The top 15 companies in the S&P 500 now account for 46.7% of the index's total market value. In March 2000, at the apex of the dot-com bubble, the equivalent figure was 27.2%. That is a 19.5 percentage point expansion of concentration into territory with no modern precedent. The Kobeissi Letter flagged the comparison on October 21, and the financial press picked it up as another bubble data point. It is not that simple, and the framing obscures more than it reveals.
Consider the internal geometry first. Nvidia carries 7.5% of the index. Apple carries 7.1%. Alphabet 5.9%. Microsoft 5.6%. Those four alone represent 26.1% of the S&P 500. The eleven names behind them add another 20.6%. And the structural punchline: the combined market capitalization of the smallest 434 constituents is roughly equal to the market value of the largest five. One index point of movement can be decided by a handful of order books while the operating conditions of 434 businesses remain irrelevant to the print.
This is the part of the story that deserves attention, because it changes what the S&P 500 is. In my audit work on protocol treasuries and institutional index products, I have watched the same mechanical dynamic unfold on-chain: a market-cap-weighted basket whose largest constituent appreciates automatically increases that constituent's weight, which forces passive vehicles to buy more of it, which pushes the price higher still. The feedback loop is not a bug. It is the design. Exchange-traded funds now hold close to half of their capital in fifteen companies, and the rebalancing arithmetic does the rest without a single human decision.
The bearish reading โ Michael Burry calling this a denial phase with six to nine months left, Peter Schiff noting 479 individual stocks at 52-week lows while the index trades near highs โ is built on a real observation. Market breadth is the weakest since the internet bubble, per Goldman Sachs. But breadth is a description, not a forecast. I sat through the 2021 cycle watching identical breadth arguments get made for eighteen months while the index compounded. Narrow leadership persisted through multiple stretches of the late 1990s for years before the reversal. A weak breadth reading tells you the median company is struggling. It does not tell you when the mean reverts.
The more serious problem with the bubble framing is the omitted variable. Any claim that 46.7% concentration equals 2000-level risk requires two numbers the analysis never supplies: the top 15's share of index earnings, and their forward valuation multiple relative to the rest of the index. In 2000, the concentrated names carried extreme multiples on thin or negative earnings. Today the concentrated names are, by and large, cash-generating monopolies with margins the 2000 cohort never approached. If fifteen companies represent 46.7% of market cap but something closer to 30% of index earnings โ a figure I have not verified and which the source material does not provide โ then the market is paying a premium for those fifteen. That premium is the falsifiable risk. Concentration by itself is a symptom of where capital has flowed, not proof of mispricing.
This is where Tom Lee's counterpoint earns its place. His argument that the current setup resembles 1997 rather than 1999, with a modeled cycle peak somewhere around 2033, is not a denial of froth. It is a claim about position on the technology diffusion curve. Infrastructure buildout in the early-to-middle stage looks exactly like this: a handful of suppliers capture the capex, the index tilts toward them, breadth collapses, and the skeptics are directionally right about structure while wrong about timing for years. Burry's six-to-nine-month horizon and Lee's multi-year horizon are not opposing views. They are answers to different questions, presented side by side to manufacture the appearance of a debate.
Here is the angle the bulls get right and the bubble chorus misses. The concentration is not purely a valuation artifact; it reflects a genuine consolidation of global profit pools into platform and compute infrastructure. That is a real economic shift, not a narrative. The index has stopped being a broad proxy for the US economy and become a leveraged instrument on AI capital expenditure and platform rent extraction. Anyone still using the S&P 500 to infer the health of the median American business is reading a thermometer that has been rewired to measure a reactor core. That methodological failure matters more than the bubble call itself, because it propagates into policy assumptions and household portfolio risk simultaneously.
The risk that actually deserves a hedge is not concentration. It is the gap between market-cap concentration and earnings concentration. If that gap is wide, the premium embedded in the top fifteen is exposed to any deceleration in their earnings growth, and there is no breadth buffer underneath to catch the index โ the 434 smallest names are already at lows and cannot rotate into leadership. If the gap is narrow, the entire panic is misdirected. Every passive retirement saver with a target-date fund is short that answer without knowing it.
Watch the quarterly S&P DJI rebalance weights, the ratio of equal-weight to cap-weight performance, the count of constituents above their 200-day moving average, and โ the signal nobody in this debate is publishing โ the top-fifteen earnings share against their 46.7% market share. Until that last number is on the table, the bubble question is being argued with half the evidence. Logic over hype. โ ๏ธ