Jump to content

Course:CONS200/2025FL1/Environmental Impact of AI

From UBC Wiki

Environmental Impact of AI

Introduction

Artificial Intelligence (AI) is generally described as "machines that can learn, reason, and act for themselves"[1]. It has taken 2025 by storm, becoming a staple used in many people’s everyday lives. Its impact on careers, academics, and ecosystems is drastic. This ranges from its advances in innovation in relation to environmental sustainability to increased water depletion for cooling purposes. AI has its origins dating back to the mid-20th century, evolving from theoretical notions to practical applications. With the increasing use of AI, investigating its environmental effects is crucial in determining its usage.

AI’s challenges can range from environmental costs, excessive e-waste, and current gaps in AI policy[1]. However, increasing pro-AI rhetoric as a tool against climate change has been a powerful counter argument against AI’s costs. As AI continues to bleed into every sector of the economy, appropriate policy is also required, and currently being enacted. The policy gap at the moment shows just how fast this technology is outpacing humanity, and needs to be quickly countered to maintain control[2].

Where AI may go in the future is unbeknownst to all. Some predict a takeover of jobs, mindless activities, and further depletion of critical resources while others predict that AI proves as a powerful assistant in innovation against climate change, bringing about a net zero effect on climate[3]. Whether AI’s net impact is a positive or negative one is up for debate, as both the pros and the cons have significant effects on society today and must be further researched and observed in order to determine responsible use.

Challenges

Two big challenges surrounding AI are its water footprint - which describes the total volume of water used to create a certain good - and e-waste[1]. Even though it is argued that AI can contribute to environmental sustainability and ideas, its compounding effects on the environment overshadow the good it may do.

Water Consumption

Teraco Data Centre in South Africa

AI consumes water through two avenues - water withdrawal and water consumption[4]. Water withdrawal is the amount of freshwater taken from reserves, and consumption refers to the net reduction of water from the immediate water environment. These two components make up AI’s water footprint. As far as AI production goes, significant amounts of water is needed to produce its microchips - roughly 2,200 gallons of water to produce a single microchip[1]. This water usage embedded in the supply chain further increases AI’s water footprint. As for daily consumption, every 10-15 responses by OpenAi’s Chat-GPT uses around 500mL of water - multiplying this number with the approximate 2 billion queries per day[3] provides the figure of 1 billion litres of water from chats alone[3]. On top of that, data processing and cooling water consumption at data centres is on the rise, where it is predicted that by 2026, all data processing centres with AI will be at least 1,000-terrawatt hours - this is the same as Japan’s total annual consumption[1].  

E-Waste

AI also discards a significant amount of technology - referred better as electronic waste (or e-waste). In order to improve efficiency, technology needs to be continuously updated in data centres, hence increasing e-waste. E-waste poses many environmental threats, such as the risk of spreading toxins and contaminating ecosystems which impacts both humans and non-humans. Old non-degradable equipment is discarded which can end up in aquatic systems, thus affecting all other ecosystems through declining biodiversity and increased water toxicity[4]. As demand for AI and computational resources rises, so does the demand for more efficient AI technology, and thus shorter replacement cycles[5]. As replacement cycles shorten, this makes Thus, e-waste is not a byproduct, but rather as a consequence of AI-growth[5].

Advances: AI as a Climate Solution

AI as a Use for Biodiversity

As governments, industries, and scientific communities race to address accelerating climate change, artificial intelligence (AI) has emerged as a powerful driver of innovation across sustainability efforts. The transition to a low-carbon future increasingly depends on tools capable of analyzing massive datasets, identifying hidden environmental patterns, and generating precise forecasts that support real-time decision-making. AI meets these demands with unmatched analytical capacity, positioning it as a critical technology for advancing global net-zero ambitions[6]. Its applications span energy systems, transportation, agriculture, biodiversity protection, disaster forecasting, and waste management-sectors that collectively determine the world’s emissions trajectory[6].

AI in Renewable Energy and Grid Optimization

The energy sector responsible for a large share of greenhouse gas emissions stands at the centre of AI-driven climate solutions. AI technologies are increasingly integrated with the Internet of Things to improve forecasting accuracy, real-time monitoring, and system-wide control across renewable energy infrastructures[6].  These systems allow AI to perform essential tasks such as solar radiation modeling, wind energy prediction, simulation of renewable energy performance, and optimization of hybrid systems capabilities crucial for stabilizing grids powered by variable renewable sources[6].

