AI can support sustainability—from cutting emissions to improving public health—but it also poses ethical and environmental risks. This article explores AI’s potential to advance progress while stressing the need for responsible design, energy efficiency, and governance. It also advocates for a national agenda to guide and accelerate AI’s role in sustainable development. This article provides a summary of conceptualizations and findings derived from a series of other articles by the same author, all originating from the “AI for Sustainability” project.
Climate Warnings
Last year was the warmest on record, with global temperature 1.18 °C above the 20th-century average. Staying within 1.5–2.0 °C requires global emissions to fall 7.5% annually—roughly 4 billion tons globally and 3 million tons in Sweden. Sweden’s Climate Law (2017) commits to halving emissions by 2030 and reaching net-zero by 2045, or risk fines up to 20 billion SEK. Yet in Q2 2024, emissions rose 5.6%, when national goals demand annual reductions of 12–15%—levels never witnessed before. AI can be a critical accelerator in this context. AI can help reversing this trend, but public debate remains scant.
Despite this urgency, Sweden lacks a national AI strategy focused on climate action. We argue that a National Agenda for AI and Sustainability is needed to coordinate innovation, investment, and regulation—rather than continuing with siloed initiatives. It must involve broad collaboration across government, industry, academia, finance, civil society, public sector and local authorities. Many compelling initiatives—like WASP and OECD’s program—lack clear climate profiles. Sweden’s “AI Commission’s Roadmap (SOU 2025:12)” also falls short on how AI can support environmental goals. International efforts—like the AI Action Summit (2025), where 58 countries (including Sweden) endorsed a statement on inclusive and sustainable AI for planet and the people—show growing recognition of AI’s sustainability role. Swedish actors—AI Sweden, RISE, Stockholm Resilience Centre, Vinnova, and WWF, among others—are making progress, but not yet at sufficient scale.
The AI for Sustainability Promise
AI has moved from niche automation to a general-purpose technology reshaping industries. Generative AI is expanding the boundaries of knowledge and productivity. The global AI market is projected to reach $253 billion by 2033, driven by advances in machine learning and adoption across industries.
Sweden has strengths to build on—innovation capacity, advanced technology, and high digital maturity. Yet it invests just 5.50 SEK per capita in AI, compared to 18,300 SEK in France. Sweden also ranks only 25th globally in AI readiness (France is 5th, the US 1st). Failing to accelerate national AI efforts, Sweden not only risks falling further behind in the global race of taking leadership in AI adoption, but also achieving sustainability targets.
As a strategic initiative, Sweden could lead through AI adoption. One promising area is AI for Sustainability: the use of AI to support sustainable development across environmental, social, and economic dimensions. This includes, but not limited to, strategic initiatives for:
• Optimizing energy in buildings and grids
• Predicting emissions in supply chains
• Enhancing agriculture via satellite and
sensor data
• Managing climate risks in finance and cities
• Improving public health and biodiversity
monitoring
• Using digital twins to optimize crop yields
If implemented successfully, our research shows that AI could reduce emissions in Sweden by 4–5 million tonnes CO₂e annually by 2045—around 10% of the national target—and deliver climate-related benefits worth 110–170 billion SEK over two decades. Table 1 illustrates a sectoral breakdown of these benefits, investment needs, and reductions. Non-climate benefits—such as improved efficiency and health gains—could contribute another 90–120 billion SEK in societal value. In total, AI for Sustainability in Sweden could yield 200–310 billion SEK over two decades. Yet only a fraction is being realized. Current deployments reduce just 0.5–1 million tonnes CO₂e and generate 10–20 billion SEK in value per decade—under 10% of the potential.
Realizing climate-related benefits alone requires 53–83 billion SEK in investment. Capturing non-climate benefits would need another 27–46 billion SEK. Combined, the return on investment) is estimated at 1.5 to 3.6 times the cost—making a strong economic and environmental case for action.
“A National Agenda for AI and Sustainability is needed to coordinate innovation, investment, and regulation”
Why Progress Remains Slow
Despite its potential, AI’s use for sustainability remains limited due to:
• Fragmented data systems – Poor data quality and siloed infrastructures limit scale and impact.
• Lack of coordination and investment – Sweden trails behind other countries in AI funding, and efforts are scattered without national direction.
