“2025 as seen by MarketScreener” (5/6). With a 17% gain in 2025, the S&P 500 is on track to post its third consecutive year of double-digit gains. The Nasdaq 100 can return over 20% over even three years. Once again, it's all thanks to artificial intelligence. But what has changed in 2025 is that AI is no longer just a market topic. It is also the driving force behind US GDP growth.
When you remember how this year started with Wall Street getting hit squarely by Donald Trump's tariffs, 2025's performance is almost unbelievable. In early April, the S&P 500 briefly fell. bear market (20% peak-to-trough correction). A few months later, Donald Trump relaxed this policy slightly. customs duty invoice. And above all, American companies continued to invest aggressively in artificial intelligence.
all in
Capital spending by the five largest hyperscalers (Amazon, Alphabet, Meta, Microsoft, and Oracle) is expected to reach nearly $400 billion by 2025. And this number is expected to increase further in the coming years. The central issue now is return on investment.
At this stage, the market seems fairly confident on this point. The cloud businesses of Amazon, Microsoft, and Google (the three major players in this market) continue to grow rapidly. In the most recent quarter, Azure's growth rate (Microsoft) was 40%, AWS's was 20%, and Google's was 32%. These are impressive numbers for a company with a market capitalization in the trillions of dollars.
Despite this growth rate, concerns about return on investment are understandable. but, hyperscaler They continue to argue that they need to accelerate, that the main risk is lack of investment, and that the main problem is lack of capacity (data centers, chips, etc.). It all feels a bit like venture capital logic. Even if some of your investments don't pay off, you can still make a lot of money in the end, so give it a try.
All the more so since these companies remain automated teller machines. Therefore, the cost of mistakes is not very high. A recent example is the Metaverse. This was a big bet for Facebook, which changed its name to Meta in late 2021. This was more of a failure, as the Reality Labs division will suffer cumulative losses of nearly $70 billion by the end of 2024, according to Financial Times calculations. Still, this failed pivot only cost the company 12 to 18 months of losses in the wilds of the market. few quarters in stock dead money (No one wanted it anymore) before the AI train got the company back on track.
don't forget to return the money
But that last point may no longer be so obvious, and the market hasn't realized it yet. For about a decade, these stocks have been popular with investors for two main reasons. asset light (It's a company with few physical assets, so it doesn't have the investment burden that comes with it.) And it's a cash machine that can recoup tens of billions of dollars through stock buybacks.
But won't the transition to AI make that investment thesis obsolete, since hundreds of billions of dollars must now be spent building physical infrastructure, i.e. data centers? Capital expenditures are amortized over several years. But most of your investment is in the Nvidia chip, which is a consumable item that will likely need to be replaced fairly quickly. Hyperscalers believe that servers can last up to 6 years, but this is not clear.
All of this raises questions about shareholder returns. Is it possible to deploy a large share buyback program again if a large portion of cash flow must go to capital expenditures? BofA calculates that nearly 70% of hyperscalers' cash flow will be consumed by capital expenditures this year and 80% next year. In the past 10 years, it has been around 30% to 50%.
Capital expenditures and cash flows of five major hyperscalers. Source: Bank of America
on the same boat
Therefore, these companies now have to rely on debt, which was not the case before. At the end of September, Oracle issued $18 billion in bonds. Meta raised $30 billion in late October. Alphabet made $25 billion in early November. And in mid-November, Amazon announced its intention to raise $15 billion. Nothing dramatic for a company with a very strong balance sheet, but it still indicates that it will no longer be able to cover all its investments on its own and will need to find funding elsewhere.
Beyond the amount of funding and increased reliance on debt, the cyclical nature of this ecosystem poses a problem. Customers are funding suppliers, equity stakes are being taken in all directions… and among them is a player called OpenAI, which is not profitable but whose funding needs are estimated at $1.4 trillion by 2029.
Ecosystem circularity may be the main risk. This can lead to a vicious cycle if one player slows down. But for now, this circle is healthy. One company's capital investment becomes another company's profit, which generates growth. Valuations are pricing in a bright future for everyone, with artificial intelligence pushing the S&P 500 to near 7,000 points and the Nasdaq over 25,000 points this year.
K-shaped economy
And it's not just Wall Street that's being lifted by artificial intelligence. It is now the main driver of growth in the United States. The GDP growth rate for the first half of 2025 was 1.6%. AI-related investments drove growth, contributing 1.4 percentage points to growth over the year. cent Bank of America estimates it rose 1.5 points in the second quarter;
AI-related investments contributed to growth. Source: Bank of America
In other words, excluding AI investment, the GDP growth rate in the first half was almost zero. So while some sectors are buoyed by the AI boom, the rest of the economy is stagnant. This finding is supported by data on capital expenditures.
Capital investment by category. Only the AI-related parts are rising sharply, the rest are falling. Source: Apollo Global Management
The AI boom has thus revived a concept that emerged with the coronavirus pandemic: the K-shaped economy. While some sectors have made a strong recovery, others continue to sink. While the US economy currently appears to be holding up well, it is actually an economy where AI is driving investment and consumption is largely dependent on the wealthiest.
And the two are linked. The AI boom is causing the stock market to soar. This surge leads to an increase in household wealth, especially among the wealthy, who own the most stocks. And as your wealth increases, you are encouraged to spend more. That is the wealth effect.
This K-shaped economic concept also explains what public opinion polls have found: a majority of Americans are dissatisfied with the state of the economy. Despite Wall Street's record and despite strong economic growth. GDP in the third quarter increased at an annual rate of 4.3%. But at the same time, the “Current Economic Conditions” component of the Michigan Consumer Confidence Index hit a 10-year low in December. That's because Americans face a cooling labor market (meaning it's harder to find a job and fewer raises) and still high prices. According to Politico, nearly half of Americans (46%) think the cost of living is the worst in living memory.