AI enhances demand-side management as well. Machine learning algorithms can predict energy consumption patterns with high precision, enabling smarter scheduling, dynamic load balancing, and reduced peak-time strain[7]. This optimization has direct climate benefits: it increases renewable energy utilization, reduces fossil fuel backup needs, and lower carbon intensity across the energy supply chain[7].

Empirical evidence supports these outcomes. Research in Italy and Japan shows widespread adoption of AI-enabled energy management systems , demonstrating measurable improvements in efficiency, reliability, and energy conservation[6]. Likewise, early studies in the United Kingdom indicate that AI used for predictive maintenance although still emerging has significantly improved system performance by preventing equipment failures and extending infrastructure lifespan[6]. Together, these findings confirm AI’s indispensable role in decarbonizing modern energy systems.

AI Beyond Energy: Multi-Sector Pathways to Net Zero

Artificial intelligence also contributes to emission reductions across a diverse set of sectors that influence national climate pathways. AI’s capacity for large-scale analytics and complex modeling makes it particularly effective at optimizing transportation, agriculture, waste management, industrial systems, and climate monitoring[8].

  • In agriculture, AI-powered precision systems reduce fertilizer overuse; optimize irrigation, and detect crop stress early, all of which lower emissions from nitrogen fertilizers and improve carbon sequestration in soil[8].
  • In transportation, algorithms can optimize traffic flow, reduce congestion, and improve route efficiency, generating significant reductions in fuel consumption and air pollution[8].
  • In transportation, algorithms can optimize traffic flow, reduce congestion, and improve route efficiency, generating significant reductions in fuel consumption and air pollution[8].
  • In waste management, AI improves sorting efficiency, identifies recyclable materials, and optimizes collection routes, supporting circular economy strategies[8].

Meanwhile, in climate modeling, machine learning enhances the accuracy of extreme weather forecasting, flood prediction, and disaster preparation, strengthening resilience measures that protect vulnerable communities[8].

These multi-sector applications demonstrate how deploying AI can reshape environmental strategies, not only by reducing emissions but by increasing the efficiency and resilience of entire systems.

A Strategic Technology for Climate Resilience

AI is not only transforming how we produce energy or monitor emissions is rapidly becoming a foundational tool for building climate resilience. By enabling more accurate climate modeling, adaptive infrastructure planning, and dynamic energy system management, AI’s potential extends beyond mitigation into adaptation and long-term sustainability.

Modern machine learning and deep learning significantly improve predictive accuracy, spatial and temporal resolution, and the integration of multi-source data (satellite, sensor, atmospheric urban infrastructure), all essential for understanding future climate risks and planning resilient cities[9].

These AI-enhanced climate models allow urban planners, policymarkers, and stakeholders to simulate a variety of climate scenarios (eg., extreme weather events, sea-level rise, heatwaves, flooding) with higher fidelity than traditional models[9].This enables data-driven decision making; for example, identifying vulnerable infrastructure, assessing exposure, optimizing land use, and designing adaptive measures to increase long-term resilience.

Ultimately, the growing body of research makes a point unmistakably clear: AI is no longer just an emerging technology it is a strategic solution and an indispensable tool for navigating a rapidly changing climate. By strengthening our ability to predict environmental risks, adapt critical infrastructure, and manage energy systems with intelligence and precision, AI expands the possibilities for both mitigation and long-term resilience. While challenges remain in data quality, governance, and equitable access, the potential of AI to guide informed decisions and support climate-adaptive planning is unparalleled. When deployed ethically and collaboratively, AI becomes not only a driver of innovation, but also a cornerstone in building societies capable of withstanding and thriving amid the impacts of climate change.

Policy and Regulatory Responses

The Policy Gap

As artificial intelligence continues to transform sectors from healthcare to supply chain management, its environmental footprint is becoming a major policy concern. The development and deployment of advanced AI systems, particularly energy-intensive models, have led to rising carbon emissions, water usage, and e-waste, outpacing existing regulatory frameworks that historically focused on digital privacy, ethics, or innovation[10]Effective policy and regulation are crucial to ensure that AI contributes to climate solutions while minimizing its role as a growing stressor. This is an urgent concern, as governments and industries rush to harness AI’s benefits without sufficient oversight of its resource demands[2].

To situate the growing policy gap, it is useful to identify how current frameworks address (or fail to address) AI’s environmental risks and how different jurisdictions are beginning to respond. This context also reveals tensions between public and private interests, since companies control much of the computing infrastructure while governments increasingly bear responsibility for environmental oversight.

