Periodic Reporting for period 1 - LIBRA (Leveraging artificial Intelligence to Balance tRade-offs in the digital economy)
Berichtszeitraum: 2024-05-01 bis 2026-04-30
Zusammenfassung vom Kontext und den Gesamtzielen des Projekts
The LIBRA project aims at quantifying the sustainability trade-offs of investments in artificial intelligence (AI). On one hand, the project assesses how AI supports sustainability objectives. On the other hand, the action computes the environmental costs of the AI economy. The ultimate goal is to strike an optimal balance to support innovation and growth without hampering the planet’s health. The proposed research approach goes beyond the state of the art in three aspects: framework, method and policy-related. First, LIBRA bridges economics of innovation, finance and sustainability sciences in a novel manner. The digital revolution for sustainable development is at core of an action agenda to achieve a transformed sustainable society. As recent literature flagged, economists have so far contributed very little to the expanding alignment area of AI in decision theory. LIBRA fills that gap by bringing a fresh and novel perspective to the literature on policy optimisation with DRL as efficient and replicable approach to solve decisions in the policy context. Third, LIBRA contributes to the ongoing and growing discussion about the impacts of AI on sustainability and the clean transition. Research has proved AI as a key enabler of environmental innovation, but there is a wide gap about trade-offs between costs and benefits of the digital economy, which LIBRA fills. The findings of my project will support actions at European level (i.e. the EU AI Act and the Digital Decade policy programme in particular) and internationally (i.e. OECD.AI Policy Observatory) well beyond the duration of the project itself. Since the award (May 2024), research and action in the sustainable AI domain has been evolving towards increased awareness of the impacts of this technology on the environment. The LIBRA project has helped closing the research-policy gap in times when parliaments around the globe, the United Nations AI Advisory Board and industry associations are in ferment to approve legally requirements or to publish good practices for the sector to grow responsibly.
Arbeit, die ab Beginn des Projekts bis zum Ende des durch den Bericht erfassten Berichtszeitraums geleistet wurde, und die wichtigsten bis dahin erzielten Ergebnisse
Mapping the AI economy: an empirical framework for analysing AI investment, industrial structure and environmental materiality
On the research side, LIBRA has managed to develop a consistent workflow to investigate the structure, evolution and potential environmental implications of the global AI economy. Rather than focusing exclusively on technological development, the workflow reconstructs AI as a complex socio-economic system characterised by interconnected investment flows, industrial specialisation, geographical concentration and environmental externalities. By integrating firm-level financial information, investment portfolios, institutional indicators and environmental materiality metrics, the methodology provides a comprehensive empirical framework for understanding how AI development interacts with broader sustainability challenges.
The analysis begins by constructing a harmonised database of AI-related firms from multiple investment datasets.
Progress: moving beyond the DoA, this workflow reconstruct the AI economy temporally, spatially and per segment. Individual company records are integrated, standardised and classified according to technological domains, industrial sectors and geographical location, creating a consistent representation of the AI economy. This harmonisation is a critical prerequisite for analysing investment patterns because AI firms often operate across multiple technological fields and industrial applications. Establishing a unified analytical database therefore enables the identification of structural trends that would remain hidden within fragmented datasets. From the perspective of sustainable development, this stage contributes primarily to Sustainable Development Goal (SDG) 9 – Industry, Innovation and Infrastructure, by enabling the systematic mapping of emerging innovation ecosystems and the infrastructure supporting the digital transformation.
Once the analytical database has been established, the workflow reconstructs the temporal evolution of AI investment by examining fundraising activities (from Prequin, figure above), company creation, technological specialisation and geographical distribution. Rather than describing isolated firms, the analysis reveals how financial resources have progressively concentrated within specific countries, industries and AI technology domains. This longitudinal perspective provides important evidence on the geography of technological innovation and on the emergence of regional AI ecosystems. Understanding where AI investment is concentrated also provides indirect insights into where future computational infrastructure, data centres and supporting digital assets are likely to expand, thereby establishing an empirical foundation for assessing future energy demand and associated environmental pressures.
