원문 보기
Everyone is asking, is AI a bubble? This is the wrong question. The real question is what part of the AI stack are you betting on. Because the risk you take looks very different depending on where you are in that stack . So, let's build an AI data center from scratch, layer by layer. Who designs the brain?
Who actually produces it? Who provides its memory? Who connects thousands of such devices into one machine? Who supports the power supply? Five companies, five different tasks, and five very different risk profiles. Nvidia shares have risen 73.5% per year over the past five years , for a total return of 1,471%.
But look at the graph: almost none of this was the result of steady growth. The stock was flat, practically dead from 2015 to 2020, then woke up in 2023 and has only been rising since then. Only one elite investor holds this position, David Tepper, and the insider share is a pleasant 4%. The return on invested capital is 78.4%, with a net profitability of about 49%.
These are the metrics of a software company derived from the hardware business. And this explosive growth is directly reflected in revenues. Minor fluctuations around 27 billion in 2022 and 2023 , then a jump to 130 billion by 2025. The graphs of free cash flow and earnings per share are very similar.
And here is a figure that is worth closer consideration. Nvidia's discounted cash flow model estimates an average cost with a safety margin of $149. They are currently trading at $206, with a P/E ratio of 31.6 and a valuation gap of about 27%. In fact, the price assumes that the next 5 years will be as successful as the last two.
But to better understand the current price of this exceptional company, let's turn to forecasts, fundamentals, and ratios. And if we look at the history of the price- to-earnings ratio, this is one of the lowest P/Es in a long time. In January, it was twice as high. A year earlier it was four times higher, around 120.
300, 250, 180. So, taking all that into account, the current P/E of 31 is a really interesting opportunity. And if we look at other ratios, such as price to sales, price to book value, or price to cash flow, the graph looks pretty much the same. So maybe the discounted cash flow model doesn't value it high enough and we can still consider the investment.
The risk here is simple. This is the risk of expectations. Almost the entire investment strategy is based on the assumption that demand for AI will remain high for many years to come. Therefore, Nvidia is not the best place to be if you think that growth could be even slightly disappointing. TSMC's graph is not at all similar to Nvidia's graph.
The 5-year return is 273%. That's a powerful figure, but only a fraction of what Nvidia has. The share of insiders here is 0%—the lowest figure of anything we've looked at today. Again, David Tepper is one of the elite investors who own stocks. The return on invested capital is 28.8%, and the net margin is 40%.
But the real story lies in the balance. Net cash has quadrupled in 2 years, reaching $54 billion. Unlike Nvidia, this is a true dividend stock. $3.51 per year and a payout ratio of 25.5%. She actually pays you while you wait. And these shares are trading at a 54.1% premium to their true value with a P/E ratio of 29.5.
The demand for AI is so strong right now that it's hiding TSMC's true nature. This is a cyclical business, which means big ups and downs. The price chart for 2022 is proof of that. The ups and downs have not disappeared. They're just hidden under all this demand for AI. But ignoring this demand can also be a significant mistake.
The risk with TSMC is actually two risks superimposed on each other. The cyclical nature of chip production, which is currently masked by a boom, and the geopolitical risk of being on an island, which is very important for both the US and China. Micron's returns look almost as extreme as Nvidia's—63.2% compound annual growth rate over 5 years.
But look at the price today. At the end of 2025, it was around $200. Today it is $831. The shares have more than quadrupled in about half a year. The insider share is only 0.3%, and David Tepper is once again the only elite investor here. Micron's return on invested capital, or how well it turns money into profit, has averaged 14.9% over the past 10 years.
And this is the most cyclical stock on the list. In 2023, Micron lost $5.34 per share, and revenue fell almost in half. Two years later, revenue returned to $37 billion, and earnings to $7.59 per share. And we can see the cyclicality very clearly if we turn to forecasting. Look back 20 years , and whether we choose sales, net income, or free cash flow, the graphs go up and down throughout the years.
And if we go into the AI prediction agent and select future expectations, we can read it. Investors seem to expect Micron to benefit significantly over the next 12-24 months from AI- driven memory demand, especially the growth of HBM and data centers. Today's valuation looks overvalued, as the stock already factor in very high profits, strong margins, and continued growth, even though memory chip manufacturing remains a cyclical business.
So, let's look at the assessment. The P/E ratio is only 18.8—the most acceptable multiple of all we considered. But if you apply discounted cash flow analysis, the price looks simply absurd. The reason for this is the volatility of free cash flow. If we take the 5- year average, we are actually near zero.
So, again , the assessment depends almost entirely on which story you believe. Micron is a commodity business that has yet to try on an AI suit. The risk here is purely cyclical risk. Memory prices fluctuate wildly in both directions, and the lowest return on invested capital on this list is a reminder that this business doesn't have a deep, stable moat to protect its profits.
Arista's profitability is also very good. 59.2% average annual growth rate over 5 years. This is another great result for the company in the AI ecosystem. But this is where the model breaks down. Every other company today has one elite investor—David Tepper. Arista has zero of them . Return on invested capital 51.1%, net profit 35.6%.
The story with pure debt looks just great. Revenue, free cash flow, and earnings per share have all shown solid double-digit growth over the decade. This is truly impressive, especially considering that revenue has been growing steadily over the years. And this stability comes at a high cost. The valuation gap is 33.8% at a P/E ratio of 63.3, which is twice that of Nvidia.
This premium is the market price for future growth, not for current cash flow. Stability is not the same as security, where the price already assumes that this stability will persist. However, quality businesses are not often sold for cheap. Arista's risk is not in the business itself. It is in the price.
Its operations look almost flawless, but you are the one taking the risk, paying upfront for years of growth that are yet to come. The Vistra chart is the most striking example of a flat and then explosive trend on this list. The stock traded in a range of $15 to $23 for years, then soared to $162 by 2025.
51.6% average annual growth over five years. The insider share is 0.8%, including, you guessed it, David Tepper. This perfectly demonstrates his strategy for AI. He doesn't know which part of this ecosystem is the best bet, so he chose almost all of them . The return on invested capital is just 4.6%—the lowest on the list by a wide margin.
Net debt has almost doubled in 2 years to $19.3 billion. The profits are quite chaotic. Losses in 2021 and 2022, a jump to $7 per share in 2024, and then a drop to $2.18 in 2025. It's also the only company here with a truly consistently growing dividend, which has increased every year since 2019 and is now 91 cents.
The gap between price and fair value here is 54.4% with a P/E ratio of 26.1. Vistra is not a cheap company with a strong, protected business. This is an energy company that has taken on a lot of debt and has always had unstable results. Stocks have already risen significantly due to expected demand from AI.
This could still be a good investment, if only the growth of AI continues uninterrupted. Vistra carries the highest legacy risk among the five companies. Leverage. It has the weakest balance sheet and the lowest return on invested capital of all. So if the demand for electricity for AI cools even slightly, this company could be in serious trouble.
If you want to analyze other companies in the technology sector, simply go to Forecaster, use the sector map, and select technology. This will give you access to over 500 companies and their data. And the best part: if you use the link in the description and enter the code bold 10 at checkout, you'll get an additional 10% off your subscription. So, thank you Forecaster for sponsoring this video, and thank you for watching.
Comments 0