Today, Skytells is joining the Coalition for Sustainable AI. We take our place alongside Nvidia, Cisco, IBM, Hugging Face, Mistral AI, SAP, and other technology companies, research institutions, and public sector organizations working to align artificial intelligence with environmental responsibility.
The coalition launched at the AI Action Summit in Paris in February 2025. It was built to answer a question the AI industry can no longer avoid: how do we expand the use of AI while keeping its energy, water, and hardware footprint within limits the planet can sustain?
A Meaningful Step in Skytells' Sustainability & Vision
AI has quietly become infrastructure. It runs inside logistics networks, healthcare diagnostics, banking systems, scientific research, public services, and the software tools most engineers use every day. As that footprint grows, the environmental cost of training and serving models grows with it.
At Skytells, we have been working on this problem for a while—from contributing 2% of our revenue to carbon removal projects through our Climate program, to building an efficient multi-agent orchestration platform designed around the principle of efficiency by design, to partnering with renewable energy providers for our data center contracts and measuring the energy use of our workloads from the early days. But the industry is at a point where individual efforts are not enough. The standards for measuring and reporting AI's environmental impact are still emerging, and the decisions being made now about architecture, deployment, and disclosure will shape the trajectory of AI's footprint for years to come.
We joined the Coalition for Sustainable AI because that cost is now a design problem, not a reporting problem. Energy efficiency, water use, hardware lifecycles, and carbon intensity all get decided early: when a team picks a model, writes an orchestration layer, provisions a cluster, or signs a data center contract. If those decisions aren't made carefully, no amount of later disclosure will fix them.
Membership gives us a seat at the table where those decisions are being standardized. It also commits us publicly to a direction we were already heading: building AI products that are measured, efficient by design, and honest about their environmental impact.
What We Are Bringing to the Table
Skytells did not arrive at the coalition with a set of aspirations. We arrived with a running sustainability program, a production platform built for efficiency, and a public commitment that our customers can verify today:
- 2% of annual revenue directed to verified carbon removal projects through the Skytells Climate program.
- Skytells Cloud Agents, a multi-agent orchestration platform that activates specialized agents only when a task requires them, so compute is provisioned to match real workload rather than reserved and left idle.
- Renewable energy sourcing as a default requirement in our data center contracts, not an optional upgrade.
- Internal energy accounting for model inference, agent orchestration, and data processing workloads, in place since early in the platform's lifecycle.
These are the foundations we are contributing into coalition workstreams—and the basis on which we will adopt the standardized metrics that come out of them.
What the Coalition for Sustainable AI Is
The Coalition for Sustainable AI is a global initiative co-initiated by France together with the UN Environment Programme (UNEP) and the International Telecommunication Union (ITU)1. Its purpose is to align AI development with environmental sustainability and with the UN's 2030 Agenda for Sustainable Development.
Members span the full AI value chain. Chip makers, cloud providers, foundation model labs, enterprise software vendors, telecom operators, national regulators, climate NGOs, research universities, and sustainability-focused investors all work inside the same coalition. That range matters, because AI's environmental impact is produced across layers that rarely talk to each other: silicon in one company, data centers in another, models in a third, applications in a fourth.
The work sits on two tracks. The first is Green AI: cutting the energy, water, and hardware cost of building and running AI systems. The second is AI for Green: applying AI to climate science, renewable energy, biodiversity monitoring, agriculture, and disaster response. Most members contribute to both2.
Why Sustainable AI Matters Now
AI has outgrown the lab. It ships products, runs customer support, writes code, reads scans, routes trucks, and informs public services. That growth has pushed the environmental cost of AI from a niche concern for data center operators into a mainstream business question that shows up in board meetings, procurement reviews, and regulatory filings.
Three pressures explain the urgency. Data center electricity demand is climbing faster than grids can add clean generation. Water consumption for cooling high-density AI clusters is rising in regions that were already water-stressed. And the supply chain behind advanced accelerators depends on materials and manufacturing capacity that cannot scale indefinitely. Ignoring any of these is how AI becomes an environmental liability instead of a productivity tool.
The coalition exists because there are two jobs to do at once, and the industry cannot pick one.
1. Build AI More Sustainably
Training large models means running clusters of accelerators continuously for weeks or months. Serving those models means keeping inference capacity online across multiple regions, every hour of every day. Behind all of that sits a supply chain of data centers, networking gear, cooling systems, and specialized semiconductors that takes years to plan and build.
The scale of the resulting impact is no longer abstract. Training a single large AI model can emit as much carbon dioxide as five cars over their lifetime—a comparison cited by the coalition itself3—and at scale those costs compound across energy consumption, water use for cooling, semiconductor manufacturing, and hardware e-waste. The coalition's response is to push for standardized measurement methods, comprehensive lifecycle analysis covering both hardware and software, and transparent reporting frameworks that let organizations disclose the environmental footprint of their AI products on comparable terms2.
The levers that actually reduce impact are well understood, even if they are rarely measured consistently. They include choosing the smallest model that solves the task, designing orchestration that only activates compute when it is needed, keeping provisioned hardware busy instead of idle, and placing workloads in regions with cleaner grids and lower cooling overhead. These choices compound. Shared measurement standards are what make the savings visible, comparable, and believable.
