The reported 1.5% figure is an estimate of AI’s effect on the level of U.S. GDP over ten years—not a prediction that the economy will grow 1.5% a year. The figure comes from a Microsoft-published essay by Nobel laureate economist Daron Acemoglu, as reported by The Decoder; the essay’s exact text was not available to verify independently. Acemoglu’s 2024 academic estimate, summarized by the White House Council of Economic Advisers, puts the ten-year U.S. GDP-level impact at 0.9% to 1.6%.
What does the 1.5% forecast mean?
It describes a potential increase in the size of the U.S. economy after ten years relative to a scenario without the relevant AI gains. It is not an annual GDP growth rate, and it is not a claim that GDP will rise by 1.5% every year. The Decoder reported the 1.5% figure as Acemoglu’s forecast in a Microsoft-published essay, but the essay itself was not available to confirm its precise wording or assumptions. The Decoder’s report is therefore the source for that specific 2026 claim.
In his 2024 paper, The Simple Macroeconomics of AI, Acemoglu estimates a 0.93% to 1.16% increase in the U.S. GDP level over ten years under one investment assumption. The White House Council of Economic Advisers summarizes the paper’s estimate as a 0.9% to 1.6% impact on the U.S. GDP level over that period. These are modeled estimates, not measured results. Acemoglu’s 2024 paper and the CEA’s 2026 report provide the academic and official comparison.
Why does Acemoglu expect a limited aggregate effect?
Acemoglu’s analysis asks how AI changes the allocation of production tasks between workers and capital, including digital tools and algorithms. It distinguishes automation, in which technology takes over tasks, from task complementarity, in which it helps people perform tasks. A tool can reduce the cost or time required for a particular task without producing a similarly large change in economy-wide output.
The paper focuses on the United States because much of the evidence it draws on about AI’s effects and which tasks are exposed comes from the U.S. economy. Its macroeconomic estimates concern aggregate productivity and output, while task-level improvements are treated as cost savings that do not automatically translate into equivalent GDP gains.
Acemoglu cautions in the 2024 paper: “AI will have implications for the macroeconomy, productivity, wages and inequality, but all of them are very hard to predict.” The estimate should be read as a conditional forecast, not a settled account of what AI will deliver.
How does the estimate compare with other forecasts?
The CEA’s 2026 comparison shows how widely estimates differ. Its table labels the listed figures as impacts on GDP levels, except for one identified study. The estimates differ in geography, horizon, and underlying assumptions, so they should not be treated as like-for-like predictions.
| Source and year | Reported estimate | Geography and horizon |
|---|---|---|
| Acemoglu (2024), as summarized by the CEA | 0.9%–1.6% GDP-level impact | United States; ten years |
| Penn Wharton (2025) | 1.5% GDP-level impact | United States; ten years |
| Oxford Economics (2024) | 1.8%–4% GDP-level impact | Horizon of eight years; geography not stated in the CEA table |
| McKinsey (2023) | 2.4%–4.1% GDP-level impact | Long run; geography not stated in the CEA table |
| Goldman Sachs (2023) | 7% GDP-level impact | Global; ten years |
The comparison comes from the CEA’s 2026 report. Because the scope and time periods vary—and forecasts may cover different technologies and adoption paths—the larger numbers do not simply refute Acemoglu’s estimate. The CEA table identifies its figures as GDP-level impacts, while the Decoder’s account of Acemoglu’s Microsoft essay reports a ten-year 1.5% GDP impact.
Rank #3
Why do AI forecasts vary so much?
The European Central Bank’s March 2026 review describes the macroeconomic literature as spanning modest to transformative outcomes. Three important drivers of that spread are how quickly firms adopt AI, how much economic activity the technology can affect, and whether AI meaningfully accelerates innovation. Forecasts can also differ in whether they count generative AI alone or a broader set of AI and automation effects.
Early evidence from individual tasks helps explain why the debate is difficult, but it does not settle the economy-wide question. The ECB review cites one writing experiment in which time spent fell 40% and output quality rose 18%, and a customer-support deployment in which issues resolved per hour rose 15%. Those are results for specific tasks and settings, not estimates of national GDP growth. The ECB’s review discusses the gap between task-level evidence and aggregate projections.
Quick Recap
What should readers take from the headline?
- The reported 1.5% is a ten-year GDP-level estimate, not 1.5% annual growth.
- Acemoglu’s underlying 2024 academic estimate is about the United States, not the global economy.
- The forecast is modeled and uncertain; task-level productivity gains alone do not establish an equivalent aggregate GDP effect.
- The 2026 essay’s specific wording and assumptions remain unverified here, so its 1.5% figure should be attributed to The Decoder’s report rather than presented as a directly checked quotation from Acemoglu.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
