Cambyses Financial Advisors LLC

NVIDIA Adds Depth and Breadth to its Artificial Intelligence Offer -- August 2023

 

NVIDIA (NVDA) positioned itself at the center of today's AI world with a new chipset that facilitates throughput for Machine Learning (ML) and Large Language Models (LLM) and the inferential training they require. Last week NVDA introduced the Grace Hopper Superchip, a CPU and GPU on a single platform that promises 7X the speed of the Intel et. al. PCIe 5.0 bus architecture, and 30X the bandwidth of NVDA's own DGX A100. 

 

Throughput and bandwidth identify-as the primary constraints on the Machine Learning and Large Language Models that are taking AI by storm this year. With 35-40% of companies already implementing AI in some form, much of it centered on massive databases and almost equally massive regression/optimization algorithms,[*1] data center overload threatens to become acute. NVDA was already near the top in the race to provide throughput. The Grace Hopper Superchip cements that industry leadership position - at least for the moment. 

 

NVDA is already the largest chip maker. Its $1.1 trillion market capitalization is roughly three times the market capital of its nearest rivals (Taiwan Semiconductor and AVGO-Broadcom) and eight times the size of industry household names, QualCom and Intel. Once seen as primarily a computer game facilitator and video editing solution, NVIDIA has become the premiere supplier of advanced GPUs and CPUs. The public and businesses IA fixation, centered on ML and LLM, make GPUs, potentially massive, parallelism and scalability, the technology of the hour - and NVDA its supplier. That combination has driven a nearly 220% appreciation in NVDA's stock price so far this year. Combine that with extraordinarily good fundamentals (Current ratio 3.4, Debt/Equity 0.4, Gross margin 56%) and it almost takes the terror out of NVDA's 228+ P/E. Indeed, 85% of analysts that we follow rate NVDA as a Buy or Strong Buy, and project stock appreciation averaging 16% for the coming year.[*2]

 

With all this going for it, we are still left with one burning question: Is the Grace Hopper Superchip actually smarter than its namesake? We find that difficult to imagine! If you haven't encountered legendary computer and programming pioneer Grace Hopper before, maybe it's time to find her Wiki? [*3]

 

 

[*1]  ChatGPT reportedly includes a 580 GB dataset and a 175 million parameter search/response algorithm. Approximately 10 million users access the software daily.

[*2] For comparison sake, CFA rates NVDA's risk as moderate or elevated (Four or Five on a 6-point scale), and 1 year share value growth as elevated (11-15%). We rank its performance (Return/Risk/Prospects) 62nd of the 246 securities we rank for our client's portfolios. Detrimental factors include high volatility (Beta = 1.8, standard deviation exceeding that of 99% of S&P securities), that extraordinary P/E (exposing NVDA to market valuation risk), and geopolitical factors that affect NVDA's sales in China.

[*3] Rear Admiral Grace Hopper is credited with, among many other things, initial concepts in linker-compiler architecture, contributions to higher level language programing (particularly COBOL), her advocacy of distributed networks, and one of the earliest mentions of a "computer bug" in the literature. Congress and Ronald Reagan extended Hopper's service well beyond the Navy's mandatory retirement age in order to preserve her contributions to the field. Hopper retired from the Navy in 1986 after 43 years in service. She continued to contribute to the field until her death in 1992 at age 85.