Nvidia's prolific rise to prominence in the past year has been mesmerizing. Its name has become a fixture in dinner conversations, and its price movements are subject to immense daily media scrutiny. The graphical processing unit (GPU) manufacturer has seen its market cap balloon from $350 billion to nearly $2 trillion. This makes it the third largest publicly traded American company by market capitalization, only trailing Apple and Microsoft. Its expansion has seen it comprise 4.6% of the S&P 500, with its daily returns regularly buoying the index to new all-time highs.
In 2023, Nvidia reported $60.9 billion of total revenue, giving the company a 32x price-to-sales multiple. Google / Alphabet, on the other hand, trades at a comparatively modest 5.5x P/S ratio. The dislocation between these backward-looking ratios and both aforementioned companies' similar market capitalizations should provide an idea of where the market is placing its optimism. The simplest narrative is that Nvidia's fiscal future looks bright because of AI's unbounded potential, large language model improvements, and computational expansion. Their GPUs and their associated APIs meet the need for all three verticals, and they are well-positioned to capture value from advancements in the space.
Qualitatively, Nvidia's 90-degree ascent feels justified. Let's take a closer look at the numbers to see if they agree.
Attractive Gross Margins
A company's gross margin measures its revenues less its cost of goods sold (COGS) as a percentage of its total revenues. Costs of goods sold references direct costs associated with manufacturing goods, like labor and materials.
Not all businesses are created equal, and certain industries[1] tend to have either lower or higher gross margins by virtue of the products that they sell. For example, automotive manufacturers usually have anemic gross margins, measuring in the high single digits to low double-digits, due to their high labor and materials costs. On the other hand, software companies receive larger earnings multiples because of their wider gross margins, often exceeding 60% or even 70%. This is because the marginal cost of producing an additional unit of software is incredibly low.
Nvidia saw its gross margins grow from 56.9% in 2022 to 72.7% in 2023. This is a marked expansion. Established, mega-cap companies don't magically save 15 cents on the dollar in just 12 months. Nvidia's explanation for their larger gross margins was “primarily driven by Data Center revenue growth and lower net inventory provisions as a percentage of revenue”.
Datageddon
AI is computationally expensive, and as it becomes the industry standard, the de facto “electricity” that powers every business, we will see a proliferation of new data centers to meet that need. Meta is investing upwards of 1 billion euros into a 30 hectare data center in Toledo, Spain. Google has been buying up tracts of land in Oklahoma and the United Kingdom for new data center complexes, and Microsoft is looking to cut the ribbons on new facilities in Wisconsin and Finland.
Nvidia categorizes its revenues in five different reportable segments:
Data Center,
Gaming,
Professional Visualization,
Automotive, and
OEM / Other.
The growth narrative becomes immediately apparent looking at the difference in Data Center revenues between 2023 and 2022. Nvidia notched $47.5 billion in Data Center revenue in 2023, more than tripling the $15 billion that it achieved in the same segment 2022.
It's evident that Nvidia rose to satisfy the rising computational needs of AI. However, these figures alone only explain one portion of the company's appreciation.
No Sprockets, Just COGS
Nvidia doesn't disaggregate its COGS between revenue segments in its annual reports. It's implied that the COGS associated with each segment differ, as given in their explanation for their larger gross margins.
We can estimate the gross margins for each individual revenue segment using a linear regression of revenue segments against the total cost of revenue. There are some assumptions that we will be making with this model:
Each revenue segment's COGS is independent of others,
Each revenue segment's COGS doesn't change over time,
All regression coefficients should be positive, and
The model will be fit without a y-intercept.
A linear regression over N=11 datapoints (annual revenue segment figures from FY 2014-2024) yields an R^2 of 97.4%. The coefficients that the regression yields correspond to the COGS for each respective revenue segment, providing the following implied gross margins (1 - COGS):
Data Center: 79.3%
Gaming: 44.0%
Professional Visualization: 9.5%
Automotive: 100.0%
OEM / Other: 100.0%
Note that for the final two reportable revenue segments, the model provided a coefficient of 0 (and therefore a 100% gross margin). While this is practically impossible and factually incorrect, this is because both the Automotive and OEM / Other segments have reported relatively flat revenues over the past 10 years, and their proportion of total revenues has shrunk from 39.5% to 2.2%. Their implied costs have become a mere footnote (pun unintended) in Nvidia's financials, and for the purposes of our exercise, we will ignore them here.
The handsome, software-like, implied gross margins for the Data Center complete the missing piece of the Nvidia growth narrative. The company grew, and has been growing revenues exponentially while scaling down costs. Furthermore, as per their provided guidance for the upcoming quarter, they expect these trends to continue.
Sunny Skies Ahead
In their most recent earnings press release[2], Nvidia provided the following outlook for the first quarter of their 2025 fiscal year:
$24 billion in total revenue ±2%, and
a 76.3% overall gross margin ±0.5%.
Since the future of Nvidia is in its data center segment, we can aggregate the remaining four revenue segments and run a new regression with two features: data center revenue and ex-data center revenue versus COGS.
This regression, with an R^2 of 97.0%, yields different gross margins:
Data Center = 77.9%
Ex-Data Center = 50.0%
In any case, a 1.4% difference in gross margins will not meaningfully affect our forecast. We will use the new gross margin figure from our second regression analysis to forecast the next quarter's data center revenues. Solving the following linear equation will give us our projection for the segment's revenues.
Let
D = Data Center revenues,
X = Ex-Data Center revenues
Nvidia's upcoming quarter gross margins = 76.3% (mean of their guidance).
Then (in millions),
Therefore, by our estimation, we expect Nvidia to garner over $22.6B in data center revenue. This would represent
94% of all revenues in the upcoming quarter,
a 23% Q/Q growth in data center revenues, and
a 428% Y/Y growth in data center revenues.
Peer Pressure
It’s a bull market, you know.
While we analyzed revenues, COGS, and gross margins in this article, it's ultimately net income and free cash flow which matters to analysts and a company's intrinsic valuation. Net margins are much harder to forecast because they incorporate more variable line items, including operating expenses like research & development, interest expenses, income taxes, and more. Forecasting these will be a topic for a future article.
Nvidia's net margin tripled from 16.2% in FY 2023 to 48.9% in FY 2024, again, in large part because of their expanding gross margin attributable to huge Data Center revenue growth. The company reported earnings per share (EPS) of 11.93, which at 776.63 gives Nvidia a P/E of 65. However, like our P/S comparison above, this is a trailing calculation; the stock market is forward-looking.
The S&P 500's estimated forward P/E ratio[3] stands at 23.4. If Nvidia were to keep in line with the mean ratio of the index, it would need to report an EPS of 33.2 for FY 2025, nearly triple that of its last twelve months. While this may seem like an insurmountable feat, Nvidia sextupled their EPS from 1.74 in FY 2023 to 11.93 in 2024. “Disappointment” is not a word in Jensen Huang's dictionary.
Disclaimer
This article does not constitute a recommendation for buying or selling a security. I, Rohit Sarathy, do not hold positions in any of the securities mentioned in this article.