The most distinctive contribution of the analytical workflow lies in the integration of environmental materiality indicators with investment and industrial datasets. This final stage moves beyond conventional financial analysis by explicitly linking patterns of AI investment to measures of direct environmental impact. Financial indicators alone provide limited information regarding the broader societal implications of AI development. For this reason, environmental materiality indicators are incorporated into the analytical framework in order to evaluate how patterns of investment correspond to environmentally relevant industrial activities. Integrating these dimensions enables the identification of AI technology domains that combine rapid financial growth with potentially significant environmental implications. Rather than treating economic growth and sustainability as separate dimensions, the workflow establishes an empirical bridge between technological expansion and environmental performance. Linking financial investment to environmental materiality creates the foundations for identifying those AI technology domains that combine rapid economic growth with potentially higher environmental relevance. This component directly supports SDG 12 – Responsible Consumption and Production, by facilitating the identification of environmentally material production systems, while simultaneously contributing to SDG 13 – Climate Action, through the development of evidence base capable of supporting future assessments of AI-related greenhouse gas emissions, resource consumption and climate risks.
Progress: provide a scalable methodological framework capable of supporting future sustainability analyses of AI. The current workflow deliberately focuses on reconstructing the economic and industrial architecture of AI rather than directly estimating emissions or resource consumption. Nevertheless, by integrating investment dynamics, industrial classifications, institutional indicators and environmental materiality information within a single empirical framework, the methodology establishes the foundations upon which additional environmental variables can subsequently be incorporated. Future extensions may integrate electricity consumption, carbon emissions, water use, critical mineral requirements, semiconductor supply chains or lifecycle assessment data, thereby transforming the present analytical framework into a comprehensive platform for evaluating the sustainability of AI-driven economic development.
The analytical framework presented so far reconstructs the financial, technological, geographical and institutional architecture of the AI economy, thereby providing an empirical map of where AI development is taking place and which technological domains exhibit the greatest environmental materiality. An alternative strategy was pursued with respect to the DoA due to data limitations beyond my control and responsibilities. However, understanding the sustainability implications of AI (Specific objectives 2) requires moving beyond the geography of investment towards the economic mechanisms through which investment translates into physical infrastructure, industrial restructuring and market concentration. These mechanisms constitute the focus of the subsequent analytical framework, which investigates how capital expenditure, mergers and acquisitions, and industrial concentration collectively drive the physical expansion of the AI economy and, ultimately, its environmental footprint.
Empirical framework for analysing the physical expansion of the AI economy: integrating capital formation, infrastructure investment, industrial concentration and sustainability indicators
Building upon the investment landscape reconstructed in the previous analytical framework, the present workflow investigates the economic mechanisms through which financial investment is transformed into productive assets, industrial restructuring and digital infrastructure. Rather than asking where AI investment is concentrated, this framework examines how the AI economy expands through capital formation, infrastructure investment, mergers and acquisitions, and increasing market concentration. Together, the two analytical frameworks provide a comprehensive methodology for understanding both the architecture and the physical evolution of the AI economy, thereby establishing the foundations for future assessments of its environmental footprint.
The analytical framework is based on the premise that the sustainability implications of artificial intelligence cannot be understood solely by analysing algorithms or computational performance. Instead, AI must be examined as a rapidly expanding industrial ecosystem supported by continuous investment in physical infrastructure, including data centres, semiconductor manufacturing, cloud computing facilities, communication networks and other capital-intensive assets. These investments determine the future scale of computational capacity while simultaneously shaping electricity demand, material consumption, critical mineral requirements and associated greenhouse gas emissions. Consequently, understanding capital formation and industrial restructuring represents a fundamental prerequisite for evaluating the environmental sustainability of AI development.
