Groq raised $350 million at a $3.5 billion valuation, capital that will fund the startup's transition from AI chipmaker to a neocloud operator running Nvidia GPUs. The round was led by investment firm Disruptive with planned participation from Nvidia, the same company that gutted Groq's founding team eight months ago in a $20 billion licensing deal. The new valuation is roughly half the $6.9 billion mark Groq held last September, before founder and CEO Jonathan Ross and other top talent departed for Nvidia.
A Groq spokesperson told TechCrunch the company doesn't view this as a down round, but rather as a fresh valuation for what it called the "post-Nvidia-licensing-deal version of Groq." That framing matters because the company today is a fundamentally different business than the one that carried the $6.9 billion tag. Then, Groq was building its own language processing units, or LPUs, purpose-built silicon meant to undercut Nvidia on inference workloads. Now it runs Nvidia's chips inside its own data centers.
The pivot began in June, when Groq raised $650 million to begin buying Nvidia systems and scaling data center capacity. Today the company operates 13 data centers across North America, Europe, the Middle East, and Asia Pacific, serving more than 6 million developers, enterprises, and AI-native companies. It plans to grow from 54 megawatts of current capacity to more than 200 megawatts by 2027, a nearly fourfold expansion aimed at customers running medium and large training and inference clusters.
Key facts
- 01Groq raised $350M led by Disruptive with Nvidia participating, valuing the company at $3.5B.
- 02The new mark is down from $6.9B in September, following Nvidia's $20B licensing deal for Groq's founder and top talent.
- 03Groq plans to scale from 54 megawatts today to more than 200 megawatts by 2027 across 13 data centers.
- 04The company serves more than 6 million developers, enterprises, and AI-native customers on Nvidia hardware.
- 05This follows a $650M round in June that kicked off Groq's shift from LPU chipmaker to neocloud operator.
Alex Davis, Groq's chairman and CEO of Disruptive, framed the strategy plainly in a statement announcing the round. The bet is that inference — the compute cycles used to actually run models in production, rather than train them — will become the dominant category of AI infrastructure spend as enterprises push generative AI into products. That thesis is shared across the neocloud category, but the economics are still unproven.
CoreWeave, the most public comparison, posted strong Q2 revenue growth and landed contracts with Meta and Anthropic. Investors have nonetheless remained skeptical of the model, citing CoreWeave's heavy capital expenditures, its dependence on debt to buy GPUs, and the fast depreciation curve on the underlying hardware. Turning top-line growth into free cash flow remains the open question for every operator in this segment, Groq included.
Groq's financials remain private, so it's not yet possible to judge unit economics against CoreWeave, Lambda, or Nebius — the three other Nvidia-aligned neoclouds absorbing the capacity buildout. Nvidia sits on both sides of these transactions, supplying the GPUs and, increasingly, taking equity stakes in the clouds that deploy them. Its participation in the Groq round fits that pattern, extending a financial web that ties model developers, cloud operators, and chip supply into a single ecosystem centered on Jensen Huang's company.
The $20 billion licensing deal that pulled Ross and Groq's core engineering team into Nvidia last winter effectively ended Groq's original mission of building an alternative to Nvidia's data center chips. That deal also paid out Groq investors, which helps explain the spokesperson's insistence that the current $3.5 billion valuation isn't a markdown — the earlier valuation was retired along with the LPU roadmap. What remains is a data center operator with brand recognition, existing customer relationships across 6 million users, and a clear runway to buy more Nvidia gear.
The competitive question is what differentiates Groq from the other neoclouds now that everyone is buying from the same supplier. Scale, siting, energy contracts, and financing terms become the levers, alongside software layers that make it easier for customers to move workloads on and off the cloud. Groq is entering the buildout race later than CoreWeave and with less capital, but with a customer base already in place from its inference API business.
The bigger structural story is how quickly the AI chip challenger category has collapsed into the Nvidia ecosystem it was meant to disrupt. Groq's LPU thesis was one of the more credible attempts to build inference silicon that could compete on price and latency; that project is now inside Nvidia. For the AI infrastructure market, this consolidation removes competitive pressure on Nvidia margins in the near term and pushes the next round of differentiation up the stack — into data center design, financing structures, and the software tooling neoclouds wrap around commodity GPU capacity.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
