Nvidia Analyst Says New, Ampere-Based Data Center GPU Makes Chipmaker 'Unassailable'

NVIDIA Corporation NVDA shares, which have been bouncing along nicely for much of this week amid positive sell-side expectations for the chipmaker's earnings report next week, rose over 3% Thursday. CEO Jensen Huang hosted the annual GTC keynote address virtually, revealing details of a slew of products.

Here's how sell-side reacted to the announcements. 

The Nvidia Analysts

SunTrust Robinson Humphrey analyst William Stein maintained a Buy rating and $327 price target.

Raymond James analyst Chris Caso reiterated an Outperform rating and increased the price target from $300 to $330.

Needham analyst Rajvindra Gill maintained a Buy rating and a Street-high price target of $360.

Rosenblatt Securities analyst Hans Mosesmann maintained a Buy rating and $340 price target.

Nvidia A Unique Growth Asset In Semis, SunTrust Says

The GTC Keynote played in line with expectations, Stein said in a note.

The significant product announcements include the A100 data center GPU; the DGX AI computer, which integrated eight A100 GPUs and delivers five PetaFLOPS; the second-gen EGX AI computer targeting edge AI applications; and others such as CUDA 11 Jarvis for conversational AI, Merlin and native GPU acceleration for Spark 3.0.

The lack of a gaming GPU announcement is a surprise, the analyst said.

An Ampere-based gaming GPU will ship in the fourth quarter, with an announcement likely to be made midyear, he said. 

Previewing earnings, Stein said he expects first-quarter results to show meaningful upside in data center, with a similar trend likely to be reflected in the second-quarter guidance.

Second-half demand trends for data center are less robust, the analyst said. 

"Our constructive long-term thesis on NVDA remains: that the company is a unique growth asset in semis that should provide strong investment returns in the long-run."  

See Also: Analysts Dissect Nvidia's Product Announcements: 'Evolutionary Rather Than Game-Changing'

Ampere's Performance Makes Nvidia's Dominance Unassailable Near-Term, Raymond James Says

Nvidia targeting Ampere for both the training and inference markets is quite surprising, especially as the previous generation Volta architecture was used mainly in training, Caso said in a note. 

The 20 times performance improvement for Ampere makes Nvidia's dominance unassailable in the near-term, the analyst said. He sees the higher ASP boosting Nvidia's top-line growth.

With the Ampere GPU capable of running seven virtualized inferencing threads, making it appropriate for inferencing market as well, Caso sees a strong potential growth avenue. 

When Nvidia reports next week, the data center and Mellanox acquisition will serve as catalysts and anew gaming GPU launch in the October quarter will also likely serve as a catalyst, the analyst said. 

KeyBanc Says Ampere Products Have Compelling Specs, Performance 

Nvidia's GTC was focused on data center and HPC applications, with no financial or consumer product updates provided, KeyBanc's Weston Twigg said in a note. 

The analyst sees the launch of the new 7nm Ampere architecture and the A100 data center chip based on Ampere as the major announcement. Nvidia also announced related products and applications, with Jarvis being the standout, as well as advancement in real-time ray tracing, he said. 

"The specs and performance of Ampere products are extremely compelling, which should provide NVDA with ongoing traction in datacenter applications, in our view."

KeyBanc expects data center demand to remain relatively strong through 2020 and for consumer/gaming GPU to experience a slower-than-expected recovery in the second half due to weakening macroeconomic conditions.

Needham Remains Constructive On Nvidia's Long-Term Prospects 

Needham singled out the announcement of the 7nm A100 GPU as the event's highlight. Gill noted the A100 data center GPU based on the Ampere architecture is Nvidia's first 7nm node, making it competitive with rival Advanced Micro Devices, Inc. AMD. 

"We believe the A100, which is able to boost performance by up to 20x over its predecessors, represents NVDA's biggest generational performance leap of its eight generations of GPUs," the analyst said. 

With the DGX A100 system featuring eight A100 GPUs, Gill said he expects the A100 to soon debut in the cloud.

"We remain constructive on NVDA L-T prospects and expect NVDA to continue benefiting from ongoing strength in data center, increased use of its chips to combat COVID-19 and other diseases, and strong demand for Gaming GPUs."

Rosenblatt's 5 Key Takeaways On Nvidia 

Mosesmann listed the following as the takeaways from the GTC Keynote:

• The momentum of accelerated computing is not slowing, with GPUs displacing CPUs in data center hierarchy.
• Nvidia is spearheading the idea that the data center, not the server, is the compute unit. 
• The Mellanox acquisition is helping Nvidia to control the direction of how data moves across the data center.
• The DGX A100s are using AMD's EPYC2 64-core CPUs.
• The lack of announcement concerning consumer version of Ampere was a slight disappointment.

NVDA Price Action

At last check, Nvidia shares were 1.9% higher at $327.31. 

Photo courtesy of Nvidia.