Groq raises $350M at $3.5B, down from $6.9B before the Nvidia deal

Groq closed a $350 million Series A on August 17, led by the investment firm Disruptive with planned participation from Nvidia. The round values the company at $3.5 billion, according to Groq’s own announcement.

That’s about half the $6.9 billion Groq carried last September, a drop of roughly 49% by our calculation. A company spokesperson told TechCrunch that Groq doesn’t read this as a down round, but as a new valuation for the “post-Nvidia-licensing-deal version of Groq”. That framing leans on a real change, because the company that raised at $6.9 billion no longer exists in the same shape.

What happened in between was the largest deal Nvidia has ever done. In December 2025, Nvidia agreed to buy assets from Groq for about $20 billion in cash, CNBC reported, citing Disruptive chief executive Alex Davis. Groq itself described the arrangement as a non-exclusive licensing agreement for its inference technology. Founder and CEO Jonathan Ross and president Sunny Madra left for Nvidia, and finance chief Simon Edwards took over as CEO of what stayed behind.

When	Event	Amount	Valuation
September 2025	Financing round led by Disruptive	$750M	$6.9B
December 2025	Nvidia buys assets, licenses inference tech	~$20B	Not applicable
June 2026	Round to fund the cloud pivot	$650M	Not disclosed
August 2026	Series A led by Disruptive	$350M	$3.5B

What stayed behind is a cloud business. Groq now runs 13 data centers across North America, Europe, the Middle East and Asia Pacific, and it says more than six million developers build on the platform alongside Fortune 500 enterprises. The release says Groq expects to scale from 54 megawatts to more than 200 megawatts in 2027.

That last figure is worth pinning down, because the coverage doesn’t agree. SiliconANGLE reported the starting point as 57 megawatts and the target year as next year. Groq’s own release says 54 megawatts and 2027, so that’s what we’ve used here.

Inference will without a doubt become the largest and most critical layer of AI infrastructure.

Alex Davis, Groq Executive Chairman and CEO of Disruptive, in the funding announcement

The money is earmarked for customers “seeking usage of medium and larger sized clusters of NVIDIA accelerated computing for training and inference”, per the release. Groq is now certified as an Nvidia Cloud Partner, which means it builds and operates to Nvidia’s own reference architecture. That places it in the same category as CoreWeave, Lambda and Nebius, which all run Nvidia silicon and several of which Nvidia has put money into, as TechCrunch noted. Nvidia’s habit of underwriting the buildout it also supplies is by now a pattern rather than an exception.

Demand isn’t really the question. TechCrunch describes inference as in high demand as enterprises scale AI workloads, and the shift toward spending compute at serving time rather than training time points the same way. The harder question is whether operators can turn that demand into returns.

Here the reporting is blunt. TechCrunch notes that CoreWeave posted strong second-quarter revenue growth and landed contracts including Meta and Anthropic, yet investors stayed worried about its heavy capital spending, its reliance on debt, and its exposure to hardware that loses value quickly. Groq’s financials are private, so nobody outside can run those same checks on it.

Two things are still unsettled. Groq says the Series A is subject to customary closing conditions, and it describes Nvidia’s participation as planned rather than committed, without naming an amount. The 54 to 200 megawatt jump is the number that would turn this from a funding headline into an operating one, and that’s the figure to check against next year’s disclosures.