Groq just raised $350 million to accelerate its shift from building AI chips to running AI infrastructure, according to TechCrunch AI. The round was led by investment firm Disruptive, with planned participation from Nvidia, and it values the company at $3.5 billion. TechCrunch AI reports that’s a steep drop from the $6.9 billion valuation Groq carried last September, before Nvidia hired away its founder and CEO.
What stands out here is that Groq is no longer trying to beat Nvidia. It’s joining it.
What actually happened
Groq built its name on custom chips called LPUs, or language processing units. The pitch was simple: challenge Nvidia on inference, the compute that runs AI models in real time. That plan cracked when Nvidia struck a $20 billion licensing deal, paid out to Groq’s investors, and pulled in CEO Jonathan Ross along with other top talent.
Without its star engineering team, Groq changed course. It’s now a cloud and data center operator running Nvidia systems, not a chipmaker competing against them. This $350 million follows a $650 million round in June that kicked off the pivot.
The numbers that matter
Here’s where Groq stands today, per TechCrunch AI:
- 13 data centers across North America, Europe, the Middle East, and Asia Pacific
- More than 6 million developers, enterprises, and AI-native companies served
- A plan to scale from 54 megawatts to over 200 megawatts by 2027
- Fresh capital aimed at customers renting medium and large clusters of Nvidia compute for training and inference
Groq insists the lower valuation isn’t a down round. A company spokesperson told TechCrunch it reflects a new starting point for the “post-Nvidia-licensing-deal version of Groq.” Read that how you want, but the market is clearly repricing what this company is now versus what it set out to be.
Why this matters
This is significant because it shows how hard it’s become to challenge Nvidia head-on. Groq had one of the more credible inference-chip stories in the market. It still ended up inside Nvidia’s ecosystem, buying and running Nvidia hardware like everyone else.
That puts Groq in crowded company. Nvidia supplies the GPUs behind CoreWeave, Lambda, and Nebius, and it invests billions into some of those same neoclouds as they race to add capacity. So Nvidia sells the chips, funds the buyers, and now backs Groq too. It’s a tight loop, and it’s worth watching who actually profits from it.
“We are building Groq into the world’s leading AI inference cloud,” said Alex Davis, Groq’s chairman and CEO of Disruptive. “Inference will without a doubt become the largest and most critical layer of AI infrastructure.”
He’s probably right about demand. Inference is exploding as enterprises push AI into production. The harder question is whether neoclouds can turn that demand into durable profit.
The catch on neoclouds
The business model carries real risk. CoreWeave posted strong second-quarter revenue growth and landed big contracts, including deals with Meta and Anthropic. Even so, investors keep flagging the same concerns TechCrunch AI notes:
- Heavy capital spending to build out data centers
- Deep reliance on debt to finance that buildout
- Exposure to hardware that depreciates fast
- An unproven path from growth to free cash flow
Groq’s financials are still private, so we can’t see how it stacks up. But it’s chasing the same expensive playbook, and the same open question hangs over it.
What to watch next
Groq’s 2027 target of 200-plus megawatts is the number to track. Hitting it means real capacity and real customers. Missing it, or funding it with too much debt, would echo the worries already dogging the sector.
For practitioners, more inference capacity is good news. More competition among neoclouds should keep pricing pressure on and options open for teams scaling AI workloads. Just don’t mistake capacity growth for a proven business.
Groq set out to unseat Nvidia. It’s now one of Nvidia’s biggest customers and partners. Whether that’s a smart survival move or a sign of how narrow the road really is, the next two years will tell. Full details are available at the original TechCrunch AI report.