By Joseph Moss, International Banker
A global equity fund can contain 2,458 holdings and still allocate almost one-quarter of its value to a mere 10 securities, as was the case with the MSCI All Country World Index (ACWI) on August 31, when its 10 largest constituents—equating to just nine companies, given that Alphabet’s two share classes are counted separately—represented a hefty 24.26 percent of the index. Information technology (IT) accounted for 31.2 percent, while US companies represented 63.63 percent. In contrast, the top-ten weight of the ACWI at the end of 2023 was 18.4 percent, while information technology’s share was 22.9 percent. With concentration increasingly amassing within US technology firms, therefore, the standalone number of holdings in a portfolio is providing a progressively less complete picture of the forces determining returns.
Concentration risk arises when a small group of holdings, sectors or common earnings drivers generates a disproportionate share of a portfolio’s gains and losses. This is increasingly the case with technology holdings—within the ACWI, for instance, Nvidia, Apple and Microsoft sit in information technology; Alphabet and Meta are classified under communication services; and Amazon is assigned to consumer discretionary. Investors may consequently appear to own several sectors while remaining heavily exposed to AI investment and demand.
The same earnings cycle running through holdings presented as different companies, sectors and markets, then, makes technology concentration a prominent concern across the investment community. A global fund may contain thousands of securities and yet remain heavily dependent on a few US companies and/or the semiconductor supply chain and/or the commercial returns generated by unprecedented artificial intelligence (AI) expenditure. This concentration does not necessarily predict an imminent downturn; rather, a disappointment in AI spending or returns could affect several parts of the equity market simultaneously and, where projects have been debt-financed, their lenders as well.
“About 40 percent of the US equity market…is in the top 10 stocks, which are virtually all technological,” Ashley Lester, MSCI’s chief research and development officer, observed in January. “And so a view on the US is very significantly a view on AI and vice versa, given the lack of material exposure to AI elsewhere.” This assessment identifies four overlapping exposures within global portfolios: a small number of companies, the technology sector, the US market and AI. Although portfolio reports present these as separate categories, their performance increasingly depends on the earnings and valuations of the same group of US technology companies.
Market-capitalisation weighting is central to this development, with a company’s index weight rising as its stock-market value increases relative to other constituents.
Market-capitalisation weighting is central to this development, with a company’s index weight rising as its stock-market value increases relative to other constituents. It is practical for index funds, in that it allows them to invest large sums in widely traded shares and track the index with relatively low transaction and management costs. However, it automatically increases investors’ exposure to companies with share prices that have already risen the most.
The growth of index investing has also reproduced those weights across retirement accounts, advisory portfolios and institutional mandates. In July 2026, the share of assets within US domestic equity mutual funds and exchange-traded funds (ETFs) held in index-tracking products was 64.1 percent, whereas the share of the entire US stock market owned by those index funds was just 19 percent (at the end of 2025), with the remaining shares being held directly by households, pension funds, insurers, active funds and other investors.
With their share prices and market values rising significantly relative to other companies, therefore, the largest technology companies have each attained heavy index weighting. Index funds have then subsequently replicated that weighting across a large majority of fund portfolios. An investor placing $100 into a fund tracking the MSCI ACWI at the end of August, therefore, was effectively allocating $24.26 to its 10 largest securities, irrespective of whether the investor had formed a distinct view on each company.
Could active fund management act to disperse this concentration? In part, at least. A manager who sharply underweights the largest technology companies, for example, can still trail the index for extended periods while their share prices continue to rise. That said, deploying such a strategy must be defensible to clients and investment committees, whereas retaining the benchmark weighting is likely to require less justification.
This concentration issue would also pose less of a problem were the largest holdings to represent separate, distinct sources of earnings. With AI reducing that independence by linking companies across multiple industries to the same investment cycle, however, the interconnectedness between companies is only strengthening. Indeed, despite MSCI having established the AI value chain through nine individual segments, including hardware, energy providers and early adopters, those businesses together represent around one-third of global equity-market capitalisation, one-quarter of corporate earnings, one-fifth of capital expenditure and 40 percent of research and development (R&D) spending as of early 2026.
