By Michael Elkins
UBS reiterated a Buy rating on Nvidia Corp (NASDAQ:NVDA) and raised their price target on the stock to $315.00 (From $270.00) following Nvidia GTC (GPU Technology Conference).
The conference hosted hundreds of panels and presentations ranging from general knowledge to very technical. UBS analysts listened to some they considered to be most relevant and, overall, confirmed UBS’ view that “NVDA is quite literally the nexus of AI.”
NVDA created necessary corridors of cloud capacity over the past few Qs, announced ~$2.2B of pre-payments to various CSPs, and announced its new DGX Cloud offering last earnings call. This appears to be quickly gaining traction with customers. During its DGX Cloud customer panel, NVDA said “phones have been ringing off the hook” since the product was announced, with many customers in the queue to gain early access to the platform.
There were also several sessions dedicated to advances in training and inference of large language models (LLMs), with a significant amount of focus being placed on efforts to improve performance, efficiency, and accuracy. Among the various avenues that are being explored to improve training and inference efficiency is quantization with lower-precision number formats. Specifically, there were several sessions that discussed the benefits of utilizing FP8 (8-bit floating point) for training and inference – an approach that was not feasible prior to NVDA’s native support of FP8. This is a significant advantage for NVDA relative to AMD, considering the shift in the industry towards FP8 training and inference.
Generative AI also serves as a pivotal tool for the development of L3/L4+ autonomous vehicles.
The analysts wrote in a note, “The power behind NVDA’s DRIVE Sim comes from the use of generative AI and Omniverse, allowing it to train models using real-world data captured from sensors and convert it into synthetic 3D models (digital twins) for testing and validation. Additionally, NVDA’s new DRIVE Thor chip integrates AV functionality into a single SoC for digital clusters, infotainment, and driver assistance for more efficient development and faster software iterations.”
“We see NVDA’s ability to provide AI capabilities and an expansive software stack alongside its hardware as a key differentiated solution with the ability to train AI in the data center to deliver high-performance centralized in-care computing.”
On the heels of the GTC, UBS increased Data Center revenues for their model update. In terms of the Data Center segment revenues, the analysts continue to take a variant view where UBS remains below the Street on FY23 and well above the Street on FY24. Overall GM moves up due to a higher mix of Data Center revenues vs. the prior model. UBS models 1Q24 estimated revenues and EPS of $6.48B/$0.95 and raised FY23/24 estimated revenues and EPS to $29.8B/$4.70 and $42.4B/$7.24 (from $29.11B/$4.54 and $39.75B/$6.53).
Shares of NVDA are up 0.51% in premarket trading on Monday.