2. Use AI to Accelerate Environmental Progress
The same models that consume energy can also help protect it. AI is already improving the accuracy and resolution of climate projections, balancing renewable supply with demand on electricity grids, enabling precision agriculture that reduces water and fertilizer use, monitoring ocean health and illegal fishing from satellite imagery, tracking deforestation, and giving emergency services earlier warnings for floods, wildfires, and heatwaves. This work aligns directly with the UN's Sustainable Development Goals, particularly SDG 13 (Climate Action), SDG 14 (Life Below Water), and SDG 15 (Life on Land)2.
Turning those capabilities into real impact takes more than a good model. It takes clean data, trusted deployment, measurement of the downstream effect, and people who understand both the software and the domain it is working in. The coalition is set up to connect those groups—climate scientists with ML engineers, ministries of environment with cloud providers, community NGOs with researchers—so that "AI for good" produces measurable outcomes and policy-grade evidence instead of isolated demos.
The Role of France, the UN Agencies, and UNESCO
The coalition emerged from the AI Action Summit held in Paris on 10–11 February 2025, a gathering that brought together heads of state, technology executives, climate scientists, and civil society leaders to address AI governance and environmental sustainability1. France's decision to co-initiate the coalition reflects a deliberate strategy: place Europe at the intersection of technological innovation and environmental stewardship, and use the convening power of a host state to keep the AI industry's fastest-moving conversations tied to climate and development agendas rather than detached from them. The coalition's shared vision, formally endorsed by all members, centers on ensuring AI "contributes positively to environmental and climate objectives and can be deployed at scale"2.
The coalition's two UN co-initiators bring complementary authority. UNEP, the UN Environment Programme, contributes decades of experience coordinating international environmental action and scientific assessment, which keeps the coalition's work aligned with frameworks such as the Paris Agreement and the Convention on Biological Diversity and ensures new AI-specific standards build on existing environmental science rather than starting from a blank page. ITU, the UN specialized agency for information and communication technologies, brings technical expertise on digital infrastructure, standards processes, and global telecommunications policy; it has been particularly active in promoting the coalition's agenda at COP conferences, including AI for Climate startup contests at COP30 in Belém3, and its role matters because the same standards-setting machinery that shaped the modern internet will shape how environmental impact is measured and reported across AI systems.
UNESCO adds a third dimension by organizing high-level sessions on sustainable AI at major international gatherings and by framing AI's environmental impact inside a wider conversation about digital rights, access, and public interest. At ADOPT AI in November 2025, Guilherme Canela, Director of UNESCO's Division for Digital Inclusion and Policies and Digital Transformation, moderated the "Greening AI and Greening with AI" session, which connected the coalition's technical work to questions of inclusion and equitable access3. The coalition is also articulated with the Coalition on Digital Environmental Sustainability (CODES), launched in 2021 with eight co-champion organizations, creating continuity between earlier digital sustainability efforts and AI-specific initiatives and ensuring the coalition's mandate sits inside a longer arc of international work on technology and the environment1.
Who Is in the Coalition
Membership spans private companies, research institutions, public sector agencies, NGOs, and investors across six continents4. The mix is deliberate. Sustainable AI is not something any one segment of the industry can deliver alone, so the coalition was built to hold the full value chain in one room—cloud providers alongside open-source AI labs, chip manufacturers in working groups with climate NGOs, telecom regulators next to impact investors and university research centers.
Technology Companies
Corporate members include Nvidia, Cisco, IBM, Arm, AMD, Baidu, Naver, LG AI Research, Mistral AI, Hugging Face, SAP, Accenture, Capgemini, Orange, Nokia, Lenovo, Autodesk, Altair, Ampere, AVEVA, Clarity AI, Gen Digital, OVHcloud, L'Oréal, and Generali, among many others4. Together they cover most of the global AI stack—semiconductor design, cloud infrastructure and networking, foundation models, enterprise applications, and customer-facing products. Getting all of those actors into the same working groups matters because the biggest efficiency gains usually happen at the seams between layers: a more efficient chip only reduces real-world emissions if cloud providers schedule workloads to use it well, if model developers pick architectures that map onto it, and if application teams retire older inference paths once better ones exist.
Research Institutions
Leading research centers in AI and sustainability participate, including INRIA (France), Stanford University's Bits & Watts Initiative (United States), MILA Quebec AI Institute (Canada), the Stockholm Environment Institute (Sweden), CEA, the AI & Society Institute at École Normale Supérieure, Ifremer, Institut Louis Bachelier, the Humboldt Institute for Internet and Society (HIIG), and universities including Loughborough, Universidad Politécnica de Madrid, Pavia, Maribor, Malta, and Vrije Universiteit Amsterdam4. Their role is to keep standards and best practices grounded in peer-reviewed work—how to measure lifecycle footprint, how to fairly compare systems trained on different hardware in different regions, how to weigh training cost against downstream benefit—rather than letting standards be shaped only by commercial or political pressure.
Governments and Public Sector Agencies
Government participation gives