US Justice Department probes Nvidia's Groq licensing deal for antitrust concerns

Key facts

• The Justice Department is investigating Nvidia's $20 billion licensing deal with AI chip startup Groq.
• Senators Elizabeth Warren and Richard Blumenthal are also probing the deal for potential antitrust violations.
• The deal involves Nvidia licensing Groq's LPU technology and hiring key Groq personnel, including its founder and president.
• Groq's LPU architecture is seen as a competitor to Nvidia's H100 and H200 GPU product lines.
• Groq's chips can process large language model inference workloads significantly faster than Nvidia GPUs, according to benchmarks.
• Regulators are concerned the deal could consolidate control over AI inference infrastructure and eliminate a nascent competitor.

The US Justice Department has launched an investigation into Nvidia's $20 billion licensing agreement with AI chip startup Groq, seeking to determine if the tech giant is attempting to circumvent antitrust regulations. This move follows scrutiny from Democratic Senators Elizabeth Warren and Richard Blumenthal, who are also examining the deal for potential antitrust violations.

The transaction, announced on Christmas Eve, is structured as a licensing agreement for Groq's Language Processing Unit (LPU) technology, but also includes Nvidia hiring Groq's founder, Jonathan Ross, its president Sunny Madra, and a significant portion of its engineering team. Groq was valued at $6.9 billion just three months prior to this deal.

Regulators and market observers are concerned that this arrangement could lead to the consolidation of control over AI inference infrastructure. Groq's LPU technology is noted for processing large language model inference workloads significantly faster than Nvidia's own GPU products, potentially removing a competitive alternative from the market. The Federal Trade Commission (FTC), under Chair Lina Khan, has been increasingly active in challenging vertical integration deals that may eliminate nascent competitors.

Venture capitalists are worried about the implications for their investments in AI infrastructure startups. The potential for regulatory intervention could lead acquirers to demand steeper discounts on competing startups, impacting exit multiples and venture capital returns. Some analysts suggest the deal's structure, while appearing as licensing, functionally eliminates a competitor, drawing parallels to previous regulatory challenges faced by Microsoft and Meta.