Groq, the inference-chip company founded by a former Google engineer, In August 2024, Groq raised $640 million at a $2.8 billion valuation in a round led by BlackRock, according to TechCrunch, to scale production of the processors it designs for running large language models quickly.
The round reflects a shift in where the AI infrastructure money is going. Training the largest models drew the first wave of capital; serving them to users, cheaply and fast, is drawing the next one. Groq's pitch is that its processors are dramatically faster than general-purpose GPUs for inference, a claim that matters as the economics of running a model move from a capital question to an operating one.
The constraint
Groq's challenge is manufacturing scale. Designing a fast inference chip is a small part of building a business around it; the rest is fabrication, packaging and a software stack that lets customers move models onto the hardware without rewriting them. The capital funds all three, and the next year will show whether the company can produce at the volume its order book implies.

