Artificial intelligence driving billions in turnover for energy firms

Artificial intelligence is driving major economic gains in the energy sector, with new evidence pointing to billions in turnover and tens of thousands of new jobs in Romania. A study published in Mathematics provides one of the most detailed measurements to date of how AI adoption reshapes both firm-level performance and the national economy.

The research, titled A Machine Learning and Econometric Framework for Credibility-Aware AI Adoption Measurement and Macroeconomic Impact Assessment in the Energy Sector, introduces a credibility-aware framework to separate genuine adoption from hype. By combining natural language processing, econometric modeling, and input–output analysis, the study estimates that AI contributed more than 3.5 percent of Romania’s GDP in 2023 through the energy sector alone.

How is AI adoption measured in the energy sector?

The authors address a long-standing problem: how to measure AI adoption in industries where claims are frequent but verification is scarce. To solve this, they design a three-layered framework.

First, they construct an AI Adoption Score using natural language processing on scientific papers and media coverage, assigning scores to energy firms based on the intensity of AI-related content. Second, they introduce a Misinformation Bias Score, which penalizes exaggerated or low-credibility reporting. This includes checks for speculative language, subjective statements, and lack of data-backed claims, using modern classification and named-entity recognition tools. Finally, they apply an econometric difference-in-differences model to estimate the causal impact of adoption on firm turnover, scaling the results up to the national economy with a Leontief input–output framework.

The analysis focuses on 36 Romanian energy companies under NACE category 35, primarily electricity producers and traders, covering about 87 percent of sector turnover. By combining firm-level data with macroeconomic multipliers, the authors create a comprehensive picture of AI’s economic footprint.

What are the firm-level and national impacts?

The study’s econometric results show a striking boost to company performance. Using fixed-effects difference-in-differences, the authors estimate that around 42.8 percent of turnover in AI-adopting energy firms in 2023 can be linked directly to AI-related activity. This translates to approximately 19.94 billion lei in revenues across the sector.

When these firm-level impacts are aggregated through Romania’s input–output tables, the macroeconomic effects become even clearer. The study finds that AI adoption in the energy sector accounted for 3.54 percent of national GDP in 2023. This figure reflects not only direct sectoral output but also indirect supply-chain effects and induced impacts from household spending generated by new income.

The employment effects are equally significant. While the energy sector itself may see some displacement of traditional roles due to automation, the overall economy benefits from spillovers. The researchers estimate a net gain of around 65,000 jobs, with the majority created indirectly through supplier industries and consumer demand. This suggests that the broader labor market impact of AI can be positive if spillovers outweigh local disruptions.

What are the risks, limits, and policy implications?

Despite the positive findings, the study points out several caveats. The sample is relatively small, though it covers the vast majority of turnover in Romania’s energy industry. The reliance on public text sources for adoption scores means that results depend heavily on the quality and credibility of reporting, which the misinformation bias adjustment seeks to correct. Moreover, the econometric model is based on two time periods, limiting insight into longer-term dynamics.

The input–output scaling assumes stable input structures, which may not hold in rapidly changing energy markets. And while the study highlights net job creation, it also acknowledges displacement risks within energy firms themselves, underscoring the need for workforce transition policies.

From a policy perspective, the findings stress that credible measurement of AI adoption is crucial. Governments and regulators should invest in frameworks that discount hype and prioritize verifiable evidence of implementation. Supporting better reporting standards and transparency can reduce misinformation risks. At the same time, AI investment should be paired with worker retraining and social safety nets to mitigate short-term disruptions.

The authors argue that strategic sectors like energy are critical testbeds for linking firm-level adoption to macroeconomic outcomes. By integrating machine learning with econometric and macroeconomic tools, the framework offers a model for other industries and countries seeking to assess the real impact of AI.