Nvidia’s earnings could answer the AI bubble question and upend global markets in moment of truth for Magnificent 7

This, in a nutshell, is the narrative of the entire global stock market right now and the conundrum that no tech CEO or asset manager can avoid addressing: Is AI a bubble or not? 

Much is at stake. 

Since the current bull market began in October 2022, roughly 75% of gains in the S&P 500 have come from just seven stocks—Alphabet, Apple, Amazon, Meta, Microsoft, Nvidia, and Tesla. Those companies, known as the Magnificent Seven, have a combined market cap of about $21.5 trillion as of mid-November.

Their strategies, Apple’s perhaps excepted, are heavily dependent on AI. But if AI does not deliver the revenues or the efficiencies that Big Tech expects, the fallout in stocks will likely be severe—because the world’s investable assets are currently concentrated in AI and AI-adjacent stocks to an unprecedented degree.

The S&P 500 had risen 14.7% this year (at the time of writing), repeatedly breaking new record highs. But 40% of the index’s value comes from the 10 biggest stocks within it, all but one of which are tech companies.

Most of those companies, in turn, are pouring vast sums into AI for the development of new data centers, large language models, and the massive amounts of electricity they guzzle. Goldman Sachs estimates that capital expenditure on AI will hit $390 billion this year and increase by another 19% in 2026. Bank of America is even more bullish: It projects that AI capex will hit $1.2 trillion in 2030. 

The recipients of the lion’s share of that money are 10 AI companies that are interlocked with one another as customers and investors in an “increasingly circular” way, as a recent research note from Morgan Stanleydescribed it. That note referenced relationships between OpenAI, Nvidia, Oracle, Microsoft, CoreWeave, and AMD, involving billions in equity stakes, revenue sharing, vendor financing, and “repurchase agreement[s]” being passed back and forth among them. 

On one level, that’s normal: Most industries feature a collection of companies that do business with one another. The problem in AI is that the revenues currently being generated by AI companies are far smaller than the amount of capex being directed at them by the Magnificent Seven.

The back-of-the-envelope math is harsh: AI capex coming from tech companies listed in the S&P 500 is $400 billion or more per year, going forward. The biggest AI company—OpenAI, the maker of ChatGPT—has disclosed revenues of just $13 billion for 2025. Altman recently said revenues were “well more” than that, hinting they may hit $100 billion in 2027. But that’s still nowhere near the level of capex it’s committing to. 

OpenAI may have lost $12 billion in the third quarter of 2025 alone, according to a disclosure by Microsoft, which has a 32.5% stake in the company. Yet it has committed to spending $1.4 trillion to develop its product—and it is valued by its venture capital and other investors at $500 billion. 

These numbers don’t add up—and won’t until the massive AI rollout starts to yield real financial benefits. “That is exactly the discussion in the market at the moment. Can the 10 AI companies generate enough revenue to justify the capex?” says Torsten Sløk, chief economist at Apollo Global Management.

If the answer to that question is no, or more relevantly, not soon enough to satisfy investors’ expectations about the future, then the fallout in global equities could be brutal. 

Take, for instance, the Russell 2000, an index of small-cap U.S. companies: 806 of them, or about 40%, have no earnings or negative earnings. Counterintuitively, the stocks of unprofitable Russell companies have outperformed those that actually made money this year, according to Apollo. Most of those unprofitable companies are tech firms, riding the AI narrative.

With the Magnificent Seven dominating large-cap stocks, and money-losing small-caps soaring from optimism around AI, any reversal in sentiment will be widespread, especially if it’s triggered by a softening in AI infrastructure spending. And the broader repercussions in equity markets could be even greater.

Since 1990, U.S. assets of all kinds—stocks, bonds, property, you name it—have become increasingly dominant globally. U.S. stocks now account for about 60% of the valuation of all stocks on the planet, according to Christian Mueller-Glissmann, Goldman Sachs’ managing director and head of asset allocation research. And technology stocks represent about 45% of all U.S. stocks, worth $26 trillion or more at the close of October, according to S&P Global.

