Overview
Artificial intelligence is everywhere, but the extent to which companies and governments invest in AI remains elusive. Despite frequent headlines, political speeches, and corporate press releases touting major AI funding commitments, these figures often lack transparency, consistency, and independent verification.
The noise around AI investment data arises from the diverse nature of reported figures. Some numbers reflect actual expenditures, while others represent soft commitments, budget allocations, or multi-year political plans. Furthermore, AI investment data often conflate different scopes, sometimes only including research and development (R&D), while at other times encompassing broader uses such as military AI. This variation makes it challenging to distinguish between real financial flows and forecasts.
A new framework to measure AI investment
On 26 September 2025, the OECD and the European Commission released Advancing the measurement of investments in artificial intelligence, a harmonised measurement framework to quantify AI spending across the European Union (EU) and benchmark it against leading economies, such as the United States, the United Kingdom, Canada, and Japan.
In this framework, AI is considered a general-purpose technology, much like electricity or the internet, whose impact extends far beyond software development. Accordingly, the report defines AI investment broadly, encompassing research and development (R&D) as well as complementary assets such as skills, data infrastructure, computing hardware, and organisational capital.
To estimate AI investment, the methodology applies a two-step approach designed to isolate genuine AI-related spending from broader ICT or R&D expenditures. It combines:
- Macroeconomic data from sources like Eurostat and the EUKLEMS & INTANProd databases.
- AI intensity coefficients, derived from patent data and education statistics, to identify what share of these investments can reasonably be attributed to AI.
In line with our definition of investment, the study identifies areas where resources are directed to foster AI development, uptake and impact. These areas are grouped into four categories and ten specific investment items (Table 1).
To estimate the AI-related share for each investment category, a set of specific intensity coefficients were developed. These coefficients are constructed as shares and serve as multipliers for the aggregate expenditure data. These coefficients proxy the AI-specific portion of each investment category. Each investment item is matched with a coefficient that best captures its AI intensity.
Table 1. Overview of investment categories and sources.
| Category | Investment Item | AI intensity coefficient applied |
| Skills | ICT specialist compensation | % AI ICT specialists in country’s total number of ICT specialists |
| Academic teacher compensation | % of AI university programmes in country’s total programmes | |
| Corporate training | % of AI patents in country’s total number of patents | |
| Other intellectual property products | Organisational capital | |
| Brand | ||
| Design | ||
| R&D | Research & Development | % of AI patents in country’s total number of patents AND % of AI publications in country’s total number of publications |
| Data and equipment | Computer hardware | % of AI patents in country’s total number of ICT patents |
| Computer software and databases | ||
| Telecommunications equipment |
Baseline estimates
The baseline analysis indicates that the aggregate amount of investments in AI across the EU in 2023 was 257 billion euros. Private sector investment in AI significantly outweighs public investment across the EU27, accounting for approximately 73% of total AI investments in 2023. In absolute terms, private sector AI investments reached 188 billion euros, compared to 69 billion euros from the public sector. These data can be explored further in Figure 1.
AI investment over time
Estimates of AI investment were also created over history, from 2015 to 2023, excluding ICT specialist compensation and Academic teacher compensation due to data limitations. These estimates reveal the rapid dynamics of AI investment over time.
While earlier years (2015-2018) experienced modest but steady growth, the post-2019 era marked a significant acceleration in AI investments across all categories. These data can be explored in Figure 2.
AI investment for selected third countries
The measurement framework was also extended to compare AI investments in the EU with those in select third countries: the United States, the United Kingdom, Canada, and Japan. Due to data constraints and considerations of statistical consistency, the comparative analysis is necessarily limited to key investment categories, specifically R&D expenditures and investments in Data and equipment. The data show that in AI R&D investment, the United States demonstrates exceptional commitment, approximately 90 billion euros – nearly an order of magnitude higher than Germany, Japan, and the United Kingdom, which follow with an allocation of approximately 10 billion euros each. These data can be further explored in Figure 3.
Future work
The newly agreed upon 2025 System of National Accounts sets out clear priorities for statistical agencies to estimate AI investment. However, the techniques for this measurement are still under consideration, making expansions and improvements to this work valuable.
There are several expansions to this work currently being considered. First steps include updating these estimates for 2024 and 2025 when the underlying data becomes fully available. Geographical expansions to relevant non-EU economies, such as OECD Member States, will enable global comparisons and a deeper understanding of relative investment dynamics.
Other improvements, such as the refinement of the intensity coefficients, specifying the import and export content, and more detailed sectoral estimates could be investigated to provide further value and guidance.