Has Artificial Intelligence Made Market Concentration Less Risky

No. Artificial intelligence (AI) has changed the shape of market concentration more than its substance. Leadership has expanded beyond the largest technology platforms into semiconductors, infrastructure, industrials, and utilities, but many of those winners remain tied to the same AI capex cycle. As a result, the market may look broader on the surface, while still being unusually exposed to a narrow set of companies and a single dominant narrative.

That concentration risk is visible in the structure of the global equity market. US equities now account for roughly 64% of the MSCI All Country World Index, up from 42% in 2010, while the top ten US companies make up about 25% of the benchmark. Information technology holds a peak 37% share of the S&P 500, above late-1990s levels. These figures show that market leadership remains unusually concentrated, even as leadership has spread to a wider set of AI-linked beneficiaries.

The earliest winners, semiconductors and hyperscalers, were joined in 2024 by memory producers, data center providers, digital infrastructure firms, electrical equipment manufacturers, industrial companies tied to power build-out, and some utilities. Recent results underscore this expansion. Both Samsung and SK Hynix benefited from rising demand for high-bandwidth memory and related AI-linked memory products in first quarter 2026, and electrical equipment suppliers reported strong order growth linked in part to data center demand. Utilities and power-related firms have also participated as investors seek exposure to AI-driven electricity demand. AI remains a capital-intensive build-out that ties a growing share of market leadership to the same investment cycle.

The hyperscalers remain central to that cycle and may be more vulnerable than markets assume. Markets have not fully rerated hyperscalers for a business model that is becoming far more capital intensive. Combined capex for Alphabet, Amazon, Meta, Microsoft, and Oracle is estimated at roughly $760 billion in 2026, with cumulative property, plant, and equipment (PP&E) potentially approaching $2 trillion by 2030. On a five-year depreciation schedule, that translates to about $400 billion of annual depreciation expense, roughly equal to their combined 2025 profits. Markets may be underestimating the earnings growth and monetization required to justify such capital intensity, particularly in a market segment where enthusiasm and fear of missing out have often pushed equity prices ahead of fundamentals. Leadership that depends on companies becoming more asset-heavy, financing-dependent, and execution-sensitive is less secure than it appears.

If AI economics disappoint or investment slows, the impact could ripple across multiple market segments that now appear diversified. Essentially, hyperscaler capex is driving much of the market earnings and price momentum in ways that make the entire market edifice dependent on spending by five hyperscalers. For example, Empirical Research Partners finds that a basket of 48 large-cap stocks benefiting from AI spending has outperformed the broad market by 174 percentage points since the start of 2024. Spanning semiconductors and related equipment, capital equipment, metals, utilities, and tech hardware, the basket has accounted for 42% of market returns over the last 12 months and is expected to contribute nearly half the market's earnings growth this year. While leadership has broadened in one sense, continued strong performance is dependent on the capital spending of a narrow set of companies. And if their capex should continue at such a frenetic pace, it becomes more challenging for these companies to earn the high returns on invested capital (ROICs) that market valuation multiples demand.

There is also an important distinction within the expanding beneficiary set. Some adjacent beneficiaries, especially at the intersection of AI and electrification, may prove more resilient than pure AI plays. Grid modernization, transmission, electrical equipment, and certain utilities enjoy support from multiple demand drivers, including industrial electrification, energy security, and broader infrastructure needs. AI helps these areas, but it is not their sole source of support, making their return outlook less tightly bound to AI monetization than that of hyperscalers or direct AI plays.

The practical implication is that investors should focus on how many truly distinct drivers of return a portfolio contains. When a single secular story drives equity concentration, capital spending, credit issuance, infrastructure demand, and venture enthusiasm at the same time, the case for diversification becomes stronger, not weaker. Investors need not reject AI, but they should recognize that the better long-term opportunity may lie in markets where expectations are lower, valuations are less demanding, and portfolios are less dependent on a single story. AI has made concentration more diffuse and stealthier, but it has not made it less risky.