$20 billion! Nvidia Acquires Core Assets of AI Chip Unicorn Groq

Nvidia has just completed its largest deal in company history, investing approximately $20 billion in cash to acquire the core assets and talent team of AI chip startup Groq.

The transaction structure is quite unique. Rather than a traditional full acquisition, it resembles a "hollowing out" partnership: Groq's founder Jonathan Ross (who was a key creator of Google's TPU) will lead the core technical team to join Nvidia, responsible for advancing technology implementation; meanwhile, Groq will continue to operate as an independent company, maintaining its cloud service business (GroqCloud).

Groq is highly valued because its developed LPU (Language Processing Unit) performs exceptionally well in AI "inference" tasks. Simply put, if training large AI models is "reading," then inference is the "exam" when AI is applied in real-world scenarios. Groq's chips enable AI to answer questions at very high speed with extremely low latency, previously making it a potential strong competitor to Nvidia in the inference market.

Future Focus Nvidia is leveraging its massive cash reserves to clear competitors through this "quasi-acquisition" model, integrating Groq's high-speed inference technology into its ecosystem, which means its moat in the AI hardware field will become even harder to cross.