Generated by Codex with GPT 5.6 Sol XHigh

The familiar warning that the stock market is not the real economy now has a second layer: benchmark indices are no longer good portraits of the stock markets they supposedly represent. A handful of enormous companies, many tied to the artificial-intelligence investment boom, increasingly determine the movements of indices containing hundreds or thousands of stocks. The result is that diversified-looking portfolios can carry far more concentrated and volatile risks than their labels suggest.

A giant can move an entire market

The S&P 500 has risen even while American consumers remain unusually gloomy, but its distance from everyday economic sentiment is only part of the problem. Nvidia alone makes up about 8% of the index. A 3% move in the chip designer’s shares can shift the entire S&P 500 by roughly 0.25%, even if the other 499 companies do nothing. No previous constituent has had greater influence over the benchmark.

Other markets are more extreme. TSMC, the Taiwanese manufacturer of Nvidia’s processors, has doubled in value to about \$2trn and now accounts for more than 40% of Taiwan’s TAIEX index. Regulators even removed a rule that had capped a single company’s weight in index-tracking funds at 30%. Although the TAIEX contains more than 1,000 stocks, a standard measure of concentration suggests that it behaves like an equal-weighted portfolio of only six companies.

South Korea’s KOSPI tells a similar story. Samsung Electronics and SK Hynix together briefly exceeded 40% of the index, and other listed companies often rise and fall with the same AI hardware cycle. During a turbulent stretch from May to July, the KOSPI was four times as volatile as its 2025 average. It moved by at least 5% about once every three trading sessions, including an 11% fall followed three days later by an 18% rise. Leveraged exchange-traded funds popular with South Korean investors amplify the swings.

Concentration is only half the danger

A benchmark dominated by a few companies does not have to be unstable. Nearly half of Switzerland’s broad SPI index is concentrated in Roche, Novartis, Nestle, ABB and UBS. Yet those firms span pharmaceuticals, food, industrial machinery and banking, industries that also represent much of the country’s listed economy. Their relatively steady earnings and dividends make the group less prone to moving in lockstep.

Taiwanese and South Korean benchmarks are different because their largest companies share exposure to chips, technology hardware and the uncertain pace of AI capital spending. Their concentration is therefore reinforced by correlation: investors may appear to own hundreds of stocks while effectively making the same economic bet repeatedly. When expectations for AI infrastructure change, much of the index changes with them.

An index is a forecast, not a census

Benchmark dominance can disappear. Novo Nordisk once towered over Denmark’s market because of Ozempic, only to lose weight when Eli Lilly produced a stronger obesity drug. That reversal illustrates what share prices actually represent: investors’ collective expectations about future profits, not a faithful snapshot of today’s corporate economy.

The distinction matters more when technological change is rapid. AI leaders have reached enormous valuations, while abundant private capital allows potential challengers to remain unlisted and therefore absent from public benchmarks. An index may thus overweight today’s celebrated incumbents while omitting companies that could displace them.

Passive investors still gain broad exposure across listed firms, but the number of holdings can create a false sense of diversification. The risk depends not only on how many names appear in an index, but on their weights and whether their fortunes rest on the same underlying story. Increasingly, the world’s most prominent benchmarks are less like maps of national stock markets and more like amplified wagers on the AI boom.