According to Vanguard, a group of roughly 70 companies globally is driving a “historically large, physically intensive” buildout centred on semiconductors and AI infrastructure. “This ‘AI complex’ comprises companies undertaking AI infrastructure buildout—including ‘hyperscalers’ Alphabet, Amazon, Meta, Microsoft, and Oracle—and companies across the semiconductor stack, data centres, and networking infrastructure that benefit from hyperscaler spending,” the investment advisor stated in a June 26 report. “For this year and next, we expect this AI complex to generate more than half of all US earnings growth.”
The composition of the MSCI ACWI Information Technology Index illustrates how an AI-related setback could spread across global technology holdings. While its 326 securities might suggest broad diversification across the index, its 10 largest constituents represented 64.47 percent of total market-capitalisation weighting at the end of August, with semiconductors and US companies accounting for 41.06 percent and 76.26 percent, respectively. Should AI adoption and/or monetisation ultimately undershoot expectations, and hyperscalers in turn moderate capital expenditure on key assets including chips, memory, networking equipment and data centres, earnings forecasts could weaken across a wide range of connected businesses that may have previously been treated as separate holdings.
Holdings across several industries can therefore depend on a small group of hyperscalers sustaining their expenditures and earning acceptable returns. The scale of the expenditure itself, meanwhile, underscores the sheer breadth of this dependence, with J.P. Morgan Asset Management forecasting Alphabet, Amazon, Meta, Microsoft and Oracle investing a combined $697 billion in total capital expenditure during 2026—$173 billion more than was projected at the beginning of the year. Their AI capital expenditure has risen from 33 percent of operating cash flow in 2023 to an estimated 93 percent in 2026, with the spending providing revenue to chip designers, memory producers, semiconductor-equipment manufacturers, network providers and data-centre developers.
The commercial returns on that infrastructure have consequently become critical to earnings across the supply chain. “Investors are now conditioning the reward for additional capex spending on evidence of greater final demand,” J.P. Morgan also noted in its “2026 Mid-Year Investment Outlook” published on June 15, observing that larger investment plans supported share prices during the latest earnings season only when analysts also raised their revenue forecasts. The trend thus suggests that investors increasingly require credible evidence that AI expenditure can generate additional sales and acceptable returns.
As AI investment becomes more reliant on borrowing, the implications will increasingly extend beyond equity markets.
As AI investment becomes more reliant on borrowing, the implications will increasingly extend beyond equity markets. “The size of anticipated investment needs will require firms to shift the source of financing from operating cash flows to debt, with private credit playing a rapidly increasing role,” Bank for International Settlements (BIS) economists Iñaki Aldasoro, Sebastian Doerr and Daniel Rees wrote in a January 7 report.
According to BIS calculations based on PitchBook data, outstanding direct private-credit loans to AI-related sectors increased from near zero in 2015 to more than $200 billion in 2025, while the share of private-credit loans to AI-related companies rose from less than 1 percent to almost 8 percent of total outstanding loan volumes.
This rising trend of borrowing means that should AI ultimately underperform, losses will be felt not only by equity investors, but also by a much broader investor base, including bondholders, private-credit funds and financial institutions.
Investors could feasibly attempt to reduce their dependence on the largest technology companies by pursuing various mitigating strategies such as equal-weighted indices, value investing, exposure to smaller companies, developed markets outside the United States, active stock selection and greater fixed-income exposure. But many of these approaches carry risks of their own—for instance, equal weighting raises exposure to smaller companies and increases rebalancing turnover; overseas allocations add currency and country risk; fixed income adds interest rate and credit risk; and active management brings selection risk and potentially higher associated management and performance fees.
Reducing concentration could also carry a persistent performance penalty in the meantime, as US technology leaders continue to outperform. As such, investors must weigh protection from a potential trend reversal against the cost of withdrawing from the companies presently driving market returns.
