How CaixaBank Research measures the penetration of new technologies in Spain

Published on 13 February 2020, CaixaBank Research’s study asks an economic question: how significant are technologies associated with the future in Spain’s economy, and how might their significance relate to growth? It does not count businesses using artificial intelligence; instead, it constructs a measure of technological capital per hour worked.

The meaning of “penetration” depends on the indicator used. In this analysis, it refers to the intensity of capital associated with new technologies in the economy. It does not directly show what proportion of companies has deployed a particular tool, how many workers use it, or what results each deployment achieves.

The study describes changes in capital associated with new technologies over the preceding two decades and compares countries and sectors using EU KLEMS data. Its findings cover that period and are not a measure of technology adoption in 2026.

To estimate the development of technologies that lacked a direct economic measure, the authors use EU KLEMS data on capital stock. They group four categories: software and databases, research and development, computers, and telecommunications equipment. The selection includes digital assets as well as items that can support their creation and operation.

The sum of these categories is divided by hours worked in each country or sector. The result is expressed in constant 2010 euros of capital associated with new technologies per hour worked. This normalisation makes it possible to compare technological capital intensity across economies and industries, and with other types of capital, such as transport or machinery.

According to the article, artificial intelligence was a relatively new phenomenon for which no data existed to quantify it in economic terms. The authors therefore use broader capital categories related directly or indirectly to technologies that may contribute to its development. This is not an AI-specific measure: some included assets may be used for purposes unrelated to AI.

In the study’s international comparison, the United States had the highest stock of technological capital per hour worked among the advanced economies analysed. France and Germany were in the middle, while Spain and the United Kingdom were at lower levels. The article described the growing importance of these technologies in every country during the period studied, although countries started from different positions.

The analysis also highlighted Spain’s rate of growth. Based on that historical trend, the authors suggested that Spain could reach France’s or Germany’s levels in about ten years if the observed pace continued. This was a conditional projection published in 2020, not confirmation that convergence has since occurred.

The study itself noted a limitation: some technologies may be classified under broader capital categories. CaixaBank Research said that some robots used in Germany could be recorded as “other machinery”, outside the technological categories selected for its measure. As a result, the indicator may not fully capture certain assets when their statistical classification does not match the analysis’s definition.

For Spain, the article compares 31 industry and service sectors and finds differences in their stock of capital associated with new technologies. The sectors with the highest intensity generally corresponded to high-technology industries and knowledge-intensive services under Eurostat’s classification. The sectoral breakdown excludes agriculture and mining.

This correspondence does not, by itself, show that greater technological capital causes more advanced production. Eurostat’s classification takes into account factors including research and development spending, the technological content of goods and services, and patents; R&D spending is also included in the study’s capital measure. The classifications share some elements, so their association should not be presented as independent causal evidence.

The 2020 article reported no convergence between the sector groups with the highest and lowest technological intensity during the period examined. That conclusion is limited to the data analysed at the time: it does not establish whether the gap persisted afterward or describe the effects of technologies that spread later.

Capital per hour worked cannot answer questions such as what percentage of businesses uses AI, which company sizes adopt it most, or what their main obstacles are. Those questions require business surveys with a defined population and reference period. More aggregate capital does not automatically mean more businesses use AI: the indicators measure different things.

For example, Spain’s National Statistics Institute (INE) reported that 21.1% of businesses with ten or more employees used AI in the first quarter of 2025, while 44.3% paid for cloud computing services. These percentages describe use reported in a business survey; they do not update the EU KLEMS series for capital per hour worked.

In October 2026, the INE reported that 33.6% of businesses with ten or more employees used AI in the first quarter of that year. The institute cautioned that, from reference year 2025, the change to the CNAE-2025 classification and the inclusion of additional activities in the survey’s scope mean the figures are not strictly comparable with earlier years. The date and definition are part of the statistic and should accompany any comparison.

The study makes it possible to compare countries or sectors by technological capital per hour worked, but it does not measure installed capacity or actual use in a specific business. Whether to deploy a tool depends on an organisation’s needs and circumstances; this macroeconomic indicator cannot replace that internal assessment.

The methodology provides a common measure for comparing technological capital intensity relative to hours worked across countries or sectors. Its interpretation depends on the categories included and how assets are recorded in each industry. It is a tool for structural comparison, not an assessment of a company’s performance or an automatic forecast of growth.

Estimated capital stock and survey-reported adoption should be kept separate: the first describes assets grouped into an economic measure; the second, the share of businesses that say they use particular technologies. CaixaBank Research’s analysis provides the first kind of measure, while the INE survey reports the second for specific periods.