This article was first published in Barron’s on 23 February 2026.

The market seems to designate a new loser of AI every day. The latest losers are software companies, which sold off sharply after Anthropic released its latest Claude Cowork model in January. That sell-off spread to the alternative asset managers that finance those companies. More recently, wealth managers, insurance brokers, and property service firms have come under scrutiny for their vulnerability to AI.

Investors might be anxious about how AI is quickly threatening these entrenched industries. But the flip side of the coin is that AI might restore competition and revive innovation within an economy that has steadily become more concentrated for decades.

Competition in the US has been declining for the past 40 years. According to the US Federal Reserve, the share of employment accounted for by newly formed firms fell 43% between 1980 and 2016. Since the late 1990s, more than three-quarters of US industries have become more concentrated. Only health care and utilities have avoided increases in top-four market share, with many industries experiencing double-digit gains.

Several forces drove this shift. Persistently low real interest rates in the early 2000s and 2010s fueled private equity and enabled well-capitalized firms to acquire smaller competitors. Since the 1980s, Supreme Court doctrine has held that mergers are anti-competitive only if they reduce consumer surplus — an interpretation that ushered in decades of lax antitrust enforcement. At the same time, advances in computing strengthened the economics of scale, reinforcing the advantages of incumbents.

Rising concentration has been a boon for US equities. Investors pay premium multiples for predictable, defensible cash flows. As Warren Buffett famously once said, he looks for “economic castles protected by unbreachable moats.”

Less competition has also supported higher profit margins. But profitability and productivity are not the same. Research shows that firms experiencing the largest increases in product-market concentration saw higher profit margins, but no corresponding gains in operational efficiency. In other words, higher profits did not reflect better production; they reflected greater pricing power.

Indeed, productivity — output per hour worked — is often negatively correlated with profitability and stock performance. Competition in oil and gas produced the shale revolution, dramatically increasing output, but equity returns disappointed. Competition among telecom firms built a robust 5G network, but at the expense of the firms’ margins.

Yet productivity is the key to long-term real wage growth. When output per worker rises, companies can afford to pay workers more. Unfortunately, productivity growth since 2010 has been roughly one-third lower than in the prior three decades. Slower productivity growth ultimately means slower improvements in living standards, and rising concentration is part of the explanation.

AI could change that trend. AI lowers barriers to entry across a range of industries. It reduces the advantages of scale in knowledge work and allows smaller firms to compete more effectively. At the same time, hyperscalers are moving beyond their traditional oligopolies in search, social media, and cloud computing. They are pouring capital into data centers and competing intensely to lead in model development.

Capital investment is the foundation of productivity growth. One estimate suggests hyperscalers could spend roughly 2.1% of GDP on capital expenditures in 2026 — a pace reminiscent of the railroad boom in the mid-19th century. The railroad analogy is instructive. Railroads transformed interstate commerce and fueled US economic expansion. But they also produced boom-bust cycles, excess capacity, and intense volatility. Capitalists may celebrate growth, but they dislike competition. Railroad executives lobbied for exclusive charters, tax exemptions, and land grants, and formed “pools” to divide traffic and fix rates.

We should expect similar dynamics today. Tech executives have already floated the idea of government support for AI infrastructure. While some have walked back explicit calls for bailouts, those arguments are likely to resurface if financing conditions tighten or fiscal stimulus fades. Increasing merger activity should also be expected as incumbents seek to neutralize emerging competitors.

Productivity can rise in two ways: by producing more with the same workforce, or by producing the same output with fewer workers. Their economic consequences differ dramatically. The former supports rising wages, stronger growth, and broader prosperity. The latter risks widening inequality and weaker real wage gains.

Whether AI becomes a force for broad-based prosperity or deeper concentration will depend largely on whether policymakers preserve competition. Encouraging entry, resisting bailouts and excessive consolidation may compress margins and equity multiples, reversing a multi-decade trend. But it could also restore dynamism, lift productivity, raise real wages, and strengthen long-term economic growth. Investors may fear competition. The economy needs it.