US AI investment has become so concentrated across the economy that any delay in promised productivity gains could trigger a prolonged reckoning, according to an analysis of America’s technology-led growth model.
The central assumption behind the current boom is that spending on chips, data centres and computing power will give way to efficiency gains across businesses by 2027. But the analysis argues that the handover is likely to take decades rather than months or a few years.
AI-related capital spending has risen from almost nothing five years ago to about 2% of US gross domestic product. Investments in information-processing equipment, software, research and development and data centres accounted for 39% of economic growth in the first three quarters of 2025, the Federal Reserve Bank of St Louis reported in January.
That was higher than the 28% contribution made by comparable activities in 2000, at the height of the dotcom boom. Pantheon Macroeconomics has calculated that, without AI spending, US corporate investment in equipment would have fallen last year.
The concentration extends beyond business investment. The seven largest technology-linked companies – Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla – have a combined market value of $25 trillion, representing 35% of the S&P 500.
That is the highest concentration in a small group of shares since the period immediately before the 1929 market crash, according to the analysis.
AI investment spreads across US markets
AI is also becoming more prominent in credit markets. Research from the Federal Reserve Bank of Kansas City found that AI-related companies’ share of investment-grade bond issuance rose from 2% in 2023 to 10% in 2026.
Venture capital has followed the same pattern. AI attracted 81% of a record $300 billion in global venture funding during the first quarter of 2026, with OpenAI, Anthropic, xAI and Waymo accounting for two-thirds of the total.
Meanwhile, the largest cloud and technology companies are expected to spend $800 billion this year on semiconductors, data centres and the electricity needed to operate them. That spending is supporting economic growth while helping to drive share prices higher.
The rise in AI stocks has increased the wealth of the highest-earning households, who are responsible for a large share of consumer demand. Moody’s Analytics estimates that the top 20% of US earners account for about 60% of all consumer spending.
As companies’ infrastructure plans outstrip their cash flow, more of the expansion is being funded through borrowing and partnerships outside their balance sheets. The result is a tightly connected system in which investment, markets, household spending and economic growth are increasingly reliant on the same theme.
The Bank for International Settlements has warned that enthusiasm for AI could end in a “lengthy investment bust”. It has also identified “circular financing” as a key weakness, with more than half of investment in AI companies coming from other AI companies, whose funding is then used to purchase products from their backers.
The long road to productivity gains
The analysis argues that the difficulty will be spreading AI beyond a small group of leading firms. Consumers tend to adopt new technologies quickly, while businesses move more slowly, and even within companies the most advanced operators can be far ahead of the rest.
That challenge is particularly acute in service industries that depend heavily on human judgement. Such sectors account for roughly half of GDP in advanced economies and have proved resistant to rapid productivity improvements.
Administrative and regulatory compliance costs are also said to be restricting companies’ ability to adopt new technology. At the same time, the concentration of investment in AI-linked businesses may leave hospitals, schools, logistics firms and retailers without enough capital to upgrade their digital systems or retrain staff.
AI could ultimately overcome that problem because it is a cognitive technology capable of reasoning and making judgements. But the potential benefits will depend on widespread adoption, with businesses investing in employees and working practices as well as in processors and data centres.
Previous investment booms in canals, railways, electricity, telecommunications and the internet ended in painful collapses, but the infrastructure created during those periods supported later economic expansion. Data centres, semiconductors and AI models could have a similar long-term legacy.
However, the analysis argues that such benefits would emerge over decades. With economic growth, consumer spending, equity markets, debt markets and venture capital all increasingly tied to AI, a delay in the expected payoff could expose the US economy to an unusually broad and concentrated shock.