Current Policy Landscape

Global Efforts At the international level, the United Nations Sustainable Development Goals and the OECD AI Principles emphasize responsible innovation and global cooperation, embedding sustainability as a core objective. The European Union’s AI Act takes steps toward environmental risk assessment by mandating risk-mitigation measures for AI models, proposing energy and environmental reporting obligations, and linking compliance to broader sustainability goals outlined in the EU Green Deal[11]. However, environmental protection remains a secondary concern in the AI Act, with reporting and risk assessment for environmental impacts often limited or fragmented[11].

National and Regional Examples

Canada’s Pan-Canadian AI Strategy 2.0 briefly references sustainability, primarily within ethical frameworks and ongoing research initiatives that aim to quantify and mitigate the negative impacts of AI.8 Future opportunities include integrating environmental sustainability directly into funding and research priorities, such as the AI Sovereign Compute Strategy and AI Compute Access Fund[12].

United States policy remains focused on “responsible innovation” through the 2023 Executive Order on Safe, Secure, and Trustworthy AI, yet it offers minimal direct environmental criteria. Recent agency actions, however, have begun to emphasize comprehensive environmental reviews for AI infrastructure, including programmatic assessments of data centres under the National Environmental Policy Act (NEPA)[13].

European Union policy links the Green Deal’s climate ambitions to the AI Act, signalling potential requirements for energy efficiency reporting tied to regulatory compliance[11]. Meanwhile, directives such as EcoDesign and WEEE mandate energy standards and e-waste management for hardware relevant to AI, highlighting regulatory gaps in model training emissions and operational impacts[11].

Beyond North America and Europe, other regions are also beginning to integrate environmental considerations into AI governance. In the Asia–Pacific region, Singapore’s Model AI Governance Framework (2019, updated 2022) incorporates energy-efficient compute guidance within broader responsible AI practices[14]. South Korea’s Green Data Center Certification Program (2019) promotes energy-efficient data centre design, which indirectly influences the sustainability of AI computing[15].

In the Middle East, countries such as the United Arab Emirates pair large-scale AI investments with sustainability targets within their national digital transformation strategies, reflecting the interaction between the rapid deployment of AI and regional climate adaptation priorities[16]. Latin American governments, including those of Chile and Brazil[17], are emphasizing renewable-energy-powered data centres as part of their digital policy roadmaps, recognizing that the growth of AI must align with climate-vulnerable contexts[18].

These examples also highlight the uneven global geography of data centres and compute resources. AI training clusters are increasingly concentrated in the U.S., Western Europe, and East Asia, while many regions hosting the mineral extraction, manufacturing, or energy supply required for AI hardware face limited regulatory power over AI developers. This geographic mismatch reinforces the need for coordinated, cross-regional approaches to environmental governance[3].

Regulatory Gaps

Despite progress, major gaps persist:

  • There is no universal requirement for tracking or reporting the carbon impact of large-scale AI model training and inference​[2]
  • Jurisdictional fragmentation occurs as AI is developed globally but regulated nationally, resulting in inconsistent standards and loopholes[11]
  • The lack of direct regulation for the environmental impact of data centres or e-waste from AI hardware, with most mandates covering hardware broadly or stopping at voluntary reporting[19].
  • Difficulty in quantifying AI’s carbon footprint, since current metrics are narrow and often obscured by voluntary renewable credits[2]
  • Overregulation risks stifling innovation and competitiveness, while under regulation jeopardizes climate targets and societal trust—this trade-off underscores the need for balanced, evidence-based frameworks.

Emerging Frameworks

Carbon Accountability: Academic and policy proposals emphasize mandatory emissions reporting for large-scale AI training, recommending standardized metrics and data harmonization across agencies such as the DOE, NIST, and EPA in the US, as well as the EIA, NTIA, and FERC, for data collection and grid integration[2].

Sustainable AI Standards: The new ISO/IEC TR 20226:2025 standard outlines life-cycle metrics for AI systems, covering workload, asset utilization, carbon impact, pollution, and waste, offering an emerging blueprint for national and sectoral regulation[16].

Government Incentives: Funding and research programs are increasingly incentivizing AI deployments that optimize energy grids, improve biodiversity, and enhance resource management, as seen in Canadian strategies and multi-stakeholder coalitions[12].

Transparency & Data Ethics: Experts argue for public disclosure of the compute, energy mix, and hardware sourcing for AI models, alongside social science-informed policy on the dual use of AI in both climate mitigation and risk acceleration[17].

