For more than four decades, the investing world relied on a central assumption: a portfolio made up of 60% stocks and 40% bonds would offer protection when markets turned lower. Over roughly the past 20 years, another assumption took hold alongside it: that a small group of dominant technology companies would keep growing into whatever valuations investors assigned them. That idea has shifted forms over time, from the FAANG stocks to the “Magnificent Seven” during the pandemic and, more recently, to a new class of AI “hyperscalers.”

Now, both assumptions are coming under pressure from sources that rarely deliver negative views about the markets they cover.

Goldman Sachs, one of Wall Street’s most consistently bullish research firms, published a note on Monday authored by chief global equity strategist Peter Oppenheimer, who said that “there does not appear to be a valuation bubble, but there may be an earnings bubble” in the technology sector.

The same day, Apollo chief economist Torsten Slok wrote in his Daily Spark note that “the 60/40 portfolio is broken,” arguing that with the AI trade slowing down and government debt projected to reach 175% of GDP, “neither the 60 nor the 40 responds to what made it work in the first place.”

The notes came after a week of Big Tech earnings that produced unusually large stock moves both higher and lower for major technology companies, while analysts debated whether the reaction reflected the true shape of the AI moat or “financial nihilism.”

Microsoft made history with a 17% stock surge, adding nearly $500 billion in market capitalization in a single day, its largest one-day move since 2008. Oppenheimer said Monday that the market is changing in a way it has not since the Great Recession.

A regime cracking, not just a ratio

The 60/40 rule was never an immutable law. Its theoretical foundation traces back to Harry Markowitz’s 1952 work on portfolio theory, which demonstrated mathematically how combining assets with different risk profiles could reduce overall portfolio risk without sacrificing expected returns and later earned Markowitz a share of the 1990 Nobel Memorial Prize in Economic Sciences. But it became institutional orthodoxy during the four-decade stretch of falling interest rates that began in the early 1980s, adopted as the default allocation framework by pension funds, university endowments, and millions of individual retirement accounts. That decline allowed bonds to serve both as income and as ballast against equities, a dynamic that market analysts have described as the “golden age” of investing.

Slok’s warning is not a reaction to one weak earnings season. He has argued since at least 2023 that rates would stay “higher for longer” than consensus expected. By late May 2026, he had sharpened that macro view into a specific yield-curve mechanism: Front-end rates were rising on sticky inflation; the middle of the curve was under pressure “because of hyperscaler issuance”; and long-end rates were climbing on “more Treasury supply and less Fed demand.” His Aug. 3 argument that “the real risk emerges if the AI trade reverses or markets become more worried about government deficits” continues that multiyear thesis rather than introducing a new one.

Oppenheimer’s note does not use Slok’s exact phrase, but it arrives at the same structural conclusion from the equity side. “More government debt, increased issuance, and persistent inflation have all contributed to a higher cost of capital, leaving earnings as the key driver of returns,” he wrote. “We think this trend will continue.”

A rotation not seen since 2009

Oppenheimer’s team said that for the first time since 2009, the equal-weighted S&P 500 has outperformed the market-cap-weighted index by more than 7.3%. Because the standard S&P 500 weights companies by their total market value, the largest technology firms have disproportionately driven index returns in recent years; equal-weighted outperformance means the average stock in the index is gaining ground against those dominant names. Since the mid-2000s, and especially since the financial crisis, U.S. markets had become increasingly dominated by a small number of megacap technology names. But now, Goldman said, “market participation has broadened beyond the largest stocks,” helped by resilient economies, a pickup in M&A, and what it described as “the sharp momentum unwind of recent weeks.”

The mechanism behind that unwind is capital expenditure. Since ChatGPT’s emergence, Oppenheimer noted, the surge in capex among hyperscalers has increasingly eroded their premium cash flows, forcing them to turn to debt and equity markets for funding. As a result, the premium that the five biggest U.S. stocks once commanded over the S&P 500’s other 495 has “almost disappeared.” Goldman called the shift “a healthy normalization following years of very high concentration in both market capitalization and performance.”

The bubble callers Wall Street ignored

Two months before Oppenheimer’s note, warnings were largely coming from outside the official research departments of large banks.

On June 3, Acadian Asset Management’s Owen Lamont argued that expected long-term S&P 500 earnings growth had reached 20.2%, above the 18.6% peak in 2000. He said the figure made “today’s optimism … yet another way in which 2026 is looking like 1999.”

JPMorgan CEO Jamie Dimon, speaking days earlier at Bernstein’s Strategic Decisions Conference, was more direct: “It’s gung ho, folks … There’s a lot of exuberance out there.” Dimon rooted his concern in 1972, 1986, 2000 and 2007, each a period when confidence was high, deal activity was strong and the consensus believed fundamentals justified the optimism, just before conditions reversed.

He also said that $10 trillion to $12 trillion in deficit spending had mechanically boosted corporate profits, warning that markets were treating a “sugar high” as if it were organic strength.

Ray Dalio took the argument further, telling Bloomberg Television that his bubble indicators showed markets were “rising close to—not at—the same level in 2000 and the same level in 1929.”

Those warnings preceded July 14, when IBM suffered the worst single-day stock crash in its 115-year history: a 25% drop that erased roughly $40 billion in value after a revenue miss of just 3.7%. The decline came on the same day JPMorgan and Goldman reported blowout earnings, a contrast that economist Steve Hanke said pointed to two bubbles.

A classic valuation bubble can be seen in measures such as the Shiller CAPE ratio, which compares current stock prices to average inflation-adjusted earnings over a rolling 10-year period, he told Fortune at the time, “but the more dangerous mispricing … isn’t in valuations at all. It’s in the earnings themselves.”

BCA Research’s Peter Berezin has been making the same argument for months, saying the AI trade is “primarily an earnings bubble rather than a valuation bubble,” the kind that has historically appeared in boom-bust sectors such as pre-2008 banks and pandemic-era work-from-home stocks.

The market has not collapsed as a whole, even as some technology stocks have been re-rated or de-rated, and shares have mostly moved sideways for several months. That has lent support to Oppenheimer’s view that the current adjustment is healthy for equities. But as Berezin noted, earnings bubbles are harder to identify than valuation bubbles because analysts “typically only cut profit estimates after stocks have already fallen.”

IBM illustrated that point, with BofA and UBS trimming estimates only after the stock had already cratered.

Weeks later, Big Tech earnings forced markets to reassess again.

During the week of July 26–31, investors split sharply on the largest AI spenders—not over earnings, but over capex credibility. Microsoft and Amazon rose 18% and 10%, respectively, while Alphabet fell as much as 4% and Meta dropped nearly 10%, despite consistently strong earnings and revenues from nearly all of them. Investors are no longer rewarding capex simply because it exists; they are starting to ask where the money is coming from and whether each company’s balance sheet can support the pace.

Why the old playbook no longer works

Oppenheimer’s own figures now show why the technology sector’s earnings story has become harder to rely on. Goldman’s forward-implied growth for the sector has risen since 2020, but the 10-year rolling earnings growth rate “has accelerated well beyond the 2000 peaks.” In other words, realized growth has already exceeded dot-com-era extremes, while forward expectations have not yet fully caught up.

That leaves investors with an uncomfortable set of facts. The four-decade rate regime that made the 60/40 rule dependable has effectively ended. The two-decade run of tech dominance that defined this generation’s bull market is being de-rated. And even the analysts most inclined to defend the AI trade are now beginning, cautiously and on the same August day, to concede that the skeptics may have been early — not wrong.

For this story, Fortune journalists used generative AI as a research tool. An editor verified the accuracy of the information before publishing.

This story was originally featured on Fortune.com