Short answer: nobody can prove a bubble until after it bursts, and the evidence on AI stocks is mixed. Prices relative to long-run earnings are close to 2000 levels and the market is more concentrated than it was then, yet forward earnings multiples are well below the dot-com peak. For your money, the useful question is how much a bubble could cost you. Measure your AI exposure, replay a 2000-style fall on your own portfolio, and act only on what you could not live with.
Who this is for: long-term investors in index funds, ETFs or tech stocks who are wondering what to do after the November 5, 2026 selloff. What it does not cover: short-term trading, or whether any single company is a buy. We do not know where the market goes next. This is education, not personal investment advice.
What happenedA chip selloff that hit everything
The selloff of November 5 came after the midterm elections. CNBC reported that investors feared the AI buildout would slow, with Democrats winning Congress and interest rates rising, while Qualcomm added a weak forecast. The Nasdaq Composite closed down 5.52%, the S&P 500 down 4.36% and the Dow down 3.44%. Early trading on Friday, November 6 showed a partial rebound, and one or two days do not settle this argument in either direction. Bubbles are judged over years.
The case forWhy it looks like a bubble
Shiller CAPE (price vs 10-year earnings)
- Dot-com peak
- 44.19
- 2026
- About 41.6
- Reads as
- Bubble-like
Top 10 stocks' share of the S&P 500
- Dot-com peak
- About 26%
- 2026
- About 40%
- Reads as
- More extreme
S&P 500 forward P/E
- Dot-com peak
- 25.2
- 2026
- 19.7
- Reads as
- Cheaper
Tech sector forward P/E
- Dot-com peak
- About 58
- 2026
- About 25
- Reads as
- Much cheaper
| Measure | Dot-com peak | 2026 | Reads as |
|---|---|---|---|
| Shiller CAPE (price vs 10-year earnings) | 44.19 | About 41.6 | Bubble-like |
| Top 10 stocks' share of the S&P 500 | About 26% | About 40% | More extreme |
| S&P 500 forward P/E | 25.2 | 19.7 | Cheaper |
| Tech sector forward P/E | About 58 | About 25 | Much cheaper |
Source: CalculatorAI · calculatorai.app · Multpl (CAPE, Oct. 8, 2026); J.P. Morgan Guide to the Markets (forward P/E, Mar. 31, 2026 and Mar. 24, 2000); Morgan Stanley and Pensions & Investments, via secondary reports (top 10 share); Fortune, May 18, 2026 (tech P/E)
These are the parts of the bubble argument that hold up best:
- Valuations are near a record on one long-run measure. The Shiller CAPE compares price with 10 years of inflation-adjusted earnings. Its long-run average is about 17, and only the late 1990s went higher.
- The market depends on a few companies. With about 40% in 10 stocks, a problem at a handful of them becomes everyone's problem. That is why so much of an S&P 500 fund is Nvidia and why this selloff spread so fast.
- The spending is huge, and payback is unproven. Reports of Reuters' analysis put capital spending at five big tech companies near $557 billion through the third quarter of 2026, about double a year earlier. We read that figure in secondary reports, not in filings. Other reports project $725 billion to $760 billion for the big four for the full year.
- Politics now matters. Voters passed many local measures against data centers this month, and the sector's growth depends on getting them built and powered.
The case againstWhy it may not be 2000
- Today's leaders earn real money. Forward multiples are far below the 2000 level because the biggest companies have large profits, while many dot-com leaders had little or none. That is the main reason the two right-hand rows in the table look so different.
- Expensive is not the same as about to fall. High valuations have historically pointed to lower long-run returns more reliably than to the timing of a crash.
- New technology can be real and still overpriced. The internet changed everything and the Nasdaq still fell 77.9%. So AI being useful does not settle the question, in either direction. What matters is what investors pay for it, and whether the spending earns a return.
The historyWhat a real bust looks like
The numbers from the last comparable episode are worth knowing, even though there is no reason to expect the next one to match.
Show these figures as a table
| Value (% fall from peak) | |
|---|---|
| S&P 500 — −49.1% | 49.1 |
| Nasdaq Composite — −77.9% | 77.9 |
| QQQ (Nasdaq-100 fund) — About −83% | 83 |
Source: CalculatorAI · calculatorai.app · Index closing levels: Nasdaq Composite 5,048.62 to 1,114.11; S&P 500 1,527.46 to 776.76. QQQ from our VOO vs QQQ guide.
