Investors should enjoy the final months of 2026 while they can as the AI-led stock market boom is due to go bust soon, according to analysts.

For now, there are still gains to be had. James Reilly, senior markets economist at Capital Economics, reiterated an earlier forecast for the S&P 500 to end this year at 8,250, up 7.7% from Friday’s close, then plunge 21% to 6,500 by the end of 2027.

“On balance, we think the data look consistent with a late-stage bubble,” he wrote in a note on Thursday. “Most of the factors we consider are at, or close to, levels that have preceded past stock market peaks.”

Reilly flagged several bubble indicators that he’s been tracking:

Stock valuations are consistent with a late-stage bubble. For example, the market’s cyclically adjusted price-to-earnings ratio is close to its dotcom peak, while the S&P 500’s valuation compared to Treasury bonds is also near dotcom extremes.

Expected earnings growth looks unsustainable. Forward 12-month earnings-per-share growth for the S&P 500 is in line with the peak of the dotcom bubble.

The sustainability of AI investment is in doubt amid massive spending and shrinking free cash flow. The combined free cash flow for the top AI hyperscalers is expected to turn negative in 2027.

Market-cap concentration of indexes in fewer stocks is at extreme levels, and that narrowness is often associated with unsustainable rallies.

Equity issuance is booming and consistent with a late-stage bubble. Given the pipeline of IPOs and follow-on offerings, another gusher of stocks is on the way. In the past, such activity has signaled a bubble’s end is just months away, not years.

Reilly didn’t mention the recent surge in Treasury yields, with the 10-year rate hitting 4.97% on Friday.

But for Rockefeller International Chairman Ruchir Sharma, it’s another major bubble-busting indicator to watch.

In a recent Financial Times op-ed, he warned the AI bubble could pop when the 10-year yield “decisively breaches” 5%, which has been the upper end of its range since the dotcom era.

“This breach would signal the start of a new era of tighter money, in which AI mega projects will be harder to fund,” Sharma added.

Borrowing costs that high would hit the AI boom in different ways. For one, hyperscalers will likely issue fewer bonds to finance their spending. They will also have more trouble issuing new equity as yields above 5% have historically been a headwind for stocks.

In addition, yields topping 5% would start to approach nominal GDP growth, making the national debt even more unsustainable, he pointed out.

While others on Wall Street have said yields are merely normalizing after years of being suppressed by central bank policies, Sharma noted the U.S. is much more addicted to debt now as the burden has exceeded 100% of GDP.

“As a result, debt-servicing costs are much higher now,” he wrote. “Rising public borrowing costs will squeeze other borrowers sooner, and hit the bubbly AI markets harder.”

Even staunch bulls are getting more anxious. Wall Street veteran Ed Yardeni lowered the odds of his “Roaring 2020s” stock market scenario for the rest of the decade from 80% to 70% and raised the odds of a bearish outcome from 20% to 30%.

“Admittedly, recent developments in the oil and bond markets are unnerving,” he said in a note Saturday.