Quarterly Outlook
Macro Outlook: The US rate cut cycle has begun
Peter Garnry
Chief Investment Strategist
Chief Investment Strategist
The rise of Nvidia is fascinating as it is not an entirely new company as it was founded in 1993 with an idea of creating 3D graphics cards for gaming and multimedia markets. In 1999, Nvidia announces the GPU, the graphics processing unit, which begins the era of reshaping the computing industry. Very early on, Nvidia sees, and maybe inspired by Steve Jobs full integration thinking, that it is about creating an ecosystem to not only add value to clients, but increase their switching costs. This leads to the development and launch of the CUDA Toolkit in 2006 which opens parallel processing capabilities of GPUs leading to the breakthrough AlexNet neural network in 2012. In 2018, Nvidia reinvents computer graphics with the RTX platform which is the first GPU capable of real-time ray tracing (which is a method of graphics rendering that simulates the physical behaviour of light).
All this innovation laid the seeds for the high quality company Nvidia is today. A quality company creates over time deep moats that are difficult for the competition to overcome and Nvidia’s intensive R&D spending over the years was the marker for this. Nvidia spent on average around 25% or more in the years leading up to the beginning of the fiscal year 2017. That represents R&D intensity that is far above the direct competition and multiples above the average company in this world. It was the high R&D spending that made Nvidia look bad on return on invested capital (ROIC) for many years making it looking less like quality in the short-term. This is an important lesson for investors. High revenue growth and high R&D spending in percentage of revenue are two important markers of future success.Nvidia saw the future earlier than anyone else and was thus well positioned to take advantage of the major shifts in the computing industry towards parallelisation and advancements in machine learning. When you look at the historical revenue trend it is clear that the fiscal year 2017 that ended in January 2017 is the moment when things changed for Nvidia. Up until this point Nvidia had made the majority of its profits from selling GPUs used for gaming which was high growth industry, but not on the scale we see today.
In 2016, two trends began to evolve fast. Bitcoin was exploding higher and with that fat profits for Bitcoin miners that needed more and more GPUs to crunch the numbers to unlock Bitcoin in the validation process embedded in the Bitcoin ecosystem. Simultaneously, the machine learning community saw major advancements to the tune in 2016 that Roger Parloff said the “deep learning revolution” built on Nvidia’s GPUs had transformed the AI industry. While the AI and machine learning community was growing GPU demand fast it was significantly eclipsed by the Bitcoin mining demand for GPUs. When Bitcoin lived its second “ice age” during 2018 it was immediately seen in demand for GPUs and Nvidia’s results for the FY19 (ending January 2019, so covering the 11 months of 2018).
As Nvidia was wounded a new dramatic event happened as the world was catapulted into its biggest pandemic in modern time in early 2020. Initially, demand came down for everything including GPUs, but when the vaccine was published in late 2020, everyone got scared of inflation that was to skyrocket in the 2021. Bitcoin was seen as a potential safe-haven against inflation and the millions of new hobby investors joining financial markets during the lockdowns speculated on everything for a quick buck including Bitcoin. With prices soaring demand for Nvidia’s GPUs took off like nothing it had seen before with revenue in FY22 jumping to $26.9bn up from $16.7bn in FY21. At this point investors were getting Nvidia on the radar.
In late 2022, the next event happened that is today defining the company. OpenAI launched ChatGPT on 30 November 2022 taking the world by a surprise. For the first time there was a useful consumer interface with a large language model (LLM) that could process questions and return useful answers to a wide range of domains. The technology sector immediately saw the potential of this AI technology and investments in LLMs rose by nothing we have seen in a generation lifting Nvidia’s revenue to $60.8bn in FY24 up from $27bn the year before. In 15 years, Nvidia has lifted its revenue from $3.4bn to $60.9bn and is projected to deliver FY25 revenue of $113.8bn and then $143.2bn in FY 26.
When we talk about a quality company the ultimate yardstick is ROIC. After the FY16, the decades of intense R&D spending is paying off with the ROIC quickly jumping from just above 10%, which is fairly average, to the 30-40% range which is superb. In its latest fiscal year, the ROIC rose to staggering 82% far eclipsing the cost of capital for Nvidia. This is the main explanation for the strong share price performance. The high ROIC and low need for capital expenditures (CAPEX) is magic ingredients for strong stock returns. Over the past 15 years Nvidia has only invested $8.2bn in CAPEX, this of course excludes the cumulative R&D spending of $42.5bn, but those investments the company has reached a revenue level above $100bn per year and free cash flows to shareholders of estimated $57bn this year. This is just staggering numbers.
