The November 2026 AI sell-off: what it did to MSCI World and S&P 500 investors

In the week to 6 November 2026 the S&P 500 lost 2.7% and MSCI World 2.5% in dollars (2.4% in euros), leaving the S&P 500 5.7% below its 6 October record and MSCI World 5.8% below its 13 August record. Getting back to those records needs gains of 6.0% and 6.2%.

The Nasdaq Composite closed 10.9% below its record, inside the usual definition of a correction. In an MSCI World ETF, six companies at the centre of the AI build-out made up 18.3% of the fund on 5 November; in an S&P 500 ETF, 26.3%. The practical task is to measure overlapping holdings and check whether a larger fall would interrupt contributions or force sales. A single bad week does not establish that moving long-term money to cash would improve its compounded growth.

Key takeaways

1. The world index peaked in August, not October. MSCI World in dollars set its record on 13 August and closed 5.8% below it on 6 November; in euros the record came on 12 October and the gap is 4.3%.
2. A fund of 1,247 shares still puts 18.3% in six companies. In iShares' S&P 500 ETF the same six weigh 26.3%, so adding one ETF to the other buys more of the same companies.
3. Crossing a 10% loss threshold does not tell you how deep the fall will become. Of 31 completed Nasdaq corrections since 1971, 18 stopped before 20%, 11 reached 20-50% and 2 went beyond 50%.
4. Deep drawdowns can delay compounding for years. $100 at the March 2000 MSCI World record was worth $82.40 ten years later on the price index, before dividends and costs.
5. You can measure your exposure and your own hurdle; you cannot read the end of a fall from its start. Look through your ETFs, find your peak, and separate money you need within a few years.

How much did a world-index or S&P 500 investor lose in the week to 6 November?

The price indices lost between 2.4% and 2.7% over the week, with their largest daily fall on Thursday 5 November. That day the S&P 500 fell 4.36%, the Nasdaq Composite 5.52% and MSCI World 3.28% in dollars. Both US indices had risen 1.4% to 1.6% on 4 November and recovered part of the fall on Friday.

Week to 6 November 2026: S&P 500 −2.68%, Nasdaq −3.37%, MSCI World −2.52% (USD) and −2.37% (EUR); below record: −5.65%, −10.94%, −5.84% and −4.27%. Source: FRED SP500 and NASDAQCOM, MSCI World index 990100, AP, 6 November 2026; weekly figures (CSV), daily levels (CSV).

These are price indices, without dividends. With net dividends reinvested, MSCI World lost 2.50% in dollars and 2.36% in euros over the week, practically the same. FRED had not yet published the S&P 500 close for 6 November when we checked, so that close is AP's 7,376.86, which equals FRED's 5 November close plus the 23.13 points AP reported.

AP attributed the fall to the US elections: AI-linked stocks "tumbled this week on worries that Democratic victories in statehouses across the country could mean a slowdown in further construction, particularly in Texas." It reported two daily falls of 9.3% and 3.2% for Nvidia, 6% falls in Tokyo and Seoul after the US plunge on Thursday, and a 1.9% loss for Germany's DAX. Those are press figures; we have not verified them against exchange data.

Why is the world index measured from August, not October?

Because MSCI World in dollars did not set a new record in October: on 6 October, when the S&P 500 closed at its high, the world index was still 0.98% below its 13 August close. The US accounts for about 73% of MSCI World, but its country mix and company weights differ from the S&P 500's. A world-index holder was already 3.4% below the August record in dollars before the week began.

For a euro investor the picture differs because the dollar strengthened. The ECB reference rate fell from $1.1534 per euro on 13 August to $1.1165 on 6 November, a 3.2% fall in the euro. That lifted the euro value of dollar assets, so MSCI World in euros set its record on 12 October and closed 4.27% below it on 6 November. Over the week itself the exchange rate barely moved, from 1.1172 to 1.1165.

Currency gains can also reverse. A euro investor in an unhedged world ETF carries equity and foreign-currency exposure, and the S&P 500 and MSCI World comparison shows that both indices are dominated by the same US market.

How much of a world index is tied to the AI build-out?

On 5 November 2026, six companies made up 18.26% of iShares' Core MSCI World UCITS ETF and 26.26% of its Core S&P 500 UCITS ETF. They are Nvidia and Broadcom, the two chip companies, and Microsoft, Amazon, Alphabet and Meta, the four cloud and platform companies among the ten largest holdings. The grouping is ours, not an index classification.

Six companies: 18.26% of the iShares Core MSCI World UCITS ETF and 26.26% of the iShares Core S&P 500 UCITS ETF, holdings as of 5 November 2026. Source: iShares Core MSCI World holdings and Core S&P 500 holdings; largest holdings (CSV).

Apple, the largest holding in both funds at 5.73% and 7.92%, is left out, as are chip companies further down the list such as Micron, AMD and ASML. A different classification, information technology, was 27.86% of the World ETF and 35.53% of the S&P 500 ETF. That sector includes companies outside our six and excludes Amazon, Alphabet and Meta, so it is not a broader measure of the same AI group. The ten largest holdings were 26.35% and 37.58%.

MSCI's own index factsheet, dated 30 October 2026, gives Nvidia 5.25% of the index, the ten largest constituents 26.58% and the US 72.77%, across 1,248 constituents. Nvidia's lower 4.74% in the fund on 5 November is consistent with its share-price fall in between. The fund and the index are different documents, so cash and timing also explain small differences.

