Big Tech's AI Investment: One Year Left—Reading the Structural Risks of the S&P 500 Through Numbers
Big Tech companies, including Alphabet, Amazon, and Meta, are continuing AI capital expenditures on an unprecedented scale. These companies account for over 35% of the entire S&P 500 index through their top 10 stocks, and their performance is directly linked to the returns of the entire U.S. stock market.Nick Kohus and Jessica Rae, co-founders of Datar Research, analyze this situation as one where "investors have until the end of 2026 to wait." The structure is such that if Big Tech does not demonstrate the justification for their AI investments within the year, capital will flow to other markets.
1. Big Tech's business model has transformed—a shift to capital-intensive
1-1. Rapid deterioration of asset efficiency
The first thing Kohus points to is "asset efficiency." This metric, calculated by dividing revenue by property, plant, and equipment (PP&E), shows how much revenue is generated from one dollar of physical capital. The average for Alphabet, Amazon, and Meta was 2.2x as of 2023. It is a highly capital-efficient business model that generates over two dollars of revenue from one dollar of capital.
That figure is expected to drop by 42% by 2026.Alphabet will see its ratio halved compared to 2023. "Ford's revenue-to-PP&E ratio is 5x, which is much higher than that of the Big Tech companies. Ford is a typical capital-intensive industry, but tech companies are now even more capital-intensive than that. This was unthinkable three or four years ago," Kohus points out.
Furthermore, regardingMeta, the PPE ratio is below 1.0—meaning they have less revenue than capital. In addition, Kohus adds that Meta is accumulating off-balance-sheet debt in the form of long-term operating leases, making the actual capital intensity even higher than the figures based on PP&E alone suggest.
1-2. Management decisions to pour all cash flow into investment
The trend in the ratio of capital expenditures to operating cash flow for each company is also noteworthy. This ratio, which was 44% in 2023, reached 50% the following year and 70% in 2025. And in their 2026 plans, Alphabet is at 103%, Meta at 106%, and Amazon is even planning investments exceeding 133% of their operating cash flow.
"Each company calculated 'how much we can earn this year' and decided to 'spend it all.' That is where these numbers come from," Kohus explains. The fact that companies are pouring more funds into AI infrastructure than they can generate themselves shows that this investment is not within the scope of 'surplus capacity' but is 'all-in'.
2. Margin compression—will stock prices be justified until investment turns into recovery?
The average operating margin for Alphabet, Amazon, and Meta over the past three years was 34–39%. In 2026, this is expected to fall to an average of 34%, a compression of about 5 percentage points year-over-year. "The way to prove the value of the investment is to show an improvement in incremental margins. If profit margins do not start to rise from 2027 onwards, this investment will not be justified," says Kohus.
The market dislikes declining profit margins because it raises doubts about competitive advantage and future profitability. In addition to the large absolute amount of capital expenditure, the fact that it is accompanied by a decline in profit margins is a double burden on the stock price.
On the other hand, Kohus points out that there is a kind of 'agency problem' between management and shareholders. Shareholders have diversified portfolios and prioritize capital efficiency over which company wins. But for corporate management, AI investment is a matter of survival. If they neglect investment, they will miss the platform shift of the AI era. "There is no option not to invest. Moreover, the current stock price valuation (a P/E ratio of 25–30x) itself means that investors agree with this process. Management can argue that 'the stock price says so'," Kohus states.
3. Investor options—stay in U.S. stocks or diversify?
3-1. Structural differences between the S&P 500 and ACWX
Rae organizes investor options within the framework of"staying in the S&P 500 or moving to global diversification (ACWX)." The S&P 500 has about 17 percentage points more tech exposure compared to ACWX (MSCI All Country World Index ex-US), and consequently less in financials, materials, and industrials.
Furthermore, the concentration levels are significantly different. In the S&P 500, the top 10 stocks account for more than one-third of the entire index (Nvidia about 8%, Apple about 7%, Alphabet about 5.5%, etc.). On the other hand, in ACWX, even the largest stock, TSMC, is at 4.1%, and the rest are scattered at 1.6% or less. "U.S. mega-cap tech alone accounts for about 35% of the S&P 500, but the equivalent group of stocks is only about 10% in ACWX," Rae points out.
3-2. Surge in non-U.S. stocks and statistical overheating
In terms of returns over the last 100 trading days, non-U.S. stocks outperformed U.S. stocks by 11 percentage points. Compared to the average since 2010 (3.5 percentage points), this is a deviation of 2–3 standard deviations, which is statistically extremely unusual. Historically, non-U.S. stocks tend to surge right after U.S. stocks have been excessively strong. In fact, in September 2025, U.S. stocks recorded an outperformance of more than 1 standard deviation, and it is possible that capital inflows into non-U.S. stocks occurred as a reaction to that.
"Buying non-U.S. stocks now means getting on board at a 2–3 standard deviation level. You should be aware of that risk," says Rae. On the other hand, she also points out, "If AI investment does not recover as expected, the S&P 500 will mathematically be unable to outperform non-U.S. stocks due to its tech-heavy structure."
3-3. The Future of European Equities and American Exceptionalism
Kohus acknowledges that there is some rationale for shifting funds into European stocks. This includes the possibility that labor market disruption from AI adoption may be smaller in Europe, where social safety nets are robust, compared to the U.S.; changes in corporate governance modeled after Japanese corporate reforms; and fiscal stimulus driven by increased defense spending. However, he cautions that one-third to 40% of the gains in European stocks in 2025 were due to currency factors (a stronger euro and pound) and did not directly reflect improvements in corporate fundamentals.
In the long term, over three, five, and ten-year periods, the S&P 500 has outperformed non-U.S. stocks by 4 to 6.5 percentage points annually. "The basis for this superiority is that U.S. companies have leveraged disruptive innovation at scale and focused on growth and profitability. With companies like SpaceX, OpenAI, and Anthropic waiting in the wings for IPOs, the U.S. growth pipeline remains robust," says Rae.
"Overweighting U.S. stocks is synonymous with betting that AI capital expenditures will turn into returns. And if they don't show that by 2026, shareholders will not wait for the answer." This statement by Kohus summarizes the essence of this debate. Big Tech management needs to justify their investments with numbers, and investors are being forced to decide where to park their capital while waiting for those results.
