Groq Seeks $650M from Existing Investors to Build Inference Neocloud: why it matters for developers

Groq is asking existing backers for $650 million in an internal round to fund a pivot from hardware sales to a hosted inference neocloud built on its homegrown AI chip and stack.

Groq has asked its existing investors for $650 million in new financing as it pivots from primarily selling hardware to building an inference neocloud that runs hosted model inference for developers and enterprises. The raise is described as an internal round intended to bankroll the company’s shift to offering hosted runtime services on top of its homegrown AI chip and systems, a move the company frames around demand for inference — the runtime processing that occurs after an AI prompt.

In December, Groq struck a so‑called not‑an‑acquisition agreement with NVIDIA for a reported $20 billion that included licensing Groq’s hardware technology to NVIDIA and saw some senior Groq employees depart for the chip maker. According to reporting, that arrangement provided cash to Groq’s investors in lieu of a full acquisition, leaving the startup to pursue a standalone operating strategy funded by follow‑on investment.

The proposed $650 million round is structured to come from existing backers, led internally by the company’s interim chief executive Adam Winter and interim CFO Matt Eng. Two investors, Disruptive and Infinitium, have committed to fill the round if other current shareholders decline to take up their pro rata shares. customers.