The assumption that owning a cap-weighted index represents a neutral, diversified starting point is no longer defensible. That is the central argument Frank Thormann, Portfolio Manager, Global Equities at Schroders, advances in a new research piece1 that examines the structural consequences of extreme market concentration and the AI capital expenditure cycle now reshaping global equity benchmarks.
Concentration Is the New Headline Risk
The numbers are arresting. As at September 2026, the ten largest companies by market capitalization represent over 40% of the S&P 500 and 28% of the MSCI World. Thormann frames this starkly: "Market-capitalisation weighted indexing can therefore overstate diversification and understate risk." The only historical parallel is the mid-1960s, when postwar industrial giants such as AT&T, General Motors and IBM held a similarly outsized share of the index. What was once a 500-company portfolio is now, in meaningful respects, a highly leveraged bet on a single theme.
The mechanism driving this concentration is well understood but underappreciated in its consequences. As passive inflows continue to allocate capital based on market cap rather than fundamental merit, top constituents absorb ever-larger flows irrespective of return on capital. The benchmark, Thormann observes, "represents a set of material and increasingly correlated investment exposures in its own right."
The AI Capex Cycle Changes the Calculus
Layered on top of that concentration risk is an investment cycle with no historical precedent in its scale. Total AI-related capital expenditure is estimated to reach $1 trillion in 2026, with five hyperscalers, Amazon, Alphabet, Microsoft, Meta and Oracle, accounting for more than 80% of that figure. Thormann projects total industry spend will double to $2 trillion by 2027, well above current consensus. SpaceX, he notes, is emerging as a significant entrant to that spending cohort.
The comparison to prior landmark national initiatives, the Manhattan Project, Apollo, the internet buildout, is offered not as an equivalence but as a measure of sheer scale. At this magnitude, Thormann argues, "even modest overestimation of future demand could create substantial excess capacity and lower marginal returns." Passive indices embed this exposure indiscriminately; they cannot distinguish capital deployed with accretive return on invested capital from capital deployed defensively, out of competitive anxiety or fear of missing out.
The Investment Anomaly Passive Investors Inherit
Empirical market history is not encouraging for the fastest spenders. Academic literature from Titman, Wei and Xie (2004) and Fama and French (2015) documents a persistent investment anomaly: companies with the highest rates of capital growth have tended to underperform over three-to-five-year horizons as overcapacity, margin compression and lower marginal returns take hold. Thormann's analysis, drawing on Empirical Research Partners data, shows the recent outperformance of large-cap AI spenders is "historically unusual" against that backdrop.
The 2022 drawdown offers a pointed precedent. When rates rose sharply, the S&P 500's ten largest constituents lost 37% of their combined market value, versus approximately 20% for the index as a whole. Nvidia, Meta and Tesla fell 51%, 66% and 63% respectively. "With the 10 largest companies now representing roughly 40% of the index," Thormann calculates, "a 30% decline in this group would reduce the index by approximately 12%, even if every other constituent were unchanged."
The Case for Active Share
The response Thormann advocates is not blanket avoidance of high-capex companies. The structural winners of a generational technology transition should not be excluded categorically. The more productive approach, the Schroders Global Equities Team argues, is to assess each incremental capital dollar separately: does it earn more than its cost of capital and convert into durable free cash flow, or does it merely extend a costly competitive arms race?
Alphabet and Oracle offer a live illustration. Their diverging share-price performance relative to the SPGLOB Index since September 2024 demonstrates that markets are already differentiating between funding quality, monetization capacity and prospective ROIC, rather than rewarding AI-related spending indiscriminately.
Thormann draws on Robert Novy-Marx's gross profitability premium and Richard Sloan's accrual anomaly to anchor the analytical framework: cash-flow sustainability, not headline earnings or balance-sheet expansion, is what persists.
Five Takeaways for Advisors and Investors
1 A cap-weighted index is no longer a risk-neutral baseline. It carries extreme concentration in a single macro theme that many investors may not have intended to own. 2. AI capex at $1 trillion in 2026, doubling to $2 trillion in 2027, is large enough to move macroeconomic aggregates; passive indices absorb this exposure without filter. 3. The historical record on high-capex quintiles is a warning: fastest capital growth has consistently correlated with weakest subsequent relative returns. 4. Active share is not merely alpha-seeking; it is a mechanism for reducing unintended correlated risk embedded in a heavily concentrated benchmark. 5. The fundamental question for each AI spender is not how much it is spending but whether that spending converts to free cash flow. That distinction is now being priced.
Footnote:
1 Thormann, Frank. "The Age of AI Has Exposed the Structural Flaws of Cap-Weighted Benchmarks." Schroders, 11 Nov. 2026, www.schroders.com/en-us/us/intermediary/insights/the-age-of-ai-has-exposed-the-structural-flaws-of-cap-weighted-benchmarks/.
