Groq Raises $647M in Growth Funding as AI Inference Infrastructure Becomes Standalone Investment Thesis

AI inference cloud company Groq has closed $647 million in new growth funding led by Disruptive and Infinitum, with participation from existing investors. The company plans to use the capital to expand processing capacity across its 13 data centers in North America, Europe, the Middle East, and Asia-Pacific, according to BigGo Finance.

The funding round follows Groq’s non-exclusive licensing deal with Nvidia in December 2025, valued at approximately $20 billion. The industry labeled the transaction a “not-acqui-hire,” in which Nvidia secured access to Groq’s core technology and talent while leaving the company as an independent legal entity. Existing investors realized significant gains through the licensing deal, and that capital has cycled back into the current growth round, BigGo Finance reported.

The LPU Inference Bet

Groq’s proprietary LPU (Language Processing Unit) is designed to increase response generation speed for large language models compared to conventional GPUs. The company claims to process trillions of tokens weekly across more than 5 million developers and thousands of enterprise customers.

Speed matters for agent workloads specifically. Multi-step agent tasks that chain multiple LLM calls together compound inference latency at each step. A 10-step agent workflow running on a slow inference backend accumulates delays that make real-time operation impractical. Groq’s pitch is that its LPU eliminates that bottleneck.

Inference as a Funding Category

Groq is not alone in attracting inference-focused capital. Baseten secured $1.5 billion in its own funding round around the same time, according to BigGo Finance. The parallel raises signal that investors now view AI inference infrastructure as a standalone market rather than a subset of GPU cloud computing.

The shift tracks a broader industry pattern: as the AI industry’s center of gravity moves from model training to inference, the companies that can serve tokens fastest and cheapest are positioning themselves against Nvidia’s GPU-dominant inference market. Groq’s weapons are speed and low cost. Nvidia is countering by expanding into networking infrastructure, building out its “AI factory” vision of integrated GPU and networking systems through strategic investments in Coherent, Lumentum, and Marvell Technology.

Groq’s last disclosed valuation was $6.9 billion when it raised $750 million in September 2025. While some personnel moved to Nvidia after the licensing deal, the company is ramping up hiring with this new round.

For agent builders, the practical implication is straightforward: inference speed and cost are becoming differentiated infrastructure choices, not commodities. The provider an agent runs on affects how fast it can think, and how much each thought costs.