Groq raises $350M at half its 2025 price, with Nvidia on the cap table

Groq has closed a $350 million Series A at a $3.5 billion valuation, announced Monday, Aug. 17, led by the Dallas investment firm Disruptive with planned participation from Nvidia ^{[1] [2] [5] [3]}. That is roughly half the $6.9 billion the company carried in September 2025 ^{[11] [23]}, and the new money comes from the incumbent that licensed Groq's technology and hired its founder eight months ago ^{[4] [19]}.

The sequence matters more than the headline number. Groq was founded in 2016 by Jonathan Ross, a former Google engineer who worked on the Tensor Processing Unit, and spent nearly a decade selling its Language Processing Units as a faster, cheaper alternative to Nvidia GPUs ^[13]. In September 2025 that pitch cleared $750 million at a $6.9 billion valuation, led by Disruptive with BlackRock, Neuberger Berman and Deutsche Telekom Capital Partners joining ^[11]. In December 2025, according to Tech Funding News, Nvidia agreed to pay a reported $20 billion to license Groq's inference technology and hire Ross, president Sunny Madra and roughly 90% of the engineering team that built it ^[4]. Groq said it would keep operating as an independent business ^[12]. It then raised $650 million in June 2026 and reset its valuation without publishing a number ^{[14] [18]}, and has now taken $1 billion in about two months ^[14].

Read the two prices together. The reported licensing payment is about 5.7 times what the entire remaining company is now worth ^[24]. The market has separated the intellectual property from the corporate wrapper and paid accordingly.

What is left is an operator, not a challenger. Groq runs 13 data centres, serves more than six million developers, Fortune 500 enterprises and thousands of AI-native companies ^[15], and plans to grow capacity from 54MW to more than 200MW by 2027, roughly a 3.7x build ^{[16] [22]}. It remains an Nvidia Cloud Partner, certified to run Nvidia accelerated computing under Nvidia's own reference architecture ^[6]. Per the company's release, the Series A will help Groq support customers "seeking usage of medium and larger sized clusters of Nvidia accelerated computing for training and inference" ^[7]. Executive chairman Alex Davis, who is also CEO of Disruptive, said the team has "unmatched experience operating LPUs at scale" and that he looks forward "to continuing our partnership with Nvidia" ^{[9] [10] [2]}.

For anyone procuring inference capacity, that is the operative fact. The vendor you may have signed for LPU price-performance is now partly funded by Nvidia and marketing Nvidia clusters. The category of genuinely non-Nvidia inference supply is thinner than the funding headlines suggest: Fireworks AI closed a $1.5 billion Series D at a $17.5 billion valuation in July 2026, backed in part by Nvidia, five times Groq's new mark ^{[21] [25]}, while Together AI raised $800 million at $8.3 billion in the same month ^[20]. Cryptopolitan notes the license-and-hire structure Nvidia used is now common, with Google and Meta taking technology and engineers from startups rather than buying them ^[17].

Three things to watch. Whether Nvidia's participation actually closes, since it is described as planned rather than funded ^[3]. Whether the 200MW build lands on schedule, because operational scale is now the whole argument ^{[16] [7]}. And what your renewal quotes look like once Groq's capacity is predominantly Nvidia silicon it must buy at market, rather than LPUs it designed ^{[6] [7]}. Davis argues inference "will without a doubt become the largest and most critical layer of AI infrastructure" ^[8]. He may be right about the layer and still be describing a reseller.