To investigate these processes, the workflow integrates multiple firm-level financial datasets (Prequin, Pitchbook and Orbis), market information and transaction databases into a unified empirical framework. Capital expenditure data are employed to reconstruct long-term investment in productive assets, while market capitalisation provides additional evidence regarding the economic scale of firms operating within software, hardware and artificial intelligence industries. The framework subsequently incorporates information on mergers and acquisitions to examine how firms expand not only through organic investment but also through corporate consolidation, technological acquisition and industrial integration. Finally, network analysis and market concentration indicators are used to characterise the evolving structure of the AI economy, identifying sectors that occupy increasingly strategic positions within the broader digital ecosystem. This progression from capital formation to industrial concentration represents the principal contribution of the framework. Capital expenditure is interpreted as a proxy for the physical expansion of AI infrastructure, reflecting sustained investment in computing facilities, digital assets and production capacity. Mergers and acquisitions (from Orbis) reveal how firms acquire technological capabilities, consolidate markets and reorganise industrial ecosystems. Network analysis further illustrates how relationships between acquiring and target sectors evolve over time, while concentration indicators provide evidence regarding the emergence of increasingly central firms and industries within the AI economy. Collectively, these analytical components reveal the structural mechanisms through which AI evolves from financial investment into tangible economic and technological infrastructure.
Progress: From a sustainability perspective, the framework establishes an essential bridge between economic development and environmental assessment. Although the analysis does not directly estimate electricity consumption, carbon emissions or water use, it identifies the economic drivers that will ultimately determine these environmental outcomes. Persistent increases in capital expenditure indicate continued investment in infrastructure requiring substantial quantities of construction materials, semiconductors, advanced electronics and computing equipment. Similarly, increasing industrial concentration may accelerate the deployment of hyperscale computing infrastructure and concentrate future resource demand among a relatively small number of firms. Consequently, the framework provides an empirical basis for anticipating where future environmental pressures are most likely to emerge before direct environmental impacts become measurable. Technically, once firm-level investment patterns have been reconstructed, the workflow identifies those organisations that consistently sustain high levels of capital expenditure over time. Rather than focusing exclusively on firms exhibiting temporary investment peaks, the analysis distinguishes organisations characterised by persistent infrastructure investment, thereby identifying the actors most likely to drive the long-term expansion of AI infrastructure. Comparative analyses are subsequently performed across software, hardware and AI-intensive firms. This distinction is particularly important because artificial intelligence depends simultaneously upon advances in software development, semiconductor production and computational infrastructure. Examining these sectors together reveals how the physical expansion of AI relies upon the interaction between complementary technological domains rather than the growth of isolated industries. The resulting evidence provides a more comprehensive understanding of the industrial foundations of AI and contributes to SDG 9 by improving knowledge of digital innovation systems while strengthening the evidence base supporting SDG 13 (Climate Action) through the identification of sectors expected to generate increasing energy demand and infrastructure expansion.
While capital expenditure captures organic investment in productive assets, firms also expand through corporate acquisitions and strategic mergers. The third stage therefore reconstructs the evolution of mergers and acquisitions (M&A) across software, hardware and AI-related firms by integrating firm-level financial information with corporate transaction databases. Analysing acquisition activity provides important insights into how technological capabilities, productive assets and market positions become consolidated over time. Acquisitions frequently accelerate the deployment of AI technologies by enabling firms to obtain complementary expertise, computational resources and intellectual property without relying exclusively on internal development. Consequently, corporate transactions represent an additional mechanism through which AI infrastructure expands and industrial ecosystems evolve. From the perspective of sustainable development, analysing corporate restructuring contributes to SDG 8 (Decent Work and Economic Growth) by examining the structural transformation of knowledge-intensive industries while simultaneously informing industrial policy aimed at promoting innovation and long-term competitiveness. The fourth stage extends the analysis beyond individual firms to examine the structural organisation of the AI economy. Rather than treating acquisitions as isolated events, corporate transactions are represented as interconnected industrial networks linking acquiring sectors with target sectors. This network perspective enables the identification of industries occupying increasingly influential positions within the AI ecosystem and reveals how technological capabilities become distributed across interconnected production systems.
To complement this relational perspective, market concentration indicators (including centrality measures and the Herfindahl-Hirschman Index (HHI)) are calculated to evaluate the degree to which productive capacity and corporate control become concentrated among relatively small numbers of firms or sectors. These indicators provide quantitative evidence regarding the evolution of industrial concentration, revealing whether the AI economy is becoming progressively dominated by a limited number of highly interconnected actors.
