On Oct. 6, the S&P 500 closed at a record 7,818.93. Ten days later it had slipped to 7,659.44 after a rough week for anything tied to artificial intelligence, and the question investors keep asking is back: is AI a bubble in 2026? CoreWeave, a data-center company backed by Nvidia, lost about a quarter of its value in five trading days, and chip stocks fell eight sessions in a row.
A 2% dip in the index is ordinary. What's unusual is what was happening underneath it. Even when the S&P 500 sat within 0.2% of its record, 84% of the stocks in it were at least 10% below their own highs, according to a tally by The Kobeissi Letter, a market newsletter. So there's been no stock market correction in October 2026 for the index, which needs a 10% drop to earn the label, but most of the stocks inside it are already in one.
Nobody knows whether this ends in a crash, a long sideways stretch or more records, including the professionals arguing about it on TV. What you can know is how much AI you already own, how today compares with past manias, and which moves make sense if the debate is keeping you up at night.
S&P 500 concentration and the AI stocks you already own
The S&P 500 weights companies by market value, so the biggest companies get the biggest slices of your money. Right now the biggest are almost all AI-linked. Bank of America groups ten of them (Apple, Microsoft, Nvidia, Amazon, Alphabet, Meta, Tesla, Broadcom, AMD and Micron) as the "AI Big 10" and puts them at close to 40% of the index. J.P. Morgan Asset Management's data show the ten largest companies at 40.6% of the S&P 500's value as of Sept. 30.
Here's what that means in dollars. Put $10,000 in an S&P 500 index fund and about $4,000 lands in roughly ten companies. Around $2,000 goes to just three: Nvidia, Apple and Microsoft. Nvidia alone gets about $800. If the fund split your money evenly across all 503 holdings, each would get about $20.
That's only the top of the list. Chip suppliers, cloud companies such as Oracle and utilities selling power to data centers ride the same wave further down. If your 401(k) holds a target-date fund, its U.S. stock slice almost certainly owns the same giants.
The narrowness shows up in performance, too. The equal-weight version of the S&P 500, which gives every company the same slice, led the regular index for much of 2026 but has fallen behind for the year after a brutal September, CNBC reported on Oct. 15. When a handful of giants carry the index, the typical stock can lag even while the headline number sets records.
Switching to a broader fund changes less than people expect. Vanguard's Total Stock Market ETF (VTI) owns roughly 3,600 companies, yet its top 10 holdings were still about 33% of the fund as of July 31, according to Schwab. We walk through that overlap in our VOO vs VTI comparison.
The case for a bubble, and the case against
The worried camp has real numbers on its side. The Shiller CAPE ratio, which compares prices with ten years of inflation-adjusted earnings, stood at 40.9 at the end of September against a 30-year average of 28.9, per J.P. Morgan. The only other time it's been this high was the dot-com era, when it peaked around 44 in late 1999. Concentration is higher than it was then, too. The top 10 were roughly a quarter of the index in 2000.
Then there's the spending. J.P. Morgan's consensus figures put capital spending at Alphabet, Amazon, Meta, Microsoft and Oracle at about $799 billion in 2026, nearly double 2025's $416 billion. That rhymes with 1999 and 2000, when telecom companies borrowed heavily to lay fiber-optic cable far ahead of demand. The CoreWeave scare was about exactly that. Reuters reported investors are worried that debt-heavy AI infrastructure companies may not turn a profit in time to repay what they've borrowed, and CoreWeave's bond yields climbed above 15%.
Rates make all of this touchier. The Fed raised its benchmark rate to 3.75% to 4.00% on Sept. 16, its first hike since 2023, and September consumer prices rose 3.7% from a year earlier, hotter than forecast. The 10-year Treasury yielded about 5.3% at the end of September. When safe bonds pay that much, investors have less reason to pay up for profits that are years away. Joachim Klement, a strategist at Panmure Liberum, expects the S&P 500 to fall to 5,000 by the end of 2027, Reuters reported on Oct. 5. That's about 35% below where the index stood, and it's an outlier. Several other brokerages expect 8,000 or more by the end of this year.
