David Rolfe Comments on Nvidia

Guru stock highlight

Author's Avatar
Oct 11, 2019
Article's Main Image

NVIDIA (NVDA, Financial) is a pioneer in the development of the graphics processing unit (GPU) – a semiconductor traditionally utilized for rendering computer graphics – and has extended the GPU beyond the graphics domain into “general purpose computing.” We attribute NVIDIA's success in general purpose computing to their proprietary computing platform and programming model, known as CUDA.

NVIDIA’s compute acceleration platform forms the backbone of a unique value proposition for steadily emerging compute-intensive applications, such as image processing, natural language processing, assisted driving, and ray tracing (the latter relates to the video game domain). The central processing unit (CPU) has been the workhorse of general-purpose computing for decades, as reliable, almost annual efficiency gains helped drive the development of increasingly complex computing applications. As those CPU efficiency gains have slowed over the past several years, developers have begun utilizing GPUs to accelerate applications. While a CPU usually has between a couple and a few dozen cores that are very fast at computation, that contrasts with a CUDA-based NVIDIA GPU that breaks a computation down across hundreds or even thousands of cores and completes it in a fraction of the time. Yet similar to CPUs, and much like Intel’s x86 standard, virtually any industry application can utilize NVIDIA’s GPUs to accelerate performance, thanks to CUDA’s programmability and rich library of software that has been developed for more than a decade.

We expect that the total addressable market served by NVIDIA’s acceleration platform can double over the next five years as data science is increasingly applied in the enterprise, similar to how it has been applied at hyperscale (e.g., Facebook, Google) and scientific computing domains, where NVIDIA has over 90% market share, we estimate. NVIDIA maintains a very high market share in PC and cloud gaming, scientific computing, and hyperscale domains as there are no other GPU-based general compute platforms with software and standardized architectures to rival NVIDIA. Of course, there will be plenty of competitive attempts from alternative silicon providers to accelerate specific computing workloads, namely field-programmable gate arrays (FPGA) as well as other application specific integrated circuits (ASICs); however, we think these offerings lack the standardization and compatibility inherent to NVIDIA’s platform.

The stock has pulled back nearly -40% from its 2018 highs, after growth decelerated from unsustainable levels – driven by a small number of hyperscale operators building inventory. Further, a not insignificant amount of NVIDIA’s revenue over the past few years was generated by cryptocurrency mining applications. As capital fled from cryptocurrency applications, NVIDIA’s revenue from crypto-mining has approached zero and should not be much of a risk going forward. Overall, we think NVDA’s business has bottomed and should be able to sustain faster growth over the next few years. As such, NVIDIA’s fiscal 2021 price-to-earnings multiple is 24X and trading well below its 5-year average of 29X.

From David Rolfe (Trades, Portfolio)'s third-quarter 2019 Wedgewood Partners letter.