Nvidia: Record Highs, Yet History Of Railroads Hints At Bleak Prospects

Summary:

• Nvidia Corporation’s stock price is near all-time highs, but I remain bearish due to its role as an infrastructure company in the AI boom.
• Historical parallels with railroads suggest infrastructure companies face long-term stagnation after initial growth, questioning Nvidia’s revenue prospects.
• The market’s uncertainty about AI capacity requirements may inflate NVDA’s valuation, but demand may eventually plateau, leading to a potential stock price decline.
• Investors should be cautious with NVDA, considering its high market cap and potential downside risks, especially for those with significant exposure.

Nvidia Corporation’s (NASDAQ:NVDA) stock price is near all-time highs again. As shown in below chart, since my three previous bearish articles, NVDA’s stock price has chugged higher (April 22, 2024, Nvidia: Significant Downside Risks As Its Customers Develop AI Chips, June 25, 2024, Nvidia: History Of CPUs Says GPUs Are Not Worth $3 Trillion August 7, 2024, Nvidia: Stock Price Likely Peaked Even If Fed Cuts Rates).

Yet, I remain steadfast bearish for another key reason (not discussed in the previous three articles): NVDA is essentially an infrastructure company in the AI boom. No matter how hot or revolutionary infrastructure seems in the heat of the moment, infrastructure companies have historically fared poorly once the initial novelty wears off and capacity reaches saturation.

In this article, we will use railroads as an example. Why railroads? Railroads were a new, hot thing in the 1800s and a good example of a revolutionary technology that had a substantial economic impact but with a cautionary tale for investors. Railroads were a key invention for industrialization and one study estimates economic productivity would have been 25% lower in 1890 without railroads. By contrast, one estimate is that AI can potentially contribute 7% to global GDP, which is of a similar magnitude with railroads, though the eventual economic contribution of AI in the future could vary. There are several parallels that can be drawn with NVDA and implications on the long-run trajectory of NVDA’s financials and stock price.

Construction of Infrastructure is a non-recurring business

The “golden age” of railroad expansion in the US was from 1865 to 1900, where total track mileage increased from around 30,000 miles to 200,000 miles, as seen in the orange line below, which is a CAGR of 5%. The annual new track laid (shown in the blue line) varied greatly with financial and economic conditions. During the railroad boom, it may have seemed like the demand for railroad track was limitless. But it was not – the annual new track laid peaked in 1887 at 13,048 miles and sharply declined afterward.

Relevance to NVDA: NVDA provides chips for AI purposes, which are essentially infrastructure for the application of AI. This role is similar to a supplier of railroad track to railroad companies. Railroad track (and other parts) suppliers would have seen large growth during the early years, but after the cumulative track mileage reached a certain level, new construction dropped off a cliff. Analyst projections of NVDA are not only for strong growth in revenues and profits in the upcoming years. They also project that NVDA will sustain revenues of hundreds of billions of dollars each year in perpetuity while maintaining strong growth in the next few years. This seems unlikely to me, whichever way you look at it:

1. Optimistic case for NVDA where revenues are sustained by AI chips requiring constant upgrading: This is possible. However, there is a glaring drawback to this thesis. Four major customers account for 46% of NVDA’s revenue. If these four companies are to maintain their level of spending on this one item for the next one or two decades, they would spend cumulatively hundreds of billions if not trillions of dollars on AI chips at a 76% gross margin for NVDA. It seems unlikely they would be willing to spend such amounts of money and not develop alternatives. Indeed, alternatives, however imperfect, may already be on the way, with AMD announcing the launch of its AI chips. AMD has achieved this with R&D expenses for the entire company (including other products) of $6B in 2023 (it seems logical AMD or other companies could cut into NVDA’s market share or gross margin with some support from other big tech companies, and it would make sense for those tech companies to do so if investments in AI chips continue at current levels for the next few decades and are not just another passing fad).
2. Pessimistic case for NVDA where AI capacity outstrips demand: NVDA revenues sharply decline once AI capacity has been built out and demand for upgrades do not fully offset the decline. This scenario is self-explanatory and NVDA’s stock price might implode the moment the market senses a loss of upward momentum in future sales/orders.

The full economic impact of new technologies may take decades

However, railroads were nowhere near reaching their full economic impact in 1887 when new track laid peaked or 1910 when cumulative track laid plateaued. As shown below, railroad freight volumes increased from around 50 billion ton-kilometers to nearly 400 billion ton-kilometers in the 100-year period from 1910 to the 2000s. The increases were uneven, with volumes flat for many years (e.g., in the post WWII period) before seeing growth in other periods.

