Groq raises $350M to build Nvidia-powered AI inference cloud

Groq has secured $350 million in fresh funding led by Disruptive, marking a strategic shift from its original mission of building AI chips (LPUs) to becoming an Nvidia-powered inference cloud provider. The new valuation of $3.5 billion reflects Groq’s post-Nvidia-licensing-deal reality—a sharp contrast to its $6.9 billion valuation just months prior, when Nvidia acquired key talent through a $20 billion deal. Groq now operates 13 data centers globally, serving over 6 million developers and enterprises, and plans to scale capacity from 54 megawatts to 200+ megawatts by 2027.

The funding will accelerate Groq’s push to become the "world’s leading AI inference cloud," focusing on medium and large Nvidia GPU clusters for training and inference. This pivot positions Groq directly within Nvidia’s ecosystem, joining competitors like CoreWeave and Lambda, which also rely on Nvidia hardware. While Groq’s financials remain private, the company’s bet hinges on inference demand outpacing the profitability challenges faced by neoclouds—such as high capex, debt reliance, and hardware depreciation. The move underscores the growing divide between chipmakers (like Nvidia) and infrastructure providers, where access to compute power, not hardware design, may dictate long-term dominance.