AI Adoption in the EU: From 7.6% to 20% in Four Years, According to Funcas

The Funcas working paper published in May 2026 reveals a structural shift in artificial intelligence adoption within the European Union. The rate of companies using AI rose from 7.6% in 2021 to 20% in 2025, an increase the institution analyzes through the lens of technology diffusion theory. This jump is not uniform, but it marks the transition from a phase of experimental curiosity to operational integration in the business processes of a significant portion of the European business fabric. The consolidation of this metric suggests that the technology has surpassed the initial testing stage to become a functional component of corporate strategy.

It is crucial to distinguish between measured diffusion and perception of adoption. While Funcas surveys capture effective use reported by companies, other sources like the INE or Deloitte offer higher figures for Spain (21.1% and 61% respectively), suggesting a possible methodological gap or greater penetration in the Iberian market. The discrepancy between the 20% EU average and the 61% Spanish figure indicates that adoption is not homogeneous and that indicators must be interpreted with caution depending on the source and the 'use' criterion applied. This variability forces analysts to contextualize each data point within its specific reference framework before drawing general conclusions about the state of the industry.

Gartner's forecasts for 2026, presented at the IT Symposium/Xpo 2025, position AI as the central axis of digital transformation. Among the ten strategic trends, native AI platforms and AI supercomputing stand out, combining CPUs, GPUs, and ASICs for intensive workloads. Gartner estimates that more than 40% of leading companies will adopt hybrid computing architectures by 2028, validating the investment trend observed in the last four years. This projection reinforces the idea that specialized computing infrastructure is becoming an indispensable strategic asset to maintain competitiveness in global markets.

For its part, MIT Technology Review has identified generative AI search and the integration of domain-specific language models (DSLM) as key emerging technologies. These tools allow smaller teams to develop applications with greater speed, reducing the technical entry barrier. The convergence of these predictions with Funcas adoption data suggests that the market is responding to the availability of more accessible and specialized infrastructures, rather than to disruptive innovation from scratch. The accessibility of these platforms is democratizing the development of intelligent solutions, allowing smaller organizations to compete in areas that previously required massive R&D resources.

Data from the Bank of Spain (EBAE 2025) shows that nearly 20% of Spanish companies use AI systems, a rate lower than Germany's but higher than Italy's. However, the Deloitte report cited by the Artificial Intelligence Magazine states that 61% of Spanish companies have already implemented AI solutions, placing Spain seven points above the European average (54%). This contradiction evidences the complexity of measuring adoption: while the Bank of Spain measures the use of 'AI systems' in a strict sense, Deloitte might include low-threshold generative AI tools. The difference in the definition of what constitutes an 'AI solution' is the determining factor that explains the breadth of this statistical gap between the two main sources.

In sectoral terms, adoption is higher in technology services and in large, productive, and young companies. Banking, insurance, and retail sectors are driving demand due to competitive pressure, where AI has moved from being a differential advantage to a survival requirement. In contrast, SMEs show much lower penetration, at 11% according to ONTSI, compared to 44% for large companies, highlighting the scale gap in investment and talent capacity. This structural disparity indicates that organization size remains a stronger predictor of adoption than the sector of activity itself, at least in the current phase of technological maturity.

Despite the growth, adoption barriers persist. According to Zoho, 33% of companies cite the lack of talent as the main obstacle. The ECLAC report on Latin America, although geographically distinct, reflects a similar structural barrier: low investment in intangible capital and the scarcity of qualified talent limit AI expansion. In Spain, 47% of respondents wish to support generative AI adoption but admit they do not have the necessary knowledge, indicating a critical skills gap. This lack of specialized profiles not only slows implementation but also increases the opportunity cost for companies that rely on external consultants to manage their artificial intelligence projects.

Other barriers include data quality, governance, and regulation. The European AI Regulation, in force since February 2025, introduces compliance obligations that can slow implementation in companies without robust governance infrastructure. Furthermore, the lack of a clear strategy is a recurring problem: 12% of decision-makers in Spain implemented AI without adequate planning, leading to ineffective integrations and results that do not meet return on investment expectations. Regulatory complexity requires a restructuring of internal workflows to ensure compliance, a process that requires time and resources that many organizations have not yet budgeted for in their annual plans.

There is a tension between actual adoption measured in surveys and future investment expectations. While 20% of the EU already uses AI, Gartner and MIT forecasts point to an acceleration in 2026-2028 driven by physical AI and multi-agent systems. This discrepancy suggests that current adoption is only the tip of the iceberg, and that most of the economic value will be generated in the scaling and optimization phase, not in the initial implementation. Future investment will be oriented less towards license acquisition and more towards the deep integration of these systems into the value chain, seeking operational efficiencies that justify the spending made in recent years.

The sustainability of this growth will depend on companies' ability to overcome talent and governance barriers. The democratization of tools like ChatGPT or Copilot has lowered the entry barrier, but the deepening of use requires investment in intangible capital and continuous training. Without active training policies and sectoral strategies, the risk is that adoption stalls in superficial pilots, incapable of generating real productivity at scale. Long-term success will not be measured by the number of tools installed, but by the organizations' ability to transform that data into actionable and measurable strategic decisions.