This analytical stage has important sustainability implications. Increasing industrial concentration may accelerate the deployment of hyperscale computing infrastructure, concentrate electricity demand and intensify competition for strategic resources such as advanced semiconductors and critical minerals. Consequently, analysing market concentration contributes not only to understanding industrial organisation but also to anticipating future patterns of resource consumption and environmental pressure. This stage therefore supports SDG 8, SDG 9 and SDG 16 (Peace, Justice and Strong Institutions) by generating evidence relevant to competition policy, industrial governance and resilient innovation ecosystems.
The framework therefore contributes explicitly to several SDGs. By reconstructing patterns of capital formation and infrastructure investment, it supports SDG 9 (Industry, Innovation and Infrastructure) through improved understanding of the physical assets underpinning digital innovation. Analysing firm growth, mergers and acquisitions and industrial restructuring contributes to SDG 8 (Decent Work and Economic Growth) by examining the mechanisms through which technological industries evolve and generate economic transformation. Interpreting capital expenditure as an indicator of infrastructure expansion provides an important contribution to SDG 12 (Responsible Consumption and P
roduction) because it identifies future demand for construction materials, semiconductors, critical minerals and other resource-intensive inputs required to sustain AI development. The framework also directly supports SDG 13 (Climate Action) by establishing the empirical foundations necessary for future assessments of electricity demand, greenhouse gas emissions and climate impacts associated with expanding AI infrastructure. Finally, the integration of heterogeneous financial, industrial and corporate datasets within a unified analytical architecture exemplifies the collaborative data integration envisaged under SDG 17 (Partnerships for the Goals), demonstrating how complementary sources of evidence can be combined to support interdisciplinary sustainability research.
Limits with respect to specific objectives in DoA: Importantly, this framework should not be interpreted as a final assessment of the environmental footprint of AI. Rather, it represents the second stage of a broader research programme. The first analytical framework reconstructed the investment geography, technological composition and environmental materiality of the AI economy. The present framework explains the economic mechanisms through which these investment patterns materialise into physical infrastructure and industrial transformation. A natural third stage will build upon both frameworks by integrating direct environmental indicators (including electricity consumption, greenhouse gas emissions, water use, critical mineral dependency and lifecycle environmental impacts) to quantify the environmental footprint of AI systems.
On the research side, LIBRA has managed to develop a consistent workflow to investigate the structure, evolution and potential environmental implications of the global AI economy. Rather than focusing exclusively on technological development, the workflow reconstructs AI as a complex socio-economic system characterised by interconnected investment flows, industrial specialisation, geographical concentration and environmental externalities. By integrating firm-level financial information, investment portfolios, institutional indicators and environmental materiality metrics, the methodology provides a comprehensive empirical framework for understanding how AI development interacts with broader sustainability challenges.
The analysis begins by constructing a harmonised database of AI-related firms from multiple investment datasets.
Progress: moving beyond the DoA, this workflow reconstruct the AI economy temporally, spatially and per segment. Individual company records are integrated, standardised and classified according to technological domains, industrial sectors and geographical location, creating a consistent representation of the AI economy. This harmonisation is a critical prerequisite for analysing investment patterns because AI firms often operate across multiple technological fields and industrial applications. Establishing a unified analytical database therefore enables the identification of structural trends that would remain hidden within fragmented datasets. From the perspective of sustainable development, this stage contributes primarily to Sustainable Development Goal (SDG) 9 – Industry, Innovation and Infrastructure, by enabling the systematic mapping of emerging innovation ecosystems and the infrastructure supporting the digital transformation.
Once the analytical database has been established, the workflow reconstructs the temporal evolution of AI investment by examining fundraising activities (from Prequin, figure above), company creation, technological specialisation and geographical distribution. Rather than describing isolated firms, the analysis reveals how financial resources have progressively concentrated within specific countries, industries and AI technology domains. This longitudinal perspective provides important evidence on the geography of technological innovation and on the emergence of regional AI ecosystems. Understanding where AI investment is concentrated also provides indirect insights into where future computational infrastructure, data centres and supporting digital assets are likely to expand, thereby establishing an empirical foundation for assessing future energy demand and associated environmental pressures.
