Artificial intelligence (AI) is proliferating across the economy, with applications ranging from hospital diagnostics to internet queries. The emergence of AI in so many areas raises the questions: Which industries spend the most on AI, and has that list changed as AI tools have become more widely available? In a recent Monthly Labor Review article, Bureau of Labor Statistics (BLS) economists tracked trends in <a href="https://www.bls.gov/opub/mlr/2026/article/ai-and-the-rise-of-software–investment.htm” rel=”nofollow noopener” target=”_blank”>aggregate business AI usage through software investment. In this week’s post, we use underlying industry-level estimates from the Bureau of Economic Analysis (BEA) to see which industries are leading in software investment in recent years.
The scale of the AI shift is hard to measure directly, but investment data may offer some insights. At its core, AI is software, and while firms don’t report AI spending as a line item, the BEA tracks software investment across industries as part of its capital stock data. The BEA also provides more detailed industry-level estimates, although they are considered less reliable than the higher-level aggregates. As a result, the findings we present below should be taken as preliminary estimates rather than an authoritative account.
In official government data, software investment is broken down into three main categories due to differences in how businesses record their expenditures in each category:
- Prepackaged software is software that can be purchased as is from other businesses, like the latest Windows operating system.
- Custom software is software that can be purchased like prepackaged software but further customized to meet the needs of the business.
- Own-account software is developed internally by a business rather than being purchased from elsewhere.
Accurately recording business expenses in each category can be challenging. For example, while prices and quantities of prepackaged software can be easy to measure, prices and quantities of own-account software are often not measured, and related expenditures have to be imputed from other data such as compensation costs for software programmers and systems analysts.
Using the BEA’s industry-level estimates, we examine real software investment across 43 private nonresidential industries from 2019 to 2024. As shown in Figure 1, software investment rose steadily across all industries, with the sharpest one-year increase coming between 2021 and 2022, coinciding with the first public release of a generative AI model. Also, investment in all three software categories grew.
