What to Make of AI Bubble Talk

Is AI a bubble? That question hangs on many minds, particularly as volatility escalated last week. But there is a rather glaring flaw in the argument: If it were a bubble, it is unlikely many people would call it that. Actual bubbles are typically characterized by “irrational exuberance”—as some guy once put it four years before a bubble truly inflated—not near-ubiquitous warnings the boom is ready to bust or a bust-in-progress.^[i] All the bubble talk looks self-deflating to us.

A bubble—like 2000’s—occurs when sentiment is euphoric, blinding investors to a fundamentally weaker reality than expected. At the turn of the millennium, newfangled dot-com companies were supposedly at the vanguard of a new economy of “clicks, not bricks.” Now, a quarter-century later, you could say this digital revolution has been at least partially realized. But at the time, the sunlit uplands of never-ending earnings growth spectacularly failed to materialize. Instead, a glut of cash-burning startups went public to satisfy euphoric investors’ craving for “the next Dell,” and the market took over two years to work through the excess as the dot-coms imploded.

Exhibit 1 shows one way to see this: S&P 500 Tech and Tech-like Communication Services profits’ and market capitalization as percentages of the whole S&P 500. In 2000, Tech’s share of market cap spiked, but its relative earnings didn’t keep pace. It wasn’t even close. Then, as the bubble popped, Tech’s modest profits turned to losses (hence the seemingly impossible negative reading for Tech’s percentage of S&P 500 earnings) and its market cap share tanked as stock prices imploded. Nowadays, Big Tech’s share of market cap is higher, but more sane relative to the earnings it commands.

Exhibit 1: Spot the Bubble

Source: FactSet, as of 11/21/2025. S&P 500 and its Tech and Communication Services sectors’ net income and market capitalization, December 1994 – October 2025.

Sentiment today is no doubt warmer than it was last April or even this time last year. And there is little doubt some AI enthusiasm is ahead of its skis. But that is chiefly concentrated in newer, private startups and pure play firms that have yet to book profits, not well-established, publicly traded multinationals sporting huge revenue streams (even without AI) and massive margins.

For the most part, though, headlines are littered with doom. We see incessant chatter about circular AI funding, akin to leveraged vendor-financing schemes, which are set to unravel once confidence frays. Many people aren’t even optimistic about AI’s prospects should it succeed, fearing the job consequences from dislocating workers, largely reprising arguments that have surrounded new technologies since the spinning jenny.

Related to bubble fears, some see the emergence of firms rising to multi-trillion-dollar market caps—and comprising a larger share of US indexes—amid higher valuations. But markets with only a handful of stocks dominating their capitalization aren’t inherently an automatic problem. Lots of countries carry concentrated market positions. For example, the 10 biggest stocks in France, Taiwan and Switzerland each all hold more sway domestically than those in America, yet few bat an eye at them.

Don’t get us wrong. When a few stocks drive a country’s performance, that is something investors should be very much aware of, as it calls for global diversification. But concentration alone isn’t necessarily problematic when it is commensurate with fundamentals. If one company’s earnings are far greater than others, then it wouldn’t exactly be a mystery if its market cap is, too. This says nothing about the broader market’s health.

Valuations, as we have written, aren’t necessarily telling, either. They reflect data and metrics stocks have already priced. Earnings data publish after a quarter’s end, so trailing P/Es can only look backward. Stocks look forward. Yes, some try to bridge this using forward P/Es, but earnings forecasts are largely opinions rooted in today’s sentiment—subject to error.

Besides, the bull market this year isn’t all about AI and related Big Tech. These companies aren’t the only high-returning S&P 500 constituents. Year to date, 170 S&P 500 stocks exceed the index’s 13.6% gain, while 271 have positive returns.^[ii] The rally is fairly broad-based. Beyond Tech-like sectors, Utilities, Industrials and Health Care are up double digits year to date, too, with only one sector (Consumer Discretionary) down.

Meanwhile, fundamentals—what matter most to markets—are better than dour headlines imply. In aggregate, Big Tech’s enormous earnings—and growth—support their market caps. Reality is mostly matching expectations, by and large. Concerns about circular investing and debt-financed AI buildouts make headlines, but that isn’t driving capital expenditures, which are being funded mainly out of operating cash flow. Most debt being deployed is backed by Tech and Tech-like firms’ solid balance sheets. Corporate credit spreads—among the most sensitive measures of default risk—have widened somewhat in sympathy with sentiment, but they remain near record lows.

Fundamentals were manifestly not fine in 2000. Not only did credit spreads blow out with dot-coms’ (and their vendors’) balance sheets far more vulnerable, their main funding mechanism at the time—IPOs—famously collapsed as the market worked off the stock supply hangover. IPOs this year? Aggregate US year-to-date proceeds are running less than $40 billion, on the low end of the historical range and less than half 2000’s $84 billion.^[iii] On that score, sentiment seems far from exuberant, much less irrational.

While we do think there are signs this bull market is later-stage, there are almost as many offsetting that. In a way, bubble sentiment is a case in point. Many still-skeptical investors cast emerging optimism as frothy euphoria. But to us, that action itself illustrates we are still some distance from that point—which a clear-eyed look at fundamentals supports.