Wall Street analysts said that Nvidia’s (NVDA) recent deal with Groq offers both “offense and defense” and is strategic.
On Wednesday, Groq said it entered a non-exclusive licensing agreement with Nvidia for its inference technology. Under the agreement, Groq’s Founder Jonathan Ross, President Sunny Madra and other members of the Groq team will join Nvidia to help advance and scale the licensed technology. Reportedly, Nvidia is acquiring assets from Groq for $20B.
Cantor said Nvidia remains a top pick and reiterated its Overweight rating and $300 price target on the stock.
“On Christmas Eve, Nvidia announced the acquisition of Groq IP and talent for $20B (think “acqui-hire”). At a high level, we view this acquisition as offering both offense and defense. On the offense side, we know Nvidia has been working with Groq for specific inference acceleration. We suspect Nvidia saw real opportunities and decided it would be better for Groq to be inside Nvidia vs. external partners,” said analysts led by C.J. Muse.
The analysts also highlighted strong talent acquisition (including Groq’s CEO who was formerly a key developer of Google’s (GOOG) (GOOGL) tensor processing unit, or TPU).
For defense, the analysts said that Nvidia dominates AI training and time-based inference, and with Groq’s low-latency, energy-efficient inference, this offering inside Nvidia’s full system stack solution (and likely eventually CUDA compatible) will enable a greater share of the inference market, particularly for the next leg of the AI infrastructure build out — real time workloads like robotics and autonomy.
“Put together, we think this acquisition only enhances Nvidia’s full system stack and overall leadership in the AI market (and only widens its competitive moat),” said Muse and his team.
BofA Securities kept its Buy rating and $275 price target on Nvidia’s stock.
Analysts led by Vivek Arya said the Groq deal is surprising and expensive but strategic.
The analysts said key takeaways are — It was a surprise news, right before Christmas, and involves a different kind of hardware called language processing unit, or LPU versus Nvidia’s well-regarded graphics processing unit, or GPU, expertise/focus; and implies Nvidia’s recognition that while GPU dominated Al training, the rapid shift towards inference could require more specialized chips.
Another takeaway is that different hardware adds complication around future GPU/LPU roadmap and pricing, but would enable Nvidia to use its balance sheet and platform incumbency to add more customer choice and conceptually address competitive threats from Groq and other specialized application specific integrated circuits, or ASIC, chips.
The analysts noted that some questions remain, but potential deal is positive in the long term.
“Longer-term we think the potential Grog deal could be strategic, similar to NVDA’s Apr’20 Mellanox acquisition that is now the foundation of NVDA’s networking/Al scaling moat,” said Arya and his team.