China dominates upstream patenting of AI for the energy transition with filings in China and the US biased toward ICT and efficiency, while Japan leans industrial; corporate patent concentration in Europe, Japan and the US versus stronger public/university patenting in China and South Korea suggests an uneven, contested directionality with notable blind spots such as buildings.

Main Finding

Patent-landscape analysis of AI applications tied to the energy transition (the “Twin Transition”) shows concentrated upstream innovation activity in China and divergent orientation across jurisdictions: China and the United States emphasize AI in ICT and efficiency applications, Japan shows stronger industrial orientation, and the buildings sector is conspicuously under-represented. Corporations dominate patenting in Europe, Japan and the U.S., while universities and public institutions are more active patenters in China and South Korea. These patterns indicate an uneven and contested directionality of the Twin Transition rather than clear downstream technological leadership.

Key Points

• Methodological contribution: proposes an anticipatory framework to read patent data along four dimensions of directionality — geographical, actor, sectoral and functional — treating patents as upstream indicators of innovation orientation.
• Data source: European Patent Office (Espacenet) patent families filed in the five largest jurisdictions (China, United States, South Korea, Europe, Japan).
• Geographic concentration: a high concentration of patent families originates from China; the rest is spread among the U.S., South Korea, Europe and Japan.
• Actor differences: private corporations dominate patenting in Europe, Japan and the U.S.; universities and government/public institutions play a larger role in China and South Korea.
• Functional/sectoral orientation: AI applications for ICT and efficiency improvements are predominant in China and the U.S.; Japan shows relatively more industrial/manufacturing-oriented AI for the transition.
• Sectoral gaps: buildings (and some other demand-side sectors) are under-represented in patenting activity — an important directional finding pointing to potential blind spots in the Twin Transition.
• Interpretation caveat: patent concentration is an upstream signal of innovation activity and does not, on its own, prove technological leadership or downstream transition outcomes.

Data & Methods

• Data: Patent families sourced from the European Patent Office’s Espacenet database, filtered to the top five jurisdictions (China, United States, South Korea, Europe, Japan).
• Analytic frame: a tentative anticipatory framework that classifies patents across four dimensions to reveal directionality:
  • Geographical (where patenting activity is concentrated)
  • Actor (who is patenting: corporations, universities, government/public institutions)
  • Sectoral (which economic sectors are targeted: energy, transport, buildings, industry, etc.)
  • Functional (what functions the AI serves: ICT, efficiency improvements, process control, monitoring, etc.)
• Outcome focus: upstream innovation patterns rather than later-stage diffusion or real-world decarbonization outcomes.
• Limitations highlighted by the study:
  • Patents are an imperfect proxy for innovation (varying propensities to patent across firms, sectors, and jurisdictions).
  • Filing/jurisdictional patterns can reflect strategy, policy incentives or IP regimes, not strictly technological capability.
  • Temporal lags: patents capture invention activity that may precede commercial deployment by years.
  • The analysis emphasizes the Global North plus China and calls for broader geographic coverage in future work.

Implications for AI Economics

• Comparative advantage and industrial strategy: jurisdictional differences in actor mix and sector orientation suggest varying comparative advantages and the influence of national innovation systems — policies should be tailored to local actor structures (e.g., leveraging public research in China vs. firms in Europe/U.S.).
• R&D and investment targeting: the dominance of ICT/efficiency-focused AI in patenting signals where upstream R&D and private investment are flowing; policymakers aiming to accelerate sectoral decarbonization (e.g., buildings, heavy industry) may need targeted incentives to redirect R&D.
• Market structure and competition: corporate concentration in patenting in some jurisdictions points to potential market power in upstream AI-for-transition technologies; antitrust and IP policy will shape downstream competition and diffusion.
• Path dependency and lock-in risks: current patenting directionality can create technological lock-ins (e.g., ICT-centric solutions) that may bias the trajectory of transition technologies and neglect under-represented sectors.
• Labor, skills and distributional effects: differing sectoral focus (industry vs. ICT vs. buildings) implies different labor and skills demand across jurisdictions; anticipatory workforce policies and re-skilling programs should align with likely technological trajectories.
• Policy and measurement: patent landscape analysis is a useful early-warning tool for anticipating where innovation is headed, but should be combined with downstream measures (adoption, deployment, standards, complementary investments) for policy design.
• Research agenda: extend analysis beyond the Global North (and beyond a China-centric lens) to capture regional variation, incorporate complementary indicators (prototypes, deployments, publications, firm behavior), and connect upstream patent signals to downstream energy and economic outcomes.