Groq Raises $350 Million to Build an Nvidia-Powered Cloud After Its Chip Pivot

Groq is no longer asking investors to view it mainly as an AI-chip rival. Its new $350 million financing will expand a cloud business based on Nvidia-accelerated computing, drawing the company further into the ecosystem it once sought to challenge. The move brings new capital and a major infrastructure goal, but leaves Groq reliant on a costly cloud model whose long-term returns remain uncertain.

The Series A was led by returning investor Disruptive. Nvidia plans to join later, according to Groq’s announcement as reported by SiliconANGLE, but the size of Nvidia’s investment was not disclosed. TechCrunch likewise described Nvidia’s participation as planned rather than completed.

TechCrunch reported that the deal values Groq at $3.5 billion. That is below the $6.9 billion valuation reported for September 2025, before Nvidia hired founder and then-chief executive Jonathan Ross and other top talent through a licensing deal. A Groq spokesperson told TechCrunch the company did not consider the comparison a down round, calling it a new valuation for the post-licensing-deal company.

Groq is funding a changed business

Groq launched in 2016 to develop AI accelerators, including language processing units, or LPUs. It has since pivoted to data-center operations and cloud infrastructure, with TechCrunch reporting that the remaining company operates Nvidia systems. That turns Groq from a prospective hardware competitor into an Nvidia customer and cloud operator.

That shift does not mean Groq’s earlier chip work has disappeared from the cloud’s technical story. SiliconANGLE reported that GroqCloud uses racks containing the Groq 3 LPU and Nvidia Rubin graphics processors. The outlet also reported that Nvidia developed the Groq 3 LPU using technology licensed from Groq, while Groq’s cloud is built on hardware based on technology it developed before the licensing deal. The reported setup complicates a simple before-and-after account: Groq is now buying into Nvidia’s infrastructure ecosystem, but its earlier technology remains part of the reported hardware picture.

The new money is meant to finance that operating model. Groq says the proceeds will support customers seeking medium and larger clusters of Nvidia-accelerated computing for both training and inference. Training is the process of building a model from data; inference is the later work of running a trained model on new requests and producing results.

What Groq says the capital supports

• Expansion of its public cloud, GroqCloud.
• Customer access to medium and larger Nvidia-accelerated clusters for training and inference.

An existing footprint, then a rapid buildout

GroqCloud runs on hardware hosted in 13 data centers worldwide. TechCrunch reported that the sites span North America, Europe, the Middle East and Asia-Pacific, and that Groq serves more than six million developers, enterprises and AI-native companies. Those customer and developer numbers are company-reported figures, rather than independently disclosed measures of revenue or utilization.

The company had already raised $650 million in June to begin the pivot, making this cash infusion its second major reported financing in less than three months. The back-to-back rounds put substantial funding behind a fast conversion from chip development to operating infrastructure at scale. They do not, however, disclose what portion of the planned capacity is already contracted or how quickly it can be brought online.

The bet rests on demand for inference, the repeated execution of AI models after training. Groq Executive Chairman Alex Davis said inference will become the largest and most critical layer of AI infrastructure. That is a company view, not a demonstrated financial outcome for Groq’s cloud business. The supplied reports do not provide Groq revenue, margins, debt or free-cash-flow figures.

Nvidia’s role adds supply and financing ties

Groq’s strategy now occupies a familiar, demanding part of the AI market: cloud companies buy or operate Nvidia equipment and sell access to that computing capacity. TechCrunch noted that Nvidia supplies GPUs to cloud providers including CoreWeave, Lambda and Nebius while investing in some of them. Nvidia’s planned participation in Groq therefore follows an established supplier-and-investor pattern rather than creating an entirely new relationship.

The constraint is that building cloud capacity consumes substantial capital. TechCrunch pointed to investor concerns around high capital expenditures, debt, hardware depreciation and free-cash-flow conversion at peer CoreWeave, even as that company reported strong second-quarter revenue growth and won large contracts. Those concerns do not establish the same outcomes for Groq, whose financials remain private, but they frame the test facing its expansion.

Groq has disclosed the direction of travel: a larger cloud, more Nvidia-accelerated capacity and a focus on training and inference customers. It has not disclosed Nvidia’s eventual check size, a single capacity baseline and deadline across published accounts, or the financial measures needed to judge whether its new cloud operation can turn infrastructure growth into durable returns. The company’s next test is execution against those targets, not simply the size of this round.