Nvidia has agreed to acquire assets from AI chip startup Groq for approximately $20 billion in cash, marking the chipmaker's largest acquisition in company history. The deal comes just three months after Groq raised $750 million at a $6.9 billion valuation, representing a nearly threefold premium.
The Nvidia Groq acquisition was confirmed by Alex Davis, CEO of Disruptive, which led Groq's September funding round. Investors included BlackRock, Neuberger Berman, Samsung, Cisco, and 1789 Capital, where Donald Trump Jr. serves as a partner.
Strategic Move Into AI Inference Market
The transaction positions Nvidia to expand beyond its dominance in AI training chips into the rapidly growing AI inference market. While Nvidia's GPUs power the training of large AI models, Groq specializes in inference technology—the stage where trained models generate responses, process queries, and deliver real-time AI applications to users.
Groq's language processing units (LPUs) offer faster, more energy-efficient inference for certain AI workloads compared to traditional GPUs. The chips utilize SRAM memory technology to enable extremely low-latency processing, a critical advantage as demand shifts from model training toward deployment and inference.
In an email to Nvidia employees obtained by CNBC, CEO Jensen Huang explained the strategic rationale: "We plan to integrate Groq's low-latency processors into the NVIDIA AI factory architecture, extending the platform to serve an even broader range of AI inference and real-time workloads."
Deal Structure Avoids Regulatory Scrutiny
Notably, the Nvidia Groq deal is structured as a non-exclusive licensing agreement rather than a traditional corporate acquisition. Groq announced Wednesday it has "entered into a non-exclusive inference technology licensing agreement with Nvidia for Groq's inference technology." The startup will continue operating independently under new CEO Simon Edwards.
As part of the arrangement, Groq founder and CEO Jonathan Ross, President Sunny Madra, and other team members will join Nvidia to integrate the licensed technology. The deal excludes Groq's nascent GroqCloud business, which will continue operating without interruption.
Analysts suggest the licensing structure helps Nvidia avoid lengthy regulatory approvals that derailed its $40 billion attempted Arm acquisition in 2022. "This transaction is essentially an acquisition without being labeled one to avoid regulators' scrutiny," noted Hedgeye Risk Management analysts.
Market Implications and Competitive Pressure
The acquisition reflects intensifying competition in AI chip markets. Tech giants including Google, Amazon, Meta, and OpenAI have all developed custom AI chips to reduce dependence on Nvidia hardware due to supply constraints, high costs, and strategic considerations.
Groq was founded by former Google engineers who developed the tech giant's Tensor Processing Units. The startup positioned its LPUs as specialized alternatives to Nvidia's general-purpose GPUs for inference workloads, representing emerging competitive pressure in a market Nvidia has dominated.
With $60.6 billion in cash and short-term investments as of October, Nvidia has substantial financial capacity for strategic acquisitions. The company's previous largest deal was its $6.9 billion purchase of Israeli networking firm Mellanox in 2019, which provided crucial technology for building AI data center infrastructure.
The Nvidia Groq acquisition surpasses that transaction by nearly threefold, signaling the chipmaker's conviction that controlling both training and inference technology is essential to maintaining dominance as the AI market matures and shifts toward deployment at scale.