AI’s environmental footprint is both a challenge and an opportunity for global climate policy. Integrating climate targets with digital governance requires a hybrid approach that combines technical regulation (metrics, standards, and compliance), corporate accountability (mandatory reporting and responsible innovation), and international coordination for policy alignment and enforcement. Only by closing the policy gap through comprehensive regulation that recognizes AI’s dual role as both a solution and a stressor can policymakers ensure that AI maximizes progress toward sustainability without exacerbating new forms of resource risk[10].

Future Directions

AI use will only increase in the future, but its impact on the environment depends on how it is used. The growing reliance on AI presents both opportunities and challenges; while it has the potential to enhance sustainability efforts, it also carries environmental costs that must be carefully managed[6].

Increased use of AI in climate mitigation technology

As technologies continue to improve and develop, AI is increasingly utilized in various fields, including the medical field, education, and, of course, climate mitigation. AI applications in climate mitigation have shown measurable benefits, such as forecasting weather extremes, lowering operational costs, and improving resource management efficiency[7]. AI will continue to evolve into a tool that many people believe will play a central role in navigating climate mitigation strategies.

Limitations

Despite its potential and increasing use, AI has specific limitations when it comes to climate action, particularly when expectations exceed current capabilities. Most importantly, AI cannot operate independently and must be integrated into existing frameworks to be effective in addressing environmental issues[7]. Though evidence has shown AI to be a valuable tool when integrated into existing renewable energy systems, there is little evidence that it alone can drive full decarbonization[7]. Similarly, in GIS applications, AI enhances spatial analysis and climate modelling, but its output is only reliable when GIS theories are deeply embedded in the AI system[20]. Further limitations of AI development include a lack of organizational or policy support, data availability, and technological barriers[21]. Addressing these challenges will be critical in shaping the realistic and responsible future use of AI for environmental sustainability. Recognizing these limitations helps set realistic expectations, ensuring AI is applied where it can genuinely add value.

Long-term environmental effects

Along with water consumption and electronic waste, there are other environmental impacts of AI that remain underexplored, particularly those that may not become apparent right away. The energy efficiency paradox highlights this risk; while AI can improve efficiency, it may unintentionally encourage greater resource use[6]. For instance, livestock management practices driven by AI may increase herd sizes as efficiency improves, unintentionally leading to higher methane emissions[6]. This underscores the importance of weighing the benefits of future AI use in climate mitigation against potential unintended consequences.

Future Use

Moving forward, it is essential to strike a balance between mitigating environmental harm and addressing environmental problems. AI must integrate with existing climate strategies and policies, supporting processes that are already in place[22]. Future research should prioritize developing AI models that are compatible with existing sustainability frameworks, validated in real-world settings, and explicitly targeted at actionable outcomes. Ultimately, responsible AI development and integration will determine whether it becomes a genuine tool for sustainability or another contributor to environmental risk.

Conclusion

Artificial intelligence holds an enormous promise for advancing environmental sustainability, from optimizing energy systems and predicting climate patterns to supporting innovative solutions in agriculture and transportation. Yet, these technological strides come with environmental costs of their own, including high energy consumption, electronic waste, and the digital rebound effect, where efficiency gains lead to greater overall resource use. To truly make AI an ally in the fight against climate change, we must strike a careful balance between innovation and responsibility. This means ensuring that the development and deployment of AI are guided by transparent policies, sustainable energy practices, and ethical frameworks that prioritize long-term planetary well-being over short-term progress.


