Wedbush analyst Matt Bryson reiterated Nvidia Corp NVDA with an Outperform and a $1,000 price target.
Nvidia has unveiled four new discrete hardware offerings as part of its Blackwell launch, all based on its latest chip technology but differentiated by packaging and systems design.
The lineup includes the B100, a direct replacement for the H100 architecture with a consistent TDP, available as an HGX offering.
Also Read: Nvidia ‘1-2 Steps Ahead of Competitors’ After Blackwell GPU Reveal At GTC 2024, Analyst Says
The B200 features two dies linked via a 10 TB/s Nvidia High Bandwidth Interface, functioning as a single chip and available in an 8 GPU HGX variant.
The GB200 combines two B200s (four dies) with a Grace processor, while the GB200 NVL72 is a liquid-cooled rack-scale system comprising 36 GB200 Superchips designed to act as a single GPU. These systems can be interconnected through Nvidia’s new X800 switches to form large clusters with minimal transport latency.
Bryson’s key takeaways include the B100 offering an easy upgrade path that could drive significant demand for Nvidia’s next-generation parts. He said the packaging and systems architectures of Nvidia’s new offerings promise substantial performance gains.
The analyst noted that the B100 doubles training performance compared to the Hopper architecture due to Nvidia’s FP4 format. Bryson stated that the B200 quadruples training performance is further enhanced when packaged with Nvidia’s Grace processor in the GB200 and GB200 NVL72 systems.
He added that the latter delivers an impressive 720 petaflops of training and 1.4 exaflops of inference power.
Nvidia’s advancements affirm its leadership in the AI accelerator market, bolstered by its system-level architecture innovations that competitors may find challenging to replicate, per the analyst.
Nvidia’s position is further strengthened by its IP/competencies, partly from the Mellanox acquisition, and its ability to set AI standards due to its scale, Bryson said. Major cloud CSPs, including Microsoft Corp MSFT Azure, Amazon.Com Inc AMZN AWS, Alphabet Inc GOOG GOOGL Google Cloud, and Oracle Corp ORCL Oracle Cloud, have committed to integrating the GB200 NVL72, signaling widespread industry support.
The introduction of the B100 and B200, each with 192 GB of HBM memory—up from the H100’s 80 GB—indicates a positive outlook for DRAM vendors due to the premium pricing and margins for HBM, as well as for the broader memory complex, given the resources allocated to meet HBM requirements, Bryson noted.
The price target is based on a target PE multiple of ~36X applied to Bryson’s fiscal 2026 estimates, plus net cash of $6.54 per share, somewhat above the 2-year average (31X) or the 5-year average (30X). The analyst noted that the higher valuation is supported by strong current fundamentals that are again likely to yield more robust growth than is currently being modeled for Nvidia.
Bryson projects first-quarter revenue and EPS of $24.01 billion and $5.41.
Rosenblatt analyst Hans Mosesmann flagged how Nvidia CEO Jensen Huang heralded the rise of generative AI as a burgeoning industry, marking a shift towards accelerated hardware and software (HW/SW) computing as the leading approach. Huang suggested that the era of prioritizing cost reduction in general-purpose computing is giving way to an era focused on the need for significantly expanded scale through accelerated computing. The complexity of Large Language Models (LLMs), in terms of parameters and tokens, is rapidly increasing, with a doubling rate every 4-6 months.
The launch of the new Blackwell GPU platform, set for later in 2024, is the most ambitious endeavor Silicon Valley has seen, as per Mosesmann.
This anticipation is so high that Blackwell will likely be sold out well into 2025 despite improvements in the supply of its predecessor, Hopper. Even with Hopper’s adoption of the 4nm Taiwan Semiconductor Manufacturing Co TSM node, the Blackwell platform achieves a 2.5 times increase in training capability and a fivefold improvement in inference performance over its predecessor.
Nvidia stock has gained over 242% last year. Investors can gain exposure to the AI beneficiary via Invesco Nasdaq Free Cash Flow Achievers ETF QOWZ and Pacer Funds Pacer Data And Digital Revolution ETF TRFK.
Price Action: NVDA shares traded higher by 1.20% at $895.13 on last check Tuesday.
Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.
Photo via Shutterstock
© 2024 Benzinga.com. Benzinga does not provide investment advice. All rights reserved.
date | ticker | name | Price Target | Upside/Downside | Recommendation | Firm |
|---|
Trade confidently with insights and alerts from analyst ratings, free reports and breaking news that affects the stocks you care about.
