This article first appeared on GuruFocus.
Nvidia (NASDAQ:NVDA) is deepening its relationship with AI startup Groq as the company raises $350 million at a $3.5 billion valuation, a financing that underscores both Nvidia's growing push into AI inference and the dramatic reshaping of Groq following their massive licensing agreement.
Warning! GuruFocus has detected 4 Warning Signs with NVDA.
Is NVDA fairly valued? Test your thesis with our free DCF calculator.
Groq develops technology designed to run AI models after they have been trained, a process known as inference. The company built specialized Language Processing Units aimed at delivering fast, low-latency AI responses, but is now shifting away from primarily selling hardware toward operating an AI inference neocloud, where customers buy access to computing capacity.
The new round, led by investment firm Disruptive, is expected to be announced Monday, Bloomberg reported. Nvidia is also expected to participate.
The valuation is particularly striking. Groq was valued at $6.9 billion after raising $750 million last September, meaning the latest $3.5 billion figure represents a decline of roughly 49% in less than a year.
But Groq has changed substantially since then.
Nvidia entered a non-exclusive licensing agreement for Groq's inference technology in December 2025 in a deal reportedly worth $20 billion. Founder Jonathan Ross, President Sunny Madra and other Groq employees were also set to join Nvidia.
That leaves the remaining Groq business increasingly centered on cloud-based inference rather than competing directly with Nvidia through proprietary hardware.
For Nvidia, the relationship is strategically important. Training has driven enormous GPU demand, but inference becomes increasingly critical as deployed AI models handle billions of user requests. Groq's technology and engineering talent potentially strengthen Nvidia's ability to defend that next phase of AI computing.
The key question is whether Nvidia can translate its Groq relationship into a stronger competitive position in AI inference, where customers increasingly prioritize cost per query, latency and energy efficiency.
Investors should watch Nvidia's upcoming data-center results for evidence that inference demand is becoming a larger growth driver alongside training.
Groq's sharply lower valuation also deserves attention. It suggests that enormous AI infrastructure spending does not guarantee rising valuations for every startup, particularly when business models change dramatically.
For Nvidia, however, the combination of licensing Groq's technology, absorbing key personnel and now investing in the company could give it another strategic lever as AI spending shifts from building models toward running them at massive scale.