The most distinctive contribution of the analytical workflow lies in the integration of environmental materiality indicators with investment and industrial datasets. This final stage moves beyond conventional financial analysis by explicitly linking patterns of AI investment to measures of direct environmental impact. Financial indicators alone provide limited information regarding the broader societal implications of AI development. For this reason, environmental materiality indicators are incorporated into the analytical framework in order to evaluate how patterns of investment correspond to environmentally relevant industrial activities. Integrating these dimensions enables the identification of AI technology domains that combine rapid financial growth with potentially significant environmental implications. Rather than treating economic growth and sustainability as separate dimensions, the workflow establishes an empirical bridge between technological expansion and environmental performance. Linking financial investment to environmental materiality creates the foundations for identifying those AI technology domains that combine rapid economic growth with potentially higher environmental relevance. This component directly supports SDG 12 – Responsible Consumption and Production, by facilitating the identification of environmentally material production systems, while simultaneously contributing to SDG 13 – Climate Action, through the development of evidence base capable of supporting future assessments of AI-related greenhouse gas emissions, resource consumption and climate risks.
Progress: provide a scalable methodological framework capable of supporting future sustainability analyses of AI. The current workflow deliberately focuses on reconstructing the economic and industrial architecture of AI rather than directly estimating emissions or resource consumption. Nevertheless, by integrating investment dynamics, industrial classifications, institutional indicators and environmental materiality information within a single empirical framework, the methodology establishes the foundations upon which additional environmental variables can subsequently be incorporated. Future extensions may integrate electricity consumption, carbon emissions, water use, critical mineral requirements, semiconductor supply chains or lifecycle assessment data, thereby transforming the present analytical framework into a comprehensive platform for evaluating the sustainability of AI-driven economic development.
The analytical framework presented so far reconstructs the financial, technological, geographical and institutional architecture of the AI economy, thereby providing an empirical map of where AI development is taking place and which technological domains exhibit the greatest environmental materiality. An alternative strategy was pursued with respect to the DoA due to data limitations beyond my control and responsibilities. However, understanding the sustainability implications of AI (Specific objectives 2) requires moving beyond the geography of investment towards the economic mechanisms through which investment translates into physical infrastructure, industrial restructuring and market concentration. These mechanisms constitute the focus of the subsequent analytical framework, which investigates how capital expenditure, mergers and acquisitions, and industrial concentration collectively drive the physical expansion of the AI economy and, ultimately, its environmental footprint.
Empirical framework for analysing the physical expansion of the AI economy: integrating capital formation, infrastructure investment, industrial concentration and sustainability indicators
Building upon the investment landscape reconstructed in the previous analytical framework, the present workflow investigates the economic mechanisms through which financial investment is transformed into productive assets, industrial restructuring and digital infrastructure. Rather than asking where AI investment is concentrated, this framework examines how the AI economy expands through capital formation, infrastructure investment, mergers and acquisitions, and increasing market concentration. Together, the two analytical frameworks provide a comprehensive methodology for understanding both the architecture and the physical evolution of the AI economy, thereby establishing the foundations for future assessments of its environmental footprint.
The analytical framework is based on the premise that the sustainability implications of artificial intelligence cannot be understood solely by analysing algorithms or computational performance. Instead, AI must be examined as a rapidly expanding industrial ecosystem supported by continuous investment in physical infrastructure, including data centres, semiconductor manufacturing, cloud computing facilities, communication networks and other capital-intensive assets. These investments determine the future scale of computational capacity while simultaneously shaping electricity demand, material consumption, critical mineral requirements and associated greenhouse gas emissions. Consequently, understanding capital formation and industrial restructuring represents a fundamental prerequisite for evaluating the environmental sustainability of AI development.
To investigate these processes, the workflow integrates multiple firm-level financial datasets (Prequin, Pitchbook and Orbis), market information and transaction databases into a unified empirical framework. Capital expenditure data are employed to reconstruct long-term investment in productive assets, while market capitalisation provides additional evidence regarding the economic scale of firms operating within software, hardware and artificial intelligence industries. The framework subsequently incorporates information on mergers and acquisitions to examine how firms expand not only through organic investment but also through corporate consolidation, technological acquisition and industrial integration. Finally, network analysis and market concentration indicators are used to characterise the evolving structure of the AI economy, identifying sectors that occupy increasingly strategic positions within the broader digital ecosystem. This progression from capital formation to industrial concentration represents the principal contribution of the framework. Capital expenditure is interpreted as a proxy for the physical expansion of AI infrastructure, reflecting sustained investment in computing facilities, digital assets and production capacity. Mergers and acquisitions (from Orbis) reveal how firms acquire technological capabilities, consolidate markets and reorganise industrial ecosystems. Network analysis further illustrates how relationships between acquiring and target sectors evolve over time, while concentration indicators provide evidence regarding the emergence of increasingly central firms and industries within the AI economy. Collectively, these analytical components reveal the structural mechanisms through which AI evolves from financial investment into tangible economic and technological infrastructure.
