Groq raises $350 million after ditching its own chips for Nvidia's

Once an Nvidia rival, the startup is now an Nvidia customer. Here is what happened and what it means for the cloud companies racing to run your AI.

AI2Day Newsdesk3 min read
A close-up of a modern office telephone handset resting on a desk, with soft warm light and abstract blue-green waveform visualisations floating gently above it
Share

Key points

  • Groq raised $350 million in a round led by investment firm Disruptive, with Nvidia among the planned participants.
  • The new deal values Groq at $3.5 billion, down from $6.9 billion in September 2024.
  • Groq shifted from building its own AI chips to operating Nvidia-powered data centres after Nvidia hired away its founder and top team.
  • The company now runs 13 data centres across four regions and serves more than 6 million developers and businesses.
  • Groq plans to grow its power capacity from 54 megawatts to more than 200 megawatts by 2027.

Groq used to want to beat Nvidia. Now it buys from them.

The California startup just closed a $350 million funding round, first reported by TechCrunch, to fund a sharp turn in its business. Groq had spent years building its own custom chips, called LPUs (language processing units, designed specifically to run AI at speed), to take on Nvidia in the market for AI computing. That plan collapsed after Nvidia hired Groq's founder Jonathan Ross and several senior colleagues as part of a licensing deal, stripping the startup of its core team.

What does Groq actually do now?

Groq is now a neocloud, a company that rents out computing power to businesses rather than building the chips that power it. It operates 13 data centres across North America, Europe, the Middle East, and Asia Pacific, and says it serves more than 6 million developers, enterprises, and AI-native companies. The twist: the chips inside those data centres are Nvidia's.

That makes Groq a customer of the company it once tried to out-compete. Groq's chairman and new CEO Alex Davis put it plainly: "Inference will without a doubt become the largest and most critical layer of AI infrastructure." Inference, in plain terms, is the computing work a system does every time it answers a question or completes a task, as opposed to the upfront work of teaching it.

Should investors be worried about the lower valuation?

Groq's value has dropped from $6.9 billion to $3.5 billion, but the company insists this is not a down round in the traditional sense. A spokesperson described it as setting a fresh starting point for "the post-Nvidia-licensing-deal version of Groq," essentially arguing the old number belonged to a different company.

Wider concerns about the neocloud business linger regardless. Rival CoreWeave reported strong revenue growth but still spooked investors with high debt levels and hardware that loses value quickly. Groq's own finances remain private.

Milestone Detail
June 2025 round $650 million raised to begin the pivot
Latest round $350 million, led by Disruptive
Current valuation $3.5 billion
Previous valuation $6.9 billion (September 2024)
Data centres today 13 across four global regions
Power target by 2027 200-plus megawatts

What does this mean for ordinary users and businesses?

If your company uses AI tools that run on rented cloud computing, the neocloud race matters to you. More competition for Nvidia capacity means pricing pressure could eventually work in your favour. For now, Groq is pitching itself at businesses that need medium to large clusters of computing for training and running AI models.

Nvidia, meanwhile, keeps winning either way: it collects revenue from Groq and rivals like CoreWeave, Lambda, and Nebius, and has invested directly in several of them.

© 2026 AI2Day