📰 Key Summary

Groq has closed a new $350M funding round led by investment firm Disruptive, with Nvidia also expected to participate, valuing the company at $3.5B — a steep markdown from its $6.9B valuation last September. A company spokesperson told TechCrunch that despite the lower valuation, this isn’t a down round — it’s a repricing for “the new Groq after the Nvidia licensing deal.” Groq originally focused on building its own LPU (Language Processing Unit) chips, aiming to compete with Nvidia in inference computing. But after founder and CEO Jonathan Ross and the core team were poached through a $20B Nvidia licensing deal, the company pivoted to running Nvidia systems as a cloud and data center provider. Groq closed a $650M round back in June to kick off this pivot, with plans to scale its data center power capacity from 54 megawatts to over 200 megawatts by 2027. Groq currently operates 13 data centers across North America, Europe, the Middle East, and Asia-Pacific, serving more than 6 million developers, enterprises, and AI-native company customers. The new funding will support training and inference workloads for mid-to-large Nvidia accelerated compute clusters. Alex Davis, Chairman and CEO of Disruptive, said the goal is to build Groq into the world’s leading AI inference cloud provider. The article also notes that while inference demand is surging as enterprises scale up their AI workloads, whether neocloud providers can turn a long-term profit remains an open question — peer CoreWeave has posted strong revenue growth and landed major deals with Meta and Anthropic, but investors remain wary of its high capex, heavy debt reliance, and hardware depreciation risk. Groq’s financials are not currently public.

💬 JudyAI Lab’s Take

Groq just closed a new $350M funding round, but its valuation dropped from $6.9B last year to $3.5B. The spokesperson insists this isn’t a down round — it’s a repricing for “the new Groq after the Nvidia licensing deal.”

What’s worth noting here isn’t the valuation number itself — it’s the pivot. Groq originally bet on building its own LPU chips to take on Nvidia, but after its core team got poached through a $20B Nvidia licensing deal, the company simply pivoted to running Nvidia systems as a cloud provider — planning to scale data center power from 54 megawatts to over 200 megawatts by 2027, already serving over 6 million developers. This reflects a broader split happening in the inference compute race: instead of building your own chips to fight the giants, “riding on the giant’s infrastructure and focusing on delivering at scale” is becoming the more pragmatic business call. That said, CoreWeave’s case shows that landing big contracts doesn’t automatically solve the profitability problem posed by high capex and hardware depreciation.

Takeaway: if you’re relying on an inference provider, it’s worth taking a closer look at the chip supply chain and capital structure stability behind them.

📅 Source Info

🔗 Further Reading