Groq raises $650 million to scale its AI inference cloud business and accelerate deployment of its custom LPU architecture.

Groq has raised $650 million in new funding to scale its AI inference cloud business and accelerate deployment of its custom LPU architecture. The funding arrives as the AI infrastructure market shows intensifying competition around inference workloads, particularly following Nvidia's recent $20 billion deal signals.

The capital will support Groq's expansion of inference services built on its proprietary Language Processing Unit technology. LPUs are designed to optimize the specific computational patterns of running trained AI models in production, an increasingly critical segment as enterprises deploy generative AI applications at scale.

For the AI buildout, Groq's push into inference represents an alternative architectural approach to Nvidia's dominance in AI chips. A successful scale-up could diversify the infrastructure options available for inference workloads—the lower-latency, cost-sensitive portion of the AI stack that operates after model training. This matters because inference will comprise the majority of long-term AI compute spending once training commoditizes, making efficiency gains in this layer economically significant.