In other words, most of the world’s assets currently look like an upside-down pyramid, teetering on its point. The wide base at the top consists mostly of U.S. equities; underneath that, the performance of those equities is driven by just seven public tech companies. Those seven companies are funding 10 or so much smaller private AI companies, upon whose fortunes they currently rely. And those AI companies—at the very bottom of the inverted triangle—are largely unprofitable. 

For that reason, owning the S&P 500 via an exchange-traded fund, traditionally one of the safest and most common bets for smaller “retail” investors, isn’t providing the diversification it used to. Today, it’s largely a bet on a few globally massive tech platforms—concentrating millions of people’s retirement savings toward the tip of that pyramid.

The world’s dependence on U.S. equities stems in part from the fact that the U.S. is simply the largest economy in the world, says Mueller-Glissmann, making the concentration a proportional reflection of economic reality. But it also means that if U.S. equities are in a bubble, then the whole world is in a bubble—whether we like it or not.

The reason U.S. stocks have become so much bigger than foreign stocks is that the U.S. has a heavier concentration of finance and tech stocks compared with the rest of the world, Mueller-Glissmann adds. Those sectors enjoy a high degree of operating cost leverage, or the ability to increase revenues without adding many new staff or resources. An app with 200 million users doesn’t have anywhere close to twice the running costs of an app with 100 million users, but if a gold mining company wants to double its capacity, it pretty much needs to double its costs in equipment and labor. 

America has an enormous amount of this “financialization,” Mueller-Glissmann says. “That makes [the] argument a bit more scary, in the sense that this world portfolio is getting more and more important for the global economy in terms of driving wealth effects and in terms of driving financial conditions.

So for the next 12 months, investors will stay closely tuned to the revenue picture in AI. For now, Goldman Sachs, J.P. Morgan, Apollo, and Bank of America have all either published research or told Fortune directly that they expect capex growth for AI to continue unabated through 2026. The bubble has a way to go, in other words. The advice—as always—is to sell…but not right now! (For some concrete investing ideas, see the following pages.) 

At some point, the tide will go out, and stock investors, venture capital firms, and Big Tech’s AI capex “hyperscalers” will all want to know who has a viable business and who is swimming naked, according to Lisa Shalett, Morgan Stanley Wealth Management’s chief investment officer. That, in turn, could be the catalyst for a much broader stock market reckoning. 

Shalett believes that the current period of American exceptionalism—in which U.S. companies and assets dominate the global economy—is due to unwind in what she calls a “great rebalancing.” And when that happens, she wrote in a recent note to clients, its impact won’t be limited to companies with direct stakes in the AI boom. 

Since the financial crisis of 2008, Shalett argues, American stock market outperformance relative to the rest of the world has been “‘supercharged’ by historic monetary policy intervention” that kept interest rates near zero; deficit spending and fiscal stimulus, including about $4.6 trillion in relief outlays at the height of the COVID pandemic; “and the fruits of globalization, which were enhanced by the privilege of having the world’s reserve currency.” Those factors, Shalett believes, are likely to phase out over the next five to 10 years, with negative implications for U.S. stocks across the board. 

And that’s where the AI bubble question takes on even more weight. Over the past year, sky-high projections about AI’s economic benefits have helped investors ignore a steady flow of troubling economic indicators. If AI reality fails to live up to the hype, gloom could spread quickly.

With global trade becoming more expensive under President Trump’s tariffs, inflation now a semipermanent feature of economic life, and the U.S. facing potential pressure from the bond market to balance its fiscal books, investors are unlikely to be generous if they detect that reliable earnings-per-share growth has been replaced by speculative folly. 

“People realize that there’s only so much of this story that you can price into the future without the ‘Show me the money’ moment,” Shalett tells Fortune. “We think we’re pretty close to the moon already.” And if asset values come back down to earth, the challenge will be to avoid burning up on reentry.

A boom for the few

This article appears in the December 2025/January 2026 issue of Fortune with the headline “How an AI bubble could ruin the party.”