Breaking down Nvidia’s unusual $20 billion deal with Groq

Investing.com -- Reports emerged on Wednesday that NVIDIA (NASDAQ: NVDA) has agreed to acquire Groq, a designer of high-performance artificial intelligence accelerator chips, in a $20 billion all-cash deal. Groq recently raised $750 million at a valuation of about $6.9 billion.

However, the Nvidia–Groq transaction does not appear to be a traditional acquisition despite CNBC reporting. Groq said it has entered into a "non-exclusive inference technology licensing agreement" with Nvidia aimed at accelerating AI inference at global scale.

Under the terms of the agreement, Nvidia will license Groq’s inference technology, reflecting what Groq described as a shared focus on expanding access to high-performance, low-cost inference. The deal does not grant exclusivity.

As part of the arrangement, Groq founder Jonathan Ross, President Sunny Madra and other members of the Groq team will join Nvidia to help advance and scale the licensed technology. Groq said it will continue to operate as an independent company, with Simon Edwards stepping into the role of chief executive officer.

That structure suggests Nvidia is not acquiring Groq itself, its intellectual property, or exclusive rights to its technology, but is instead paying for licensed use of the technology and the addition of key personnel. Nvidia is yet to comment on the deal.

Shares in the AI chipmaking behemoth are up 0.7% in thin premarket trade Friday.

Groq was founded by former engineers behind Google’s tensor processing unit, or TPU, a chip designed to compete with Nvidia in artificial intelligence workloads.

Wall Street weighs in

Bank of America analyst Vivek Arya said that the deal "implies NVDA recognition that while GPU dominated AI training, the rapid shift towards inference could require more specialized chips."

Arya describes Nvidia’s GPUs as general-purpose platforms while Groq’s LPUs are positioned as "specialized," ASIC-like chips optimized for fast and highly predictable AI inference.

The LPUs rely on large amounts of on-chip SRAM to store model weights and working data, enabling extremely fast per-token access, but with more limited scalability compared with Nvidia’s GPU platforms that use high-bandwidth memory to maximize throughput, the analyst explained.

Arya envisions future Nvidia systems where GPUs and LPUs coexist within the same rack, connected via NVLink.

"Longer-term we think the potential Groq deal could be strategic, similar to NVDA’s Apr’20 Mellanox acquisition that is now the foundation of NVDA’s networking/AI scaling moat," he concluded.

Baird analyst Tristan Gerra said that while he believes that Nvidia’s GPUs "will retain the majority of the AI processor market by 2030, custom ASIC could be accretive to Nvidia’s TAM over time."

Similarly, Bernstein analyst Stacy Rasgon commented that "$20B seems expensive for a licensing deal," especially for a “non-exclusive” agreement. However, he notes that the money invoived "is still pocket change for NVDA given their current $61B cash balance (and massive future free cash flow) and $4.6T market capitalization (it’s about 82 cents per share)."

"We’re inclined to give them the benefit of the doubt," Rasgon wrote.

You May Also Be Interested In

• Gamma Resources raises private placement ceiling to $1.75M
• ONEOK to buy Brazos Midstream Permian assets for $4.43B
• Baird Upgrades CNH Industrial (CNH) to Outperform