According to the below index of railroad stock prices, the returns to shareholders from railroads from 1900 to the 1970s were largely flat. It turns out while railroads massively boosted the economy, changes in transportation patterns and demand meant the benefits were not necessarily captured by railroad companies (or reflected in the stock prices). Railroad stocks did not see a sustained bull move until after the 1980s bull market, possibly for other reasons such as market share becoming more concentrated in the top railroads. As shown in the below chart of railroad grain freight market share, in 1980 this was relatively dispersed among many companies while by 2011, 85% of volumes were concentrated in 4 companies.

Relevance to NVDA:

The above example shows that the gains of new technologies such as railroads may take a long time to accrue to shareholders. For decades after the railway boom peaked, railroad shareholders largely saw flat share prices, until the value of railroads as an investment were unlocked and rediscovered post the 1980s.

Moreover, even during the most intense ramp-up phase of railroad capacity, the increase in share price from the 1870s to 1900 (when 80% of the railroad track was laid) was not significant by today’s standards: share prices in 1900 were largely unchanged compared to 1870, compared to NVDA, which appreciated 1,000% in 2 years because of AI.

Broader Takeaway

Currently, it is difficult to project how much AI processing capacity is needed, this uncertainty may have enabled the market to justify a $3.4 trillion market cap for NVDA:

1. It is quite apparent to me that the market has no idea how much AI capacity is needed. Just witness how much NVDA’s stock price has been impacted by hearsay (example: stock price declining upon news of Chatgpt usage decline in an investment bank’s report, which some have suggested is inaccurate) and guesstimates about next quarter/year’s financials. Example: a recent report from Morgan Stanley suggests Blackwell capacity is sold out in the next 12 months, however given that 2025 forward P/E is nearly 50, whether demand is strong for another quarter or year has very little bearing on the long-term value of the company. Case in point: remember all the “shortages” in COVID-19 that turned out mostly to be one-off?
2. However, the actual demand for AI is probably finite, it just isn’t apparent at the moment when NVDA revenues will start to slow down, stagnate or even decline. The uncertainty gives the market room to imagine and fantasize.
3. The inflection point when the market decides NVDA’s gravy train has come to an end may come sooner than expected. The top tech companies are better capitalized and their investments in AI have surged. Railroad capacity measured by cumulative track mileage increased at 5% CAGR over a 35-year period from 1865 to 1900. Had the same amount of railroad mileage increase been accomplished in 20 years, the CAGR would’ve been 10%. In 10 years, the CAGR would correspond to 20% – at current rollout rates maybe AI capacity (for current purposes) is saturated in a far shorter timeframe than the forward earnings estimates for NVDA (which correspond to a P/E of 26 by 2028).

Risks to bearish view:

NVDA’s stock price has managed to plod higher and is near all-time highs. Given the uncertainties over future AI demand and how big of a lead NVDA has over competitors, NVDA’s stock price could continue upwards. I wouldn’t hold my breath on it, but maybe NVDA manages to achieve and sustain annual revenues of $200 billion a year at 76% gross margins for the next 20 or 30 years, as would be needed to support its $3.4 trillion market cap.

Conclusion:

Based on the above, I continue to believe that NVDA faces far larger risks to the downside than to the upside with a market cap of $3.4 trillion (which is nearly 12.5% of US total 2023 GDP, by comparison all railroads at their heyday in 1900 were an estimated 20% of GDP (railroads estimated as 40% of total market cap multiplied by market cap to GDP of 50%)).

I would definitely not recommend shorting (as no one knows how long this train could chug along for). But again, I would strongly caution investors who are investing money they cannot afford to lose (e.g., retirement savings) to be careful with NVDA, or at least consider keeping risk exposure to a manageable level. If an individual investor has significant sums of money in NVDA, he or she may consider selling at current prices near record highs and get out with a small profit or breakeven, rather than hoping the gravy train continues.

---

Analyst’s Disclosure: I/we have no stock, option or similar derivative position in any of the companies mentioned, and no plans to initiate any such positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

Seeking Alpha’s Disclosure: Past performance is no guarantee of future results. No recommendation or advice is being given as to whether any investment is suitable for a particular investor. Any views or opinions expressed above may not reflect those of Seeking Alpha as a whole. Seeking Alpha is not a licensed securities dealer, broker or US investment adviser or investment bank. Our analysts are third party authors that include both professional investors and individual investors who may not be licensed or certified by any institute or regulatory body.

---