What does that concentration do to a fund? Take a scenario, not a forecast: the six fall by 30% and every other holding is unchanged. Arithmetic alone then takes 5.48% off the World ETF and 7.88% off the S&P 500 ETF. In a real sell-off the other holdings need not stay still: the crash study across asset classes found that spreading equities across regions did not protect in 2008.

Is a 2.7% weekly fall unusual?

It is uncommon but not rare: since November 2016, 37 of 521 weeks, including this one, ended with an S&P 500 loss of 2.68% or more. The previous one was the week to 4 April 2025, which matches AP's description of the worst week since the April 2025 tariff shock. The worst week in that span cost 14.98%, in March 2020.

For MSCI World in dollars, 150 of 1,557 weeks since January 1997 lost at least as much as this week's 2.52%, most recently in the week to 6 March 2026. Its worst week, to 10 October 2008, cost 20.05%. The S&P 500's 4.36% fall on 5 November was matched or exceeded on 10 earlier days since November 2016, eight of them in 2020 and two in April 2025.

A weekly loss of this size changes compounded wealth very little on its own: a 2.5% fall needs a 2.6% gain to undo. The risk for a long-term investor is the path that sometimes follows.

What happened after earlier Nasdaq corrections?

Most completed Nasdaq corrections since 1971 ended before a 20% fall, but crossing the 10% threshold alone did not identify which would become deeper crashes. We counted every episode in which the index closed at least 10% below its previous record, until a new record close. The current episode, open since 6 October 2026, is the 32nd.

The episode list (CSV) gives the dates of all 32 Nasdaq episodes, and of MSCI World's corrections since 1997. The March 2000 episode first closed 10% below its record on 30 March 2000, within three weeks of the peak, and only reached its trough in October 2002. The 2026 episode is also the second this year: the Nasdaq fell 13.2% from October 2025 to March 2026 and regained its record in April.

These counts describe the past, not probabilities for this episode. They come from one price index without dividends, and the 1970s and 1980s Nasdaq was a different market from today's.

What did the 2000-2002 and 2022 tech-led falls mean for compounded growth?

The 2000-2002 fall left a world-index investor below the 2000 price level ten years later; the 2022 fall cost about two years. In both episodes the Nasdaq fell further than the broad indices we measured, so they show what a technology-led fall did to a diversified world index.

Nasdaq −77.9% in 2000-02 (15.1 years to a new record); MSCI World −51.4% in dollars (6.7 years) and −58.9% in euros (14.5 years); in 2022, −36.4%, −27.1% and −17.6%, each recovered within 2.3 years. Source: FRED NASDAQCOM and SP500, MSCI World index 990100; drawdown table (CSV).

$100 in MSCI World at its 27 March 2000 dollar record was worth $79.20 five years later, $82.40 after ten years and $120.40 after fifteen, an annualised 1.25% on the price index. The Nasdaq turned $100 at its March 2000 record into $96.30 after fifteen years. A 77.9% loss needs a 353% gain to recover, the arithmetic explained in our guide to recovery time and loss maths.

Dividends matter over such periods. MSCI publishes daily net-return levels only from 29 December 2000, so the full episode cannot be measured with dividends. From that date to 27 March 2015, $100 grew to $142.90 on the price index and to $187.30 with net dividends reinvested, before fund costs and taxes.

The euro investor's 2000-2003 loss was deeper and longer because the euro rose 27.1% against the dollar between the euro index's September 2000 peak and its March 2003 trough, at ECB reference rates. In 2022 the euro fell 14.0% from January to the dollar index's October trough, which cushioned the fall in euros.

With net dividends, MSCI World fell 26.1% in dollars in 2022 and regained its record in December 2023. In euros it fell 16.9% and regained its record in September 2023.

What can you check in your own portfolio this week?

These checks use your own statements and the issuers' holdings files; none of them predicts where this fall ends.

1. Look through your ETFs. Multiply each fund's weight in your portfolio by its share in the six companies, using the issuer's latest holdings file.
2. Add single shares you own directly. A separate Nvidia or Microsoft position comes on top of the same companies inside your ETFs.
3. Measure losses separately from cash flows. Use your broker's time-weighted, dividend-inclusive return series; peak divided by current level minus one gives the recovery hurdle. A raw account balance changes when you add or withdraw money and cannot measure the investment loss.
4. Name the money you need within a few years. A fall matters most when spending or debt payments force you to sell; the withdrawal study shows how selling during a fall leaves fewer units for the recovery.
5. Check underlying currencies and hedging. Trading an ETF in euros does not hedge its foreign assets. A world ETF has several currency exposures, with substantial US-dollar exposure.
6. Replay losses alongside contributions and spending. Use total returns after fees where available, and check whether the 2000-2002 and 2022 paths would have forced sales. Price-only recovery times omit dividends and are not a forecast.

For example, a portfolio split evenly between the two iShares ETFs above would hold 22.26% in the six companies, the average of 18.26% and 26.26%. That is an illustration of the calculation, not a suggested mix. The independent-bets analysis explains why the number of funds is a poor guide to the number of return sources.

Before adding a defensive holding, compare its fees, expected return and behaviour during shared losses with the allocation it replaces. Test compounded wealth over the same horizon, including any spending it helps fund. A lower-volatility portfolio that reduces expected growth without meeting a funding need fails that test; so does an expensive hedge whose payoff cannot justify its recurring cost.