The other camp points to earnings, which are strong. At the March 2000 peak, the S&P 500 traded at 25.2 times the next 12 months of expected earnings, a measure called the forward P/E. As of Oct. 16, FactSet put it at 19.0, below its own five-year average of 19.8. Prices are high, but profits have risen fast enough to keep pace. FactSet's analyst tally expects S&P 500 earnings to grow 32.8% in 2026, and with 9% of companies reported, third-quarter earnings were tracking 29% growth.
The giants aren't priced like the dreamers of 1999, either. J.P. Morgan shows the top 10 trading at 20.9 times forward earnings, in line with their long-run average of 20.8, while producing about 36% of the index's earnings. Plenty of dot-com darlings had no earnings at all.
History offers a middle lesson. In the early 1970s, investors piled into the "Nifty Fifty," blue-chip growth stocks considered so reliable you could buy them at any price. In 1972 they averaged about 42 times earnings, versus 19 for the S&P 500, by Wharton professor Jeremy Siegel's math. In the 1973-74 bear market the S&P 500 fell 48%, and Polaroid, which had traded near 90 times earnings, lost 91%. Yet Siegel found that someone who bought the whole group at the top roughly kept pace with the S&P 500 over the next quarter century, trailing by about half a percentage point a year. Great companies, a bad starting price and a rough ride. The dot-com bust was harsher: the S&P 500 fell 49% from March 2000 to October 2002, and the Nasdaq lost about 78%.
What the Anthropic IPO and Q3 earnings say about the mood
One of the clearest reads on sentiment is Anthropic, the maker of Claude, which is expected to go public in November. Anthropic told prospective investors its third-quarter revenue was $16.8 billion, up 45% from the second quarter, according to reports from CNBC and Reuters on Oct. 15. That's huge growth by any normal standard. But revenue had more than doubled the quarter before, and Reuters reported some investors are now discussing a valuation around $1.5 trillion, down from the roughly $2 trillion that had been expected. Neither figure is official. The takeaway for the wider market is that at these prices, investors want AI growth to stay spectacular, and even a strong quarter that's slower than the last one can reset expectations. Our Anthropic IPO guide covers how the listing works.
Third-quarter earnings season is the next test. Early reporters have mostly beaten estimates, though by a slimmer margin than usual, according to FactSet. The Fed meets again Oct. 27-28, and the biggest AI spenders typically report around the end of the month. Whatever they say about next year's spending will move the whole index, because they are so much of it.
Should I sell my index funds?
For most long-term investors, no. If you want less AI risk, there are better tools than an exit.
Check your real exposure. Look up the top 10 holdings of every fund you own across your 401(k), IRA and brokerage account. Plenty of people own the same five companies three times over without realizing it.
Know your fund options. A total-market fund dilutes the giants slightly. An equal weight S&P 500 ETF such as Invesco's RSP holds the same companies but gives each about 0.2% and rebalances every quarter. It charges 0.20% a year versus 0.03% for VOO, or $20 versus $3 on $10,000. Expect it to trail whenever the mega-caps lead, as it has since September. Some investors split money between the two instead of picking one.
Look abroad and down the size ladder, with eyes open. Small-cap and international funds hold far less Nvidia, but they aren't AI-free. In emerging markets, the top 10 companies, led by chipmakers TSMC, Samsung and SK Hynix, make up about 40% of the MSCI index, per J.P. Morgan. In Europe and Japan the top 10 are about 31%. Small caps also tend to be more sensitive to rising rates.
Rebalance instead of predicting. If your plan is 80% stocks and 20% bonds and a rally has pushed you to 86% and 14%, selling back to 80/20 takes some AI gains off the table on a rule, not a hunch.
Keep your automatic contributions going. Investing a fixed amount every month means a dip buys more shares at lower prices. Here's why dollar-cost averaging works best in choppy markets.
Don't cash out to wait it out. A $10,000 investment in an S&P 500 ETF in October 2006 grew to about $82,300 by this October with dividends reinvested, according to a Motley Fool calculation. Miss just the 10 best days and it finished near $35,700. Seven of those best days came within two weeks of one of the 10 worst, which is why people who sell during a scare usually miss the rebound. In a taxable account, selling can also trigger a capital gains bill.
Bubble or boom, the AI tilt in your index fund isn't hidden. Spend 20 minutes this week finding your largest holdings, decide how much of your money you're comfortable having in a dozen tech giants, and set your mix and contributions to match. Then let the headlines argue among themselves.
This article is general information, not personalized investment advice.