Is NVIDIA Stock Still a Good Way to Invest in AI Chips?
We recently wrote about “How AI Chips Are Changing the Semiconductor Industry,” noting that AI chips come in all shapes and sizes. From smartphone AI chips that perform computations at the edge to larger chips that are used in data centers, the trend is moving toward increased specialization. Large lots of general-purpose AI chips are taking a back seat to small batches of chips that are highly customized. As long-time investors in NVIDIA (NVDA), we’re curious about how they’re reacting to these changes.
For quite a while NVIDIA was the only pure-play AI stock for retail investors who wanted exposure to the growth of artificial intelligence. They were a market-leading producer of graphical processing units (GPUs) that were used ubiquitously by anyone training AI algorithms. Now, we’re seeing specialized AI chips like ASICs being used for training as well. In order to understand how this affects NVIDIA, it’s helpful to know where their revenues are coming from.
What Does NVIDIA Actually Do?
Trying to figure out exactly what NVIDIA does is becoming increasingly difficult, depending on how you look at it. First, there are two reportable segments you’ll find in their financial filings as “GPU” and “Tegra Processor” which commanded 87% and 13% of revenues in 2019, respectively. The company defines these segments as follows:
- GPU product brands – GeForce for gamers; Quadro for designers; Tesla and DGX for AI data scientists and big data researchers; and GRID for cloud-based visual computing users.
- Tegra brand – integrates an entire computer onto a single chip, and incorporates GPUs and multi-core CPUs to drive supercomputing for autonomous robots, drones, and cars, as well as for game consoles and mobile gaming and entertainment devices.
Then there are “market platforms,” a mix which continues to evolve over time and consists of gaming, datacenter, professional visualization, automotive, and OEM /IP (where that temporary crypto boost was recorded.)
Gaming continues to be a huge part of what NVIDIA does with Dell being their largest customer at 11% of revenues (Dell sells a popular line of high-end PCs for gaming under the brand name Alienware.) Market share is a key driver for profitability, something that Harvard Business Review wrote about decades ago. NVIDIA happens to be #1 in PC gaming with more than 3X the revenue of the other major GPU vendor, AMD. NVIDIA doesn’t disclose profitability for their business segments, but gaming seems to be a cash cow business that helps fund the other market platforms that may be less profitable as they require loads of investment to capture market share.
While esports have made gaming a promising investment thesis, the company is over-reliant on this segment for more than half their revenues. With advancements being made in areas such as ray tracing (simulating the behavior of light), NVIDIA is making sure they maintain their dominant market share in gaming while continuing to advance other areas of the business.
Then there are the industry verticals NVIDIA is focused on addressing, like healthcare, which they recently devoted an entire presentation to at JP Morgan Healthcare 2020. In that deck, they talked about how the CUDA-X platform released last year will address other industry verticals like factory automation, robotics, and even 5G.
The term “full stack” means that they don’t just build the hardware, they also build the software and everything else needed to deploy solutions in these areas such as processing, networking, and storage. Just this past week, NVIDIA closed on their acquisition of Mellanox which broadens their stack in the area of datacenter networking.
NVIDIA’s Datacenter Segment
The last time we checked in with NVIDIA was back in November of 2018 in a piece titled “NVIDIA Stock Price Hits a 52-Week Low – Buy More?” In that piece, we talked about how the OEM and IP segment was affected by the collapse of the cryptocurrency bubble and a backlog of graphics cards affected gaming revenues. Both of these represented only temporary setbacks, and the “datacenter” segment appeared to be where the opportunity for AI would manifest itself.
The growth of Datacenter revenues is a simple metric we can use to see if the “picks and shovels” of artificial intelligence are growing as we expect them to grow over time.Credit: Nanalyze
This past week, NVIDIA issued a press release titled “NVIDIA Completes Acquisition of Mellanox, Creating Major Force Driving Next-Gen Data Centers.” At a purchase price of around $7 billion, that was the largest acquisition in the history of NVIDIA.
In looking at the latest investor deck for Mellanox, we see a company that was enjoying some tremendous growth around selling both hardware and software used in datacenters. Machine learning algorithms are only as good as the data you give them, and they can only work as fast as you can stream the data. Mellanox builds the high-speed interconnectivity hardware – ethernet adapters and switches – and software that is tightly coupled with NVIDIA’s existing platform. According to the company’s fearless leader:
AI is now driving an architecture change from Hyper-Converged Infrastructures to Accelerated-Disaggregated Infrastructure.
We have no idea what any of that means but all you CTOs out there can feel free to plaster this on slide decks and everyone will just nod their heads and pretend like they know what you’re talking about. Said more simply, high-speed connectivity is a key component in tomorrow’s datacenters.
Just to give you an idea of what sort of horsepower we’re talking about here, last year the largest cloud-based simulation in history took place with 50,000 GPUs spread out across the globe from all the leading cloud computing providers. The simulation involved interpreting messages from the universe and took several hours to run using around 3 terabytes of input data. Amazon wrote about the experiment stating that the “cloud-based cluster provided almost 95% of the performance of Summit.” In the world of high-performance computing, Summit is literally the fastest computer on this planet. Speaking of which, NVIDIA and Mellanox combined have their technology in more than half of the world’s TOP500 supercomputers.
While the datacenter segment becomes increasingly important to NVIDIA following this acquisition, the investment thesis becomes much bigger when you consider how Moore’s Law is being resurrected by accelerated computing.
Accelerated Computing Platforms
In the latest investor deck, NVIDIA refers to itself as “the leading accelerated computing platform” which “is broadly recognized as the way to advance computing as Moore’s law ends and AI lifts off.” AI chips are now cannibalizing the entire semiconductor business with software replacing hardware as the key driver for continuing Moore’s Law. The use of AI has become ubiquitous, and so are AI chips, which are now being developed in various form factors – from smaller edge-computing chips found in smartphones to large powerful chips found in supercomputers.
NVIDIA’s strategy around continuing to drive Moore’s Law using their “accelerated computing platform” is clearly where we want to be as investors if AI chips will cannibalize the traditional semiconductor industry offerings. If you’re someone who thinks that the stock market will test new lows as the real impact of the WooHoo Flu becomes apparent, tech stocks will sink even lower given their inherent volatility. NVIDIA announces earnings later this month so we’ll see what sort of color they provide around future expectations. If you’re looking to invest in NVIDIA on dips, always use dollar cost averaging.
NVIDIA and Microsoft
One last thing that’s interesting to note here. In digging around the filings we saw that twenty years ago NVIDIA did some technology transfer with Microsoft relating to the Xbox. (If you bought shares of NVIDIA at that time they would have been $3 a pop giving you a return of +9,300% if you sold them today.) At that time, NVIDIA would have had a market capitalization of around $1.8 billion (they’re a $173 billion company today, almost exactly the same size as Cisco Systems) and Microsoft would have had loads of leverage for negotiations. Here’s a stipulation that was made at that time.
Under the agreement, if an individual or corporation makes an offer to purchase shares equal to or greater than 30% of the outstanding shares of our common stock, Microsoft may have first and last rights of refusal to purchase the stock.Credit: Latest NVIDIA 10-K
If anyone ever moves to acquire a controlling stake in NVIDIA, it might be Microsoft who the company ends up being sold to.
As investors, we want to make sure NVIDIA isn’t being displaced by the many companies – both large and small – that are now developing their own AI chips. As investors, we’ve always been somewhat concerned that the company is overly reliant on gaming for just over half their revenues, and it’s good to see progress is being made to grow their other verticals through acquisitions making gaming less of a majority contributor to revenues.
Companies that command large market shares enjoy profitability which generates plenty of cash to fuel continued growth. With $10 billion in cash on hand, NVIDIA is in a good position to keep making acquisitions and growing other parts of their business. We’ll keep checking in every year to see how their business segments evolve as NVIDIA continues to benefit from the growth of artificial intelligence.
Despite what the pundits say, FAANG stocks (Facebook, Apple, Amazon, Netflix, Google) don't give you real exposure to AI. Read about 7 stocks that give you true pure-play exposure to AI in our guide to investing in AI healthcare companies, freely available to Nanalyze Premium subscribers.