Progress: From a sustainability perspective, the framework establishes an essential bridge between economic development and environmental assessment. Although the analysis does not directly estimate electricity consumption, carbon emissions or water use, it identifies the economic drivers that will ultimately determine these environmental outcomes. Persistent increases in capital expenditure indicate continued investment in infrastructure requiring substantial quantities of construction materials, semiconductors, advanced electronics and computing equipment. Similarly, increasing industrial concentration may accelerate the deployment of hyperscale computing infrastructure and concentrate future resource demand among a relatively small number of firms. Consequently, the framework provides an empirical basis for anticipating where future environmental pressures are most likely to emerge before direct environmental impacts become measurable. Technically, once firm-level investment patterns have been reconstructed, the workflow identifies those organisations that consistently sustain high levels of capital expenditure over time. Rather than focusing exclusively on firms exhibiting temporary investment peaks, the analysis distinguishes organisations characterised by persistent infrastructure investment, thereby identifying the actors most likely to drive the long-term expansion of AI infrastructure. Comparative analyses are subsequently performed across software, hardware and AI-intensive firms. This distinction is particularly important because artificial intelligence depends simultaneously upon advances in software development, semiconductor production and computational infrastructure. Examining these sectors together reveals how the physical expansion of AI relies upon the interaction between complementary technological domains rather than the growth of isolated industries. The resulting evidence provides a more comprehensive understanding of the industrial foundations of AI and contributes to SDG 9 by improving knowledge of digital innovation systems while strengthening the evidence base supporting SDG 13 (Climate Action) through the identification of sectors expected to generate increasing energy demand and infrastructure expansion.
While capital expenditure captures organic investment in productive assets, firms also expand through corporate acquisitions and strategic mergers. The third stage therefore reconstructs the evolution of mergers and acquisitions (M&A) across software, hardware and AI-related firms by integrating firm-level financial information with corporate transaction databases. Analysing acquisition activity provides important insights into how technological capabilities, productive assets and market positions become consolidated over time. Acquisitions frequently accelerate the deployment of AI technologies by enabling firms to obtain complementary expertise, computational resources and intellectual property without relying exclusively on internal development. Consequently, corporate transactions represent an additional mechanism through which AI infrastructure expands and industrial ecosystems evolve. From the perspective of sustainable development, analysing corporate restructuring contributes to SDG 8 (Decent Work and Economic Growth) by examining the structural transformation of knowledge-intensive industries while simultaneously informing industrial policy aimed at promoting innovation and long-term competitiveness. The fourth stage extends the analysis beyond individual firms to examine the structural organisation of the AI economy. Rather than treating acquisitions as isolated events, corporate transactions are represented as interconnected industrial networks linking acquiring sectors with target sectors. This network perspective enables the identification of industries occupying increasingly influential positions within the AI ecosystem and reveals how technological capabilities become distributed across interconnected production systems.
To complement this relational perspective, market concentration indicators (including centrality measures and the Herfindahl-Hirschman Index (HHI)) are calculated to evaluate the degree to which productive capacity and corporate control become concentrated among relatively small numbers of firms or sectors. These indicators provide quantitative evidence regarding the evolution of industrial concentration, revealing whether the AI economy is becoming progressively dominated by a limited number of highly interconnected actors.
This analytical stage has important sustainability implications. Increasing industrial concentration may accelerate the deployment of hyperscale computing infrastructure, concentrate electricity demand and intensify competition for strategic resources such as advanced semiconductors and critical minerals. Consequently, analysing market concentration contributes not only to understanding industrial organisation but also to anticipating future patterns of resource consumption and environmental pressure. This stage therefore supports SDG 8, SDG 9 and SDG 16 (Peace, Justice and Strong Institutions) by generating evidence relevant to competition policy, industrial governance and resilient innovation ecosystems.