References

 This conservation resource was created by Course:CONS200.
  1. ↑ 1.0 1.1 1.2 1.3 1.4 Simpson, Whitni (Winter 2025). "A Need to Regulate the Environmental Impacts of Artificial Intelligence (AI)". Tulane Environmental Law Journal. 38: 133–148 – via JSTOR.
  2. ↑ 2.0 2.1 2.2 2.3 2.4 Jhaveri, Mitul (June 2025). "Measuring and Standardizing AI's Energy and Environmental Footprint to Accurately Access Impacts". Federation of American Scientists.
  3. ↑ 3.0 3.1 3.2 3.3 Luers, Amy (August 22 2025). "Net zero needs AI — five actions to realize its promise". nature. Check date values in: |date= (help)
  4. ↑ 4.0 4.1 Berreby, David (February 2024). "As Use of A.I. Soars, So Does the Energy and Water It Requires". YaleEnvironment360.
  5. ↑ 5.0 5.1 Winsta, Jenis (July 2025). "The Hidden Costs of AI: A Review of Energy, E-Waste, and Inequality in Model Development". Cornell University.
  6. ↑ 6.0 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 Chen, Lin; Chen, Zhonghao; Zhang, Yubing; Liu, Yunfei; Osman, Ahmed I; Farghali, Mohamed; Hua, Jianmin; AI-Fatesh, Ahmed; Ihara, Ikko (13 June 2023). "Artificial intelligence- based solutions for climate change: a review". Environmental Chemistry Letters. 21: 2525–2557 – via Springer Nature Link. Cite error: Invalid <ref> tag; name ":0" defined multiple times with different content
  7. ↑ 7.0 7.1 7.2 7.3 7.4 Shahverdi, Nikan; Saffari, Arina; Amiri, Babak (June 2025). "A systematic review of artificial intelligence and machine learning in energy sustainability: Research topics and trends". Energy Reports. 13: 5551–5578 – via Science Direct. Cite error: Invalid <ref> tag; name ":1" defined multiple times with different content
  8. ↑ 8.0 8.1 8.2 8.3 8.4 8.5 Olawade, David B; Wada, Ojima Z; David-Olawade, Aanuoluwapo; Fapohunda, Oluwaseun; Ige, Abimbola; Ling, Jonathan (2024). "Artificial intelligence potential for net zero sustainability: Current evidence and prospects". Next Sustainability. 4 – via Science direct.
  9. ↑ 9.0 9.1 Amuaylojaroen, Teerchai (19 May 2025). "Advancements and challenges of artificial intelligence in climate modeling for sustainable urban planning". Frontiers in Artificial Intelligence. 8.
  10. ↑ 10.0 10.1 Alnafrah, Ibrahim (September 2025). "The Two Tales of AI: A Global assessment of the environmental impacts of artificial intelligence from a multidimensional policy perspective". Journal of Environmental Management. 392.
  11. ↑ 11.0 11.1 11.2 11.3 11.4 Ebert, Kai (May 2025). "AI, Climate, and Regulation: From Data Centers to the AI Act".
  12. ↑ 12.0 12.1 "Canada's AI Strategy Powers Sustainable Industry Innovation". Business and Industry: Canada. 2025. |first= missing |last= (help)
  13. ↑ Biden, Joseph (January 14, 2025). "Executive Order 14141- Advancing United States Leadership in Artificial Intelligence Infrastructure". The American Presidency Projec.
  14. ↑ "Regulations and Licensing". Infocomm Media Development Authority.
  15. ↑ "Korea Internet & Security Agency (KISA): Homepage". Korea Internet & Security Agency.
  16. ↑ 16.0 16.1 "Information technology — Artificial intelligence — Environmental sustainability aspects of AI systems (ISO/IEC TR 20226:2025)". International Organization for Standardization. 2025.
  17. ↑ 17.0 17.1 Javed, H (2025). "AI and the Scope 3 Emissions Challenge: Closing the Carbon Accountability Gap". The AI Journal.
  18. ↑ "InvestChile: Homepage". InvestChile.
  19. ↑ Zhuk, A (2023). "Artificial Intelligence Impact on the Environment: Hidden Ecological Costs and Ethical-Legal Issues". Journal of Digital Technologies and Law. 1: 932–954.
  20. ↑ Dalal, S; Chaudhary, J (2025). "Review of artificial intelligence and GIS in climate change and ecosystem research: Applications and future directions". Recent Trends in Intelligent Computing and Communication. 1st ed. – via Taylor & Francis.
  21. ↑ Abdulameer, L; Al-Khafaji, M. S.; Al-Awadi, A. T.; Al-Maimuri, N. M. L.; Al-Shammari, M.; Al-Dujaili, A. N.; Al‑Jumeily, A. N. (2025). "Artificial Intelligence in Climate-Resilient Water Management: A Systematic Review of Applications, Challenges, and Future Directions". Water Conservation Science and Engineering. 10 – via Springer.
  22. ↑ Tan, X; Peng, Z; Cheng, Y; Wang, Y; Chao, Q; Huang, X; Yan, H; Chen, D (2025). "Leveraging artificial intelligence for research and action on climate change: Opportunities, challenges, and future directions". Science Bulletin. 70 – via Elsevier Science Direct.
Retrieved from "https://wiki.ubc.ca/index.php?title=Course:CONS200/2025FL1/Environmental_Impact_of_AI&oldid=881974"