Startup Groq has raised $350 million in a round led by investment firm Disruptive, with planned participation from Nvidia, in a move supporting its transition from a company developing custom artificial intelligence chips into a cloud and infrastructure provider relying on Nvidia graphics processing units.
The round raises Groq’s valuation to $3.5 billion, below the $6.9 billion valuation it reached in September 2025. The previous valuation came only months before the company’s founder and CEO, Jonathan Ross, and several senior employees moved to Nvidia as part of a licensing deal worth $20 billion, with the proceeds paid to Groq’s investors.
The company said it does not consider the new valuation a down round, but rather a valuation of the new version of Groq following the licensing deal with Nvidia. This reflects a change in the nature of the business more than the continuation of its previous chip-company model.
From Chips to the Cloud
Groq had focused on developing its own chips, which it called language processing units, or LPUs, with the aim of competing with Nvidia in inference operations—that is, running artificial intelligence models and responding to their requests in real time. But after losing its founder and core team, the company shifted from a specialized chipmaker model to operating cloud services and data centers using Nvidia systems.
Groq had raised $650 million in June to begin this transition. The company currently operates 13 data centers across North America, Europe, the Middle East, and the Asia-Pacific region, and says it serves more than 6 million developers, companies, and organizations relying on artificial intelligence.
What Changes Practically for Customers?
The new funds target customers that need to use medium- and large-sized clusters of Nvidia accelerated computing to train and run artificial intelligence models. Groq also plans to increase its operational capacity from 54 megawatts currently to more than 200 megawatts in 2027, showing that its expansion depends on building substantial physical data center capacity rather than offering only a software service.
The company’s messaging focuses on making it “the world’s leading artificial intelligence cloud for inference.” Alex Davis, Groq’s chairman and the CEO of Disruptive, said inference will become the largest and most important layer of the artificial intelligence infrastructure. This focus indicates that Groq is betting on growing demand from enterprises as artificial intelligence usage shifts from experimentation to continuous operational workloads.
A Model That Requires Significant Capital Expenditure
Groq’s financial information remains private, so the material does not provide details about its revenue or profitability. However, its transition places it directly within Nvidia’s artificial intelligence infrastructure ecosystem, a relationship also shared by specialized cloud providers CoreWeave, Lambda, and Nebius, as Nvidia supplies these companies with graphics processing units and invests in some of them amid the race to increase capacity.
The ability of specialized cloud companies to generate long-term returns remains under scrutiny. CoreWeave recorded strong revenue growth in the second quarter and signed large contracts with Meta and Anthropic, but investors remained concerned about its high capital expenditures, heavy reliance on debt, exposure to the risk of hardware rapidly losing value, and its ability to convert growth into free cash flow.