The recovery matters more than the fall. After a 77.9% loss, the Nasdaq needed a gain of 353% to get back to even, and it did not regain its 2000 high until 2015. The S&P 500 needed 96.6% and regained its old closing high in 2007.
And there is the opposite mistake. Alan Greenspan asked about "irrational exuberance" on December 5, 1996, with the S&P 500 at 744. It then rose 105% to its 2000 peak. Selling at that speech would have meant missing the doubling, and the 2002 low of 777 was still above where the index stood on the day he spoke.
Your portfolioRun the stress test
Multiply what you hold by each fall. This is a replay of the worst stretch of the last bust, not a forecast, and it ignores dividends and new contributions.
100% QQQ
- Value at the low
- $17,000
- Fall
- −83%
70% S&P 500 fund, 30% QQQ
- Value at the low
- $40,730
- Fall
- −59%
100% S&P 500 fund
- Value at the low
- $50,900
- Fall
- −49%
60% S&P 500 fund, 40% cash and bonds
- Value at the low
- $70,540
- Fall
- −29%
| Portfolio | Value at the low | Fall |
|---|---|---|
| 100% QQQ | $17,000 | −83% |
| 70% S&P 500 fund, 30% QQQ | $40,730 | −59% |
| 100% S&P 500 fund | $50,900 | −49% |
| 60% S&P 500 fund, 40% cash and bonds | $70,540 | −29% |
Source: CalculatorAI · calculatorai.app · CalculatorAI calculation; historical replay, not a prediction
A second way to see the concentration: if the top 10 stocks fell 50% and the rest of the S&P 500 stayed flat, a fund with 40% in those stocks would lose 20%. With the 26% share of 2000, the same fall would cost 13%.
Add up your true exposure
Count the look-through holdings of each fund, plus any shares you own directly. Funds with different names often hold the same companies, as our guide to ETF overlap shows. The portfolio tracker keeps all your accounts in one view.
Check when you need the money
Money needed within about five years should not rely on stocks recovering in time. After 2000, the Nasdaq took 15 years to come back.
Pick the fall you could sit through
Look at the table and choose the row you could hold through without selling. If your portfolio sits in a row that would make you sell at the bottom, your mix is too aggressive for you, bubble or no bubble.
Change the mix by rule
Use a target and a rebalancing band, not a forecast. Trimming back to target is a decision you can defend whichever way markets go next. Selling everything requires being right twice: on the way out and on the way back in.
Mind the tax
In a taxable account, selling winners triggers capital gains tax. Use new contributions or tax-advantaged accounts to rebalance first, and check tax-loss harvesting before December 31 if you hold losers.
What to watchSigns that would change the picture
- Big tech's capital spending guidance: a cut would be the most direct sign that the buildout is slowing. Nvidia usually reports its fall quarter in the second half of November, so check its investor relations page for the date.
- Who pays for the buildout: more spending funded by debt instead of cash flow raises the risk. Heavily indebted data center companies are the weakest link if demand disappoints.
- Long-term interest rates: higher Treasury yields make distant profits worth less and borrowing more expensive.
- Revenue from AI customers: the bull case needs buyers of AI services to earn back what they spend.
Common questionsFAQ
Is the AI bubble about to burst?
Nobody knows. There is good evidence on both sides, and valuation measures have a poor record at timing. Plan so that you would be fine in either case.
Should I sell my AI stocks now?
That depends on how much of your portfolio they are, when you need the money and what tax you would owe. Trimming a position that has grown beyond your target is a reasonable rule. Selling everything is a bet on timing, and it has often been wrong.
Is an S&P 500 fund safe if AI crashes?
It is diversified across about 500 companies, but it is not protected. About 40% sits in its top 10 stocks, and in the last bust the index fell 49.1%. It took until 2007 to get back to its old high, not counting dividends.
Should I buy the dip?
If you already invest regularly, keep your plan and avoid increasing your risk just because prices fell. Our look at the first 3% down day in over a month found that the next month was close to a coin flip.
What would be a good hedge?
For most long-term investors, the simple answers are bonds and cash for near-term needs, and broader exposure outside the largest US stocks. Options and inverse funds cost money over time and are hard to time.
The takeawaySize the risk instead of predicting it
You do not have to win the bubble debate. Work out how much AI you own, look at what a 2000-style fall would do to your money, and set the mix you could hold through it. If you have done that, the next headline changes much less about what you should do.
