AI chip startup Groq reportedly raising $650M after Nvidia's $20B deal

AI chip startup Groq is reportedly raising $650 million from existing backers to expand its inference cloud business, following Nvidia's reported $20 billion 'not-acqui-hire' of the company. The round is being led by interim CEO/CFO Adam Winter, per TechCrunch, and signals that even after Nvidia's massive talent-and-tech deal, Groq's standalone inference-cloud ambitions continue with fresh capital.

The structure is notable: a $20B not-acqui-hire — where Nvidia absorbs key talent and technology without a full acquisition — followed by a separate $650M growth round suggests Groq's inference-cloud entity persists as a going concern. Groq's pitch has long been ultra-low-latency inference via its LPU architecture, positioned against GPU-based serving for high-throughput, latency-sensitive workloads.

Mechanically, the timing dovetails with the industry's pivot toward inference economics. As enterprises battle runaway token bills (Amazon's tokenmaxxing crackdown, Google's Flash pitch), specialized inference hardware that lowers cost-per-token becomes strategically valuable — which helps explain both Nvidia's interest and the fresh capital.

Competitively, the arrangement keeps Groq in the market while giving Nvidia inference optionality and hedging its GPU-centric model. Skeptics will question the unusual deal structure and whether a $650M raise can meaningfully scale an inference cloud against Nvidia's own offerings and AWS/Azure/Google's managed inference. Watch how Groq deploys the capital, whether the Nvidia relationship constrains or enables its cloud business, and how inference-specialist economics hold up as frontier models keep cutting prices.