech giants secure talent and intellectual property while stopping short of outright acquisitions. Nvidia dominates the market for training AI models but faces stronger competition in inference, where rivals such as AMD and startups like Groq and Cerebras are vying for share.
Analysts have noted potential regulatory scrutiny. Bernstein analyst Stacy Rasgon wrote that antitrust concerns appear to be the primary risk, although structuring the agreement as a non exclusive licence may help preserve the appearance of competition, even as Groq’s leadership and technical expertise move to Nvidia.
Groq’s valuation more than doubled over the past year, rising from $2.8 billion in August to $6.9 billion after its September funding round. The company’s approach, which relies on on chip memory rather than external high bandwidth memory, allows faster interactions for chatbots and other AI applications, though it limits the size of models that can be served.
Nvidia has argued that it is well positioned to maintain its lead as AI markets shift from training to inference. In his biggest keynote of 2025, Huang said Nvidia’s architecture and ecosystem would allow it to stay ahead as demand evolves. The Groq deal appears to be a decisive step in that strategy.
