Groq Raises $350M at $3.5B Valuation to Expand AI Inference Cloud

Groq has raised $350 million at a $3.5 billion valuation as the startup accelerates its shift from designing AI chips toward operating an AI inference cloud built around large-scale computing infrastructure.

Investment firm Disruptive led the financing, with Nvidia planning to participate. The new valuation is below the $6.9 billion that Groq reached in September, before a licensing agreement that led Nvidia to hire founder Jonathan Ross and other senior Groq employees.

Groq has said the lower figure reflects the company’s value after that transaction, rather than a conventional down round. The latest financing follows a $650 million raise in June, bringing the company’s recent funding to about $1 billion.

Groq shifts deeper into AI infrastructure

Groq originally developed its own Language Processing Units, or LPUs, as an alternative to GPUs for AI inference, the computing required to run trained AI models and generate responses. Following the Nvidia licensing agreement and the departure of key personnel, the company has increasingly focused on cloud infrastructure and operating Nvidia accelerated computing systems.

The startup now operates 13 data centres across North America, Europe, the Middle East and Asia-Pacific. Groq says its infrastructure serves more than 6 million developers, enterprises and AI-focused companies.

The new funding will support customers seeking medium and larger Nvidia-powered computing clusters for both AI training and inference. Groq also plans to expand its power capacity from about 54 megawatts to more than 200 megawatts by 2027.

“We are building Groq into the world’s leading AI inference cloud,” Alex Davis, Groq’s chairman and CEO of Disruptive, said in a statement. Davis said the company expects inference to become one of the most important layers of AI infrastructure as more organisations deploy models at scale.

Groq enters a competitive neocloud market

The pivot places Groq more directly alongside specialised AI cloud providers such as CoreWeave, Lambda and Nebius. These companies operate large fleets of accelerated computing hardware and sell access to businesses that need substantial AI capacity without having to build their own data centres.

Nvidia plays a significant role across that market by supplying GPUs to many neocloud providers while also investing in parts of the ecosystem. Groq’s transformation therefore changes its relationship with Nvidia from primarily competing in AI inference hardware to becoming a customer and infrastructure partner.

The model also carries significant capital requirements. Building data centres and acquiring large numbers of GPUs requires substantial spending, while hardware can depreciate quickly as newer generations arrive. Those pressures have remained a focus for investors evaluating rapidly growing AI infrastructure providers.

For Groq, the $350 million round provides additional capital to expand capacity and pursue larger AI workloads as it establishes a new identity after the Nvidia licensing deal. The company’s next phase will depend less on selling its own processors and more on attracting customers to an increasingly competitive market for AI compute and inference services.