Measuring Spanish Corporate Tech Adoption: What Invested Capital Reveals

Saying a company has adopted a technology can refer to very different situations: buying a license, testing a tool, using it regularly, or integrating it into processes that change production. The purchase alone does not prove integration nor does it reveal whether the technology has a noticeable effect on the business.

This distinction matters when interpreting indicators. A tech capital figure describes resources associated with certain asset categories, whereas a measurement of corporate use would have to specify which organizations count as users and under what definition. They are not interchangeable measures: investment and habitual use answer different questions.

The CaixaBank Research analysis published on February 13, 2020, studies the economic relevance of capital associated with new technologies in Spain and other advanced countries. It does not offer a direct figure for Spanish companies using artificial intelligence. Its utility lies in showing how to approximate technological intensity when statistics do not directly quantify that specific technology.

To approximate the presence of emerging technologies, the study uses four capital categories from the EU KLEMS database: software and databases, research and development, computers, and telecommunications equipment. It combines them and relates them to hours worked. The result is expressed as constant 2010 euros of technological capital per hour worked.

This choice avoids presenting an AI usage figure that the data employed do not offer. In the article's framework, AI is too recent to have a comparable direct economic measure; therefore, the selected capital represents a broader set of assets that can contribute to developing or using digital technologies. It is an indirect indicator, not a measurement of installed algorithms, production projects, or business results.

The denominator also changes the reading: capital per hour worked does not equal the total spending of a country or sector. It allows comparing the relative intensity of technological capital in relation to labor, but it does not inform by itself of the number of companies investing, the size of each investment, or its profitability. It also does not determine by itself the degree of utilization of the accounted assets.

With the historical data analyzed, CaixaBank Research places Spain below the United States, France, and Germany in technological capital per hour worked, although it highlights that the measure grew rapidly in Spain during the two decades examined. The comparison describes a past trajectory, not a guarantee of convergence: the article suggested that Spain could approach France or Germany in about ten years if the observed trends were maintained.

The sectoral comparison for Spain covers 31 branches of industry and services; the study excludes agriculture and mining from that breakdown. Its authors identify differences between sectors with high and medium-low levels of technological capital, and relate the former to activities classified by Eurostat as high-tech or knowledge-intensive. The national average hides that heterogeneity: the indicator does not imply that all activities adopt technologies at the same pace.

The text also warns that machinery classification can hide technological components. As an example, it notes that some industrial robots are accounted for within general machinery categories, not necessarily in the items used to build its indicator. Therefore, a comparison based on accounting categories can incompletely represent physical technologies whose digital function is not identified separately.

The CaixaBank Research analysis observes accumulated technological capital and relates it to hours worked; it does not directly measure how many companies have incorporated a technology into their operations. Therefore, its results should not be presented as a corporate adoption rate. The type of data delimits the conclusion: the indicator allows speaking of capital intensity, not of use declared by companies.

The study is also not a market forecast. Its comparison is based on capital categories and historical data, whereas a projection on a specific technology may refer to a future scenario. To put two figures in context, one would have to check that they share period, population, technological definition, and unit of measure; the 2020 article, by itself, does not allow validating a commercial forecast.

When reading a claim about adoption, it is advisable to identify what it counts: organizations, employees, licenses, projects, spending, or accumulated capital. The reference year, included sectors, and whether the data distinguishes between a one-off test and habitual use are also important. Without that precision, a figure may describe a different phenomenon than the one it seems to measure.

The capital indicator does not allow deducing which technology a specific SME should buy. It does help understand a structural issue: the technological capacity of an economy depends on investment in complementary assets, such as software, data, equipment, and R&D, and that capacity is not distributed uniformly across sectors. The business decision must start from the process to be improved, not from a list of trends.

A 2025 study by the Universitat de València based on 30 interviews with companies in the Valencian Community that use AI examines how they move from intention to practice and what organizational frictions appear. Due to its method and scope, it provides qualitative experiences, not an estimate of the proportion of Spanish companies adopting AI. Organizational integration and the presence of capital are different issues.

To decide whether to test a solution, an SME can ask specific questions: what task changes? What data does it require? With which existing system must it connect? Who is responsible for its use and maintenance? A limited test can provide practical information on compatibility and costs. It does not prove, by itself, that the solution will be integrated on a larger scale or that its results will be replicated in other companies.

The CaixaBank Research work uses EU KLEMS and compares capital categories with historical data. Its results describe differences in capital intensity and trends of the period studied, but they do not directly measure AI adoption. To assess the impact of an investment on productivity, it would be necessary to relate both variables with an appropriate method, something different from observing the evolution of capital per hour worked.

There are also coverage limits: the cited sectoral breakdown includes industry and services, not agriculture or mining, and the accounting classification does not always identify technological assets with precision. The aggregate indicator can vary by sectoral composition and by the way capital is recorded. Furthermore, the data per hour worked does not equal a direct measurement of corporate use of a technology.

The useful conclusion is limited: the analysis documents that capital associated with digital technologies gained relevance in Spain in the observed period and that its weight differs between countries and sectors. To speak of current corporate adoption, use data with a clear reference date are needed. Capital, habitual use, and results are different levels; distinguishing them avoids turning investment or expectation into a claim about widespread implementation.