The framework therefore contributes explicitly to several SDGs. By reconstructing patterns of capital formation and infrastructure investment, it supports SDG 9 (Industry, Innovation and Infrastructure) through improved understanding of the physical assets underpinning digital innovation. Analysing firm growth, mergers and acquisitions and industrial restructuring contributes to SDG 8 (Decent Work and Economic Growth) by examining the mechanisms through which technological industries evolve and generate economic transformation. Interpreting capital expenditure as an indicator of infrastructure expansion provides an important contribution to SDG 12 (Responsible Consumption and P
roduction) because it identifies future demand for construction materials, semiconductors, critical minerals and other resource-intensive inputs required to sustain AI development. The framework also directly supports SDG 13 (Climate Action) by establishing the empirical foundations necessary for future assessments of electricity demand, greenhouse gas emissions and climate impacts associated with expanding AI infrastructure. Finally, the integration of heterogeneous financial, industrial and corporate datasets within a unified analytical architecture exemplifies the collaborative data integration envisaged under SDG 17 (Partnerships for the Goals), demonstrating how complementary sources of evidence can be combined to support interdisciplinary sustainability research.
Limits with respect to specific objectives in DoA: Importantly, this framework should not be interpreted as a final assessment of the environmental footprint of AI. Rather, it represents the second stage of a broader research programme. The first analytical framework reconstructed the investment geography, technological composition and environmental materiality of the AI economy. The present framework explains the economic mechanisms through which these investment patterns materialise into physical infrastructure and industrial transformation. A natural third stage will build upon both frameworks by integrating direct environmental indicators (including electricity consumption, greenhouse gas emissions, water use, critical mineral dependency and lifecycle environmental impacts) to quantify the environmental footprint of AI systems.
Fortschritte, die über den aktuellen Stand der Technik hinausgehen und voraussichtliche potenzielle Auswirkungen (einschließlich der bis dato erzielten sozioökonomischen Auswirkungen und weiter gefassten gesellschaftlichen Auswirkungen des Projekts)
On the research side, LIBRA has managed to develop a consistent workflow to investigate the structure, evolution and potential environmental implications of the global AI economy. Rather than focusing exclusively on technological development, the workflow reconstructs AI as a complex socio-economic system characterised by interconnected investment flows, industrial specialisation, geographical concentration and environmental externalities. By integrating firm-level financial information, investment portfolios, institutional indicators and environmental materiality metrics, the methodology provides a comprehensive empirical framework for understanding how AI development interacts with broader sustainability challenges.
Concretely, the framework contributes to the following SDGs:
SDG Contribution of the empirical framework
SDG 8 Analyses capital formation, firm growth and industrial restructuring generated by the AI economy.
SDG 9 Maps investment in AI infrastructure, technological capabilities and industrial innovation ecosystems.
SDG 12 Interprets capital expenditure as a proxy for infrastructure expansion and future material demand, including semiconductors, servers and critical minerals.
SDG 13 Provides an empirical basis for estimating future electricity demand, carbon emissions and climate impacts associated with AI infrastructure.
SDG 17 Integrates heterogeneous datasets (PitchBook, Orbis, peer groups and financial statements) into a unified analytical framework, demonstrating the value of data integration for sustainability research.
Concretely, the framework contributes to the following SDGs:
SDG Contribution of the empirical framework
SDG 8 Analyses capital formation, firm growth and industrial restructuring generated by the AI economy.
SDG 9 Maps investment in AI infrastructure, technological capabilities and industrial innovation ecosystems.
SDG 12 Interprets capital expenditure as a proxy for infrastructure expansion and future material demand, including semiconductors, servers and critical minerals.
SDG 13 Provides an empirical basis for estimating future electricity demand, carbon emissions and climate impacts associated with AI infrastructure.
SDG 17 Integrates heterogeneous datasets (PitchBook, Orbis, peer groups and financial statements) into a unified analytical framework, demonstrating the value of data integration for sustainability research.