Nvidia acquires AI inference chip maker Groq for $20 billion to consolidate talent and develop competing inference architectures.

Nvidia Corp (NASDAQ:NVDA, XETRA:NVD) is reportedly acquiring most of the assets of AI chip company Groq in a transaction valued at roughly $20 billion in cash. The deal is also said to include a non-exclusive licensing agreement for Groq's inference technology.

Wedbush analysts highlighted the strategic value of talent acquisition in the deal, noting that Nvidia will add Jonathan Ross, one of the key designers behind Google's TPU architecture and Groq's LPU, to its team. "One of our friends likes to say that getting silicon right is not about having the largest chip design team, but rather having the right small group of individuals," the analysts wrote. The analysts suggested the transaction is aimed at long-term positioning in AI inference rather than addressing immediate competitive pressures.

"We, however, think purchasing Groq is less about competitive dynamics today and more about what could happen over the next decade," Wedbush wrote, adding that the future success of the deal will depend on how Groq's technology integrates with Nvidia's broader software and hardware stack. The analysts noted that while $20 billion is a large sum, it represents just over half of Nvidia's estimated net income for the current quarter. "The Groq deal alone doesn't affect NVDA's financials significantly," they wrote.

The analysts additionally flagged the remaining assets of Groq not included in the deal, led by Simon Edwards, Groq's former CFO. They characterized the leftover business as "largely aimed at minimizing regulatory risk" and questioned the long-term prospects for the remaining cloud-focused operations. Wedbush also highlighted the optionality the deal provides Nvidia, with potential new architectural offerings that could address future AI use cases across data center inference, edge applications, or other workloads. Nvidia shares were down 2% in early trade on Monday.