• Low public and investor trust – Green-washing and vague ESG claims erode confidence in AI-driven sustainability initiatives.
• Regulatory uncertainty – Absence of clear, climate-specific AI policies delays both innovation and adoption
• Talent gaps– Few professionals combine expertise in AI, sustainability, and systems thinking.
• AI’s own environmental footprint – Energy-intensive models risk offsetting sustainability gains.
A national agenda is required to overcome these barriers and coordinate actors, align investments, and unlock AI’s full potential for sustainable development. However, to reap this potential, analysist and institutions must account for the sustainability of AI itself.
Table 1: Sweden’s Projected Climate AI Benefits by Sector. Cross-sector Infra & Skills” refers to enabling investments in digital infrastructure, data systems, and workforce development that support AI deployment across multiple sectors. These are foundational but do not directly reduce emissions. Shares (%) of total emissions are reported in parentheses in the rightmost column.
Sustainable AI
Sustainable AI refers to the development and deployment of AI technologies in an environmentally and socially responsible manner across their life cycle. It emphasizes energy efficiency, ethical design, and reducing environmental harm. To some extent, sustainable AI is supported by EU frameworks:
• The European Green Deal (2019) calls for “sustainability-by-design.”
• The EU AI Act (2021) emphasizes safety and transparency but includes limited environmental criteria.
• The Corporate Sustainability Reporting Directive (CSRD, 2023) mandates disclosure of environmental and social impacts, including emissions from digital infrastructure.
Despite these frameworks, emissions from AI systems remain largely unregulated and poorly tracked. Data centers—key to AI—are a fast-growing source of GHG emissions due to high electricity use, which accounts for 90% of their emissions. According to the IEA (2025), global data centers consume 415 TWh of electricity per year, generating 180 million tonnes of CO₂e (about 0.5% of global emissions). This footprint may triple by 2030, mainly due to rising AI workloads. In Sweden, data centers consume 3 TWh annually (about 2.5% of total electricity use), resulting in approximately 0.15–0.18 million tonnes of CO₂e, depending on the electricity mix. If current trends continue, demand could exceed 10 TWh by 2040—potentially generating 0.5–0.6 million tonnes of CO₂e annually, even with a relatively clean energy grid.
To minimize AI’s environmental footprint, Sweden must adopt several strategic actions that go beyond current proposals—supplementing the AI Commission’s Roadmap:
• Accelerate energy-efficient AI models, including frugal algorithms and model compression to reduce computational demands.
• Invest in green infrastructure, such as renewable-powered data centers to support AI workloads with clean energy.
• Implement lifecycle assessments and mandatory emissions reporting to monitor and reduce environmental impacts.
These efforts must be supported by clearly defined, shared responsibilities across all sectors. Industry must lead by prioritizing energy-aware design and transitioning to clean power sources. Governments play a key role in enforcing carbon performance standards for data centers and enabling the decarbonization of national electricity grids. Financial institutions must fund climate-aligned innovation and ensure transparent tracking of environmental performance. Civil society must hold institutions accountable by demanding transparency, responsibility, and measurable sustainability outcomes. Such actions can help ensure that AI supports—rather than undermines—environmental goals.
To conclude, Sweden can lead in adopting AI solutions for sustainability—in a sustainable way. This article has outlined key recommendations and critical pillars to supplement the AI Commission’s Roadmap, demonstrating compelling rewards: a significant environmental impact and strong economic returns, with the potential to cut emissions by 10% and save around 250 billion SEK. We should no longer wait for such an initiative—the time to take sustainable action is now.
Situated under the umbrella of the House of Innovation at the Stockholm School of Economics, the Center for Data Analytics (CDA) is a center that provides applied, theoretical, and simulation-based research in statistics, econometrics, and data science, with applications in Business Administration, Economics, and Finance, and a special focus on AI for Sustainability. CDA has several research collaborations with other institutions as well as with the Swedish industry.
1Estimates are based on the author’s cost-benefit- and scenario analysis, using conservative assumptions on annual reductions, carbon pricing, and a 20-year horizon. Investment needs cover infrastructure, deployment, skills, and governance. Non-climate value draws from efficiency, public services, and health gains.