I’m optimistic about the long-term potential of artificial intelligence (AI) to power big increases in worker productivity1 and economic growth. But I’m pessimistic that AI can justify this year’s rally in US share prices or ward off an economic slowdown this year or next.
As is often the case with new technology, it’s likely to take many years to realise the full potential of AI. While substantial benefits appear likely, a meaningful risk of disappointment remains.
In this article, I attempt to connect the dots between the current level of US share prices, an approaching slowdown in the US economy and the long-term promise of AI to command the world’s attention.
$1 trillion may be invested in AI, but not by the end of 2025
A common narrative in AI circles is that tech firms, utilities and other businesses will spend a combined $1 trillion or more to advance the technology in the coming years2. Such a sum may be spent, but it’s not going to happen by the end of next year, by which time we expect the US economy to have slowed. We estimate that AI-related spending would need to reach $1 trillion by the end of 2025 to push US economic growth in 2025 above the trend of roughly 2%.
As shown in the chart below, spending on AI totalled an estimated $67 billion in 2023. To estimate spending in 2024 and 2025, we looked at the rates of growth in AI-related spending over the last decade as well as the rates of spending growth on other types of technologies in their heydays. The chart shows that if spending on AI grows by 13% a year – as spending on software did between 1990 and 2000 – it would amount to $76 billion this year and $86 billion next year. If spending on AI grows 34% a year – as it did between 2013 and 2023 – it would amount to $90 billion this year and $121 billion next year.
Even if investment in AI nearly doubled this year and next – mirroring the near doubling of Nvidia’s data centre revenues in recent years – AI spending would amount to ‘only’ about $129 billion in 2024 and $248 billion in 2025. Those would be tremendous outlays. Perhaps unprecedented. But $1 trillion in AI investment by the end of 2025 would require 286% growth. That’s probably not going to happen, which means we’re unlikely to experience an AI-driven economic boom in 2025.
Investments in AI: Unprecedented but well shy of $1 trillion
Notes: The chart shows the amount that would be spent on AI in 2024 and 2025 if the rate of spending growth matched historical rates of spending on a range of technologies. Historical figures reflect compounded annual growth rates (i.e. growth for one year is added onto growth from the year before, and so on).
Sources: Vanguard; historical telecommunications and cloud spending data are from the US Federal Reserve’s FEDS Notes, Own-Account IT Equipment Investment, October 2017; Data for historical AI spending and estimate of actual 2023 spending in the United States are from Stanford University’s Artificial Intelligence Index Report 2024; Data for historical software spending are from the U.S. Bureau of Economic Analysis; and data for Nvidia revenue are from Ycharts.com.
Enthusiasm for AI may explain much of the recent love for US shares
We have been cautioning investors for some time that US shares are expensive, or ‘overvalued’. We base this on a measure called the ‘cyclically adjusted price-to-earnings ratio’ (CAPE), which shows share prices relative to 10-year average historical earnings.
As the chart below shows, the CAPE for US shares is about 32% above our own estimate of what is a ‘fair’ value for US shares, based on things like interest rates and inflation.
US shares remain expensive despite recent pullbacks
Past performance is not a reliable indicator of future results.
Notes: Our ‘fair value’ measure is based on a model that adjusts CAPE measures for the level of inflation and interest rates. It analyses company earnings, inflation and US bonds from 1 January 1940 to 5 August 2024. Details were published in the 2017 Vanguard research paper, Global Macro Matters: As US Stock Prices Rise, the Risk-Return Trade-off Gets Tricky.
Sources: Vanguard calculations, based on data as at 5 August 2024, from Robert Shiller’s website, the U.S. Bureau of Labor Statistics, the Federal Reserve Board, Refinitiv and Global Financial Data.
Profits would have to soar to stop shares looking overvalued
We wondered how fast company profits would have to grow to stop US shares looking overvalued. We calculated that to return to a ‘fair’ value in three years, profits would need to grow by about 40% a year. This is double the annualised growth rate3 of the 1920s, when electricity lit up the nation. And it’s significantly higher than the 4% average rate of growth since 1871.
With profit margins close to record highs, most of a hypothetical 40% jump in annualised profits would have to come from soaring company revenues. But slowing economic growth tends to prevent soaring sales. My team’s forecast of US economic growth in 2025 is 1%-1.5%, which is lower than our expectation of 2% growth this year.
Human intelligence remains irreplaceable
The promise of AI is real. Our research suggests that the odds of an AI-driven surge in workforce productivity are between 45% and 55%. In that scenario, we believe the US economy would grow at an annualised rate of about 3.1% between 2028 and 2040. The intervening years reflect the need for additional investments in the technology and time for them to pay off.
We also see meaningful risk – a 30% to 40% chance – that AI produces more modest benefits that are insufficient to offset the increase in government spending required to meet the needs of an ageing population. In that case, long-term economic growth might reach only about 1% a year.
Investors looking to connect the dots between the current level of share prices, probable levels of economic activity and the widespread enthusiasm for AI would be well-advised to temper any expectations that economic growth and company profits are set for near-term acceleration.
Instead, as ever, they’d be well-served to apply good sense in building and maintaining balanced portfolios that reflect their goals and attitude to risk. They should also be prepared to endure periodic downturns that would push share prices closer to their fair values.
1 Worker productivity is the output per worker or per hour worked.
2 See, for example, Goldman Sachs’ Gen AI: Too Much Spend, Too Little Benefit? (June 2024).
3 The annualised rate represents growth on a 12-month basis.
Investment risk information
The value of investments, and the income from them, may fall or rise and investors may get back less than they invested.
Past performance is not a reliable indicator of future results.
Important information
Vanguard only gives information on products and services and does not give investment advice based on individual circumstances. If you have any questions related to your investment decision or the suitability or appropriateness for you of the product[s] described, please contact your financial adviser.
This is designed for use by, and is directed only at persons resident in the UK.
The information contained herein is not to be regarded as an offer to buy or sell or the solicitation of any offer to buy or sell securities in any jurisdiction where such an offer or solicitation is against the law, or to anyone to whom it is unlawful to make such an offer or solicitation, or if the person making the offer or solicitation is not qualified to do so. The information does not constitute legal, tax, or investment advice. You must not, therefore, rely on it when making any investment decisions.
Issued by Vanguard Asset Management Limited, which is authorised and regulated in the UK by the Financial Conduct Authority.
© 2024 Vanguard Asset Management Limited. All rights reserved.
3968239