6 minEconomy
US Markets' Record AI Bet May Take Decades to Pay Off, History Suggests
AI-related spending has reached about 2% of US GDP and drives a historically concentrated share of growth and investment, but economic history and the slow spread of technology inside firms suggest the payoff will arrive far later than markets expect.
American investors have placed an unprecedented wager on artificial intelligence, and the concentration of that bet across stocks, credit and venture funding has few parallels in US financial history. The question now facing markets is not whether AI will matter, but whether it can deliver returns on the timetable investors have priced in. Five decades of economic history suggest it will not, at least not as quickly as the market assumes.
The scale of the commitment is striking. AI-related capital spending has climbed to roughly 2% of US gross domestic product, up from near zero five years ago. Research from the Federal Reserve Bank of St. Louis found that investment in information processing equipment, software, research and development, and data centers — four categories economists treat as the closest proxies for AI spending — accounted for 39% of US economic growth in the first three quarters of 2025. That exceeds the 28% contribution those activities made in 2000, at the height of the dot-com boom. Pantheon Macroeconomics calculates that without AI spending, US corporate investment in equipment was negative last year.
The comparison that looms largest is the Gilded Age. In 1885, railroad stocks made up more than 80% of active trading on the New York Stock Exchange, a concentration unmatched until now. When that bubble burst in the Panic of 1893, a third of railroads failed, more than 500 banks collapsed and the economy slid into a four-year depression. Today's AI buildout may be following a similar arc, though it is concentrated in far fewer companies than the railway mania ever was.
That concentration runs through every segment of US capital markets. The so-called Magnificent Seven — Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla — carry a combined market value of about $25 trillion, or 35% of the S&P 500, the highest share held by a handful of stocks since just before the 1929 crash. In credit markets, research from the Federal Reserve Bank of Kansas City shows AI-related companies' share of investment-grade bond issuance rising from 2% in 2023 to 10% in 2026. In venture capital, AI captured 81% of a record $300 billion in global funding in the first quarter of 2026, with just four companies — OpenAI, Anthropic, xAI and Waymo — absorbing two-thirds of the total.
The result is a self-reinforcing loop. Hyperscalers are set to spend $800 billion this year on chips, data centers and the power to run them, propping up GDP growth. Investors bid up AI stocks, enriching the top 20% of earners, who now account for roughly 60% of all US consumer spending, according to Moody's Analytics. As the buildout outgrows hyperscalers' cash flow, new investment is increasingly financed by debt and off-balance-sheet partnerships. The growth engine now runs on a narrow set of mutually reinforcing drivers, and if one stalls, the others lose power.
The Bank for International Settlements warned in its 2026 Annual Economic Report that AI exuberance risks ending in a «lengthy investment bust.» It flagged «circular financing» as a core vulnerability: more than half of investment in AI firms comes from other AI firms, whose recipients use the money to buy their backers' products. Profits, in effect, are being funded by investors rather than end customers.
The central assumption underpinning the bet is that by 2027, growth driven by infrastructure will hand off to growth driven by AI-powered efficiency gains across the wider economy. That handoff is likely to come much later than the market expects. Five decades of digital innovation achieved broad consumer adoption yet coincided with a decline in aggregate productivity growth — a pattern sometimes called the Innovation Paradox, in which more innovation produces slower growth. Consumers adopt new technology almost immediately, while enterprises move at a glacial pace, and within enterprises a handful of superstar firms embrace change quickly while the rest struggle.
The structure of the US economy compounds the problem. It has shifted from vertically linked industrial sectors, where innovations traveled naturally through supply chains, to horizontally spread service sectors that resist transformation because they depend on human judgment. These «stagnant» sectors, as economist William Baumol called them, now account for roughly half of GDP in advanced economies. Administrative and regulatory compliance costs add further drag.
Capital concentrated in AI-linked companies also starves the rest of the economy of the investment needed to become AI-ready. If hospitals, schools, logistics firms and retailers cannot invest in digital infrastructure and workforce retraining, the broad-based productivity gains that would justify current valuations will not materialize. AI could still defy the paradox: unlike earlier digital waves, it is a cognitive technology capable of reasoning and judgment, precisely what labor-intensive sectors need. But realizing that potential requires deep adoption across the economy, with firms allocating capital to workers and workflows, not just chips and data centers. If the bubble bursts, the assets financed by exuberant investors may still end up reshaping the economy for the better — as canals, railroads, electrification and the internet eventually did.
8