The low CAPEX needs are one of the secret sauces of Nvidia because it reduces how much capital Nvidia must reinvest into the business. Nvidia designs and develops the IP for its advanced AI chips, but it gets TSMC to manufacture its GPUs which effectively means that Nvidia’s has outsourced the capital intensive part of the business to the world’s leading semiconductor foundry business. This is one of the key ingredients of its secret sauce.
Another ingredient is all the software that Nvidia has developed over the years to speed up researching and development of applications using its GPUs with the CUDA Toolkit platform as the crown jewel. This is the deepest point of Nvidia’s moat and the one Sam Altman has openly said is what the industry wants to change by developing joint alternative.
One thing is a high ROIC for generating strong shareholder returns, but sometimes a business can have a high ROIC but being in a low growth industry. An example of this is Walmart. It has consistently a ROIC of around 12% compared to a cost of capital of 7%, but revenue is only growing modestly above the US inflation rate. Returning to Nvidia, its revenue is growing very fast which is reflected in its invested capital which is essentially all the capital Nvidia is deploying to run its business. Nvidia’s invested capital is growing very fast and approached close to $50bn in the recent fiscal year. This is the ultimately engine for high shareholder returns. An expansion of ROIC combined with high or even increasing growth rates.
The ultimate indicator of Nvidia’s success has been its share price performance. Nvidia’s total return has been 35.6% annualised since December 2003 compared to 7.9% for the MSCI World over the same period. The semiconductor industry group has delivered 13% annualised in the same period. The outperformance is probably the greatest over this 20-year period. The past is obviously not an indicator of future performance.
The main question for investors is whether Nvidia will turn out to be an enduring company that can fend off competition for years and become as powerful as Apple and Microsoft. Time will tell and the only thing we know from history is that no company remains at the top forever. Competition or technological change will eventually disrupt everyone.
Jeff Bezos, the founder of Amazon, is quoted for saying that “your margin is my opportunity” which can be interpreted in multiple ways. In retailing it can be mean cutting out one layer in the distribution network. In technology it can also mean that if a critical supplier is earning a fat margin then there is a lot of money to be saved for finding a cheaper alternative. This is exactly what might be brewing in the GPU industry.
It is no secret that many US technology companies are pursuing their own purpose-built AI chips just like Apple developed their own M1 chip. If you can develop the design and IP for the AI chip you can get TSMC to manufacture the AI chip. But as we described in the beginning the AI industry is not only about hardware, but also about the software. Nvidia’s CUDA Toolkit is a development environment for creating high-performance GPU-accelerated applications. Even if you find an alternative AI chip you need specialized software for making your applications.
As the FT writes today competition in the AI development environment is coming for Nvidia’s lunch as Meta, Microsoft, and Google are collaborating on the development of Triton which will be able to run on a large range of AI chips. Whether these US technology companies can build a good enough alternative time will tell, because Nvidia has invested billions over the years in nicely integrated software. So while competition is coming for Nvidia it might be that Nvidia’s moat is too deep to overcome and that Nvidia just is the next new natural “monopoly” such as Microsoft and Apple.
Existing or potential new investors in Nvidia should consider what we regard as the three most obvious risk factors highlighted below.
Quality companies can be defined in many different ways just like value. MSCI, which is world’s largest equity index provider, has defined it using three fundamental variables return on equity, debt to equity, and earnings variability. This definition makes sense because it can be applied to all companies regardless of which sector they are part of. The definition puts emphasis on profitability relative to the deployed equity, leverage ratio (less debt leverage relative to equity is good), and finally the predictability of the business with less variance in earnings being a good thing. In our past equity research we have also found that the lower earnings variability a company has the higher its valuation becomes, so this is a quality marker.
In our equity research note Top quality companies and how to decode their traits we focused on return on invested capital (ROIC) relative to the cost of capital (WACC) as the key measure to identify quality. Next we explained, that around half of those companies with the highest ROIC see their ROIC falling from the top to outside the top over three years due to competition or changing technologies. This is the quality trap that investors need to avoid. It is about finding enduring quality. The “7 Powers” framework is a good approach to analyse whether a company has enduring characteristics or not. Finally, a company can have a stable spread in ROIC minus WACC with a ROIC not in the absolute top and still be a phenomenal stock for shareholders. All it requires is that the business can invest a lot into the business. Historically, Walmart was exactly such a case.
The interesting thing about researching quality companies is that you cannot put it all into a formula. You must apply discretionary thinking about the business, its products, the company’s strategy, the industry drivers, technologies strengthening or weakening the business, because in the end the big returns about changes in expectations for the company.