Key Highlights
- Federal investigators are examining whether Nvidia designed its $20 billion Groq licensing agreement to bypass antitrust review processes.
- The December agreement allowed Nvidia to license Groq’s intellectual property while maintaining Groq’s operational independence, with key executives joining Nvidia.
- Congressional members have condemned the structure, suggesting it functions as a de facto acquisition that may reduce competition in AI semiconductors.
- Groq’s latest 3 LPX rack has entered full-scale production, achieving 3,400 tokens per second performance, scheduled for neocloud Nebius deployment this year.
- Federal authorities may conclude the investigation without taking enforcement measures, while Nvidia maintains the transaction advances technological progress.
Shares of Nvidia (NVDA) declined 0.91% following reports that federal antitrust authorities have opened an investigation into the semiconductor giant’s $20 billion licensing deal with artificial intelligence chip developer Groq.
Federal investigators are examining whether the company deliberately crafted the arrangement to avoid mandatory antitrust oversight. The New York Times initially broke the story, with Bloomberg subsequently verifying the confidential probe.
The agreement was publicly disclosed in December. According to its framework, Nvidia secured intellectual property rights to Groq’s technology while Groq continued operating independently. CEO Jonathan Ross and COO Sunny Madra both transitioned to Nvidia roles following the arrangement.
Groq chose not to file the agreement for regulatory antitrust review, attracting government scrutiny. The Justice Department launched its inquiry into the transaction earlier this year as part of an expanded antitrust examination of Nvidia.
Congressional leaders have criticized the arrangement’s framework, characterizing it as a disguised acquisition designed to diminish competitive forces in the AI semiconductor sector. Nvidia has vigorously defended its position.
“The Groq story is a prime example of the American system working as designed to promote innovation, reward entrepreneurs, and benefit consumers,” an Nvidia spokesperson said.
The investigation may conclude without resulting in formal enforcement proceedings.
Groq Rack Enters Mass Production
From a technological standpoint, the Groq 3 LPX rack has entered mass production and will arrive at neocloud Nebius during the coming months. Each liquid-cooled unit contains 256 Groq processors and integrates with Nvidia’s Vera CPUs and Rubin GPUs.
According to Nvidia, the rack achieves 3,400 tokens per second throughput, significantly exceeding the 750 tokens per second OpenAI has indicated for its Cerebras-based Ultrafast mode. Samsung fabricates Groq’s processors, contrasting with Nvidia’s GPUs, which Taiwan Semiconductor Manufacturing produces.
This $20 billion transaction represents Nvidia’s largest corporate deal to date. Bernstein’s Stacy Rasgon informed CNBC that Nvidia possesses sufficient financial resources to complete a transaction of this magnitude without compromising its fiscal position.
Institutional Investors Remain Confident
Hedge fund ownership of Nvidia increased to 285 funds during Q2 2026, compared to 275 in the previous quarter. Competitor AMD attracted 164 hedge fund investors during the identical timeframe.
Nebius remains the sole confirmed cloud infrastructure partner scheduled to implement the Groq rack system. Nvidia simultaneously manages multiple significant investments, including potential commitments reaching $100 billion to OpenAI and $5 billion to Intel.
Federal authorities continue their examination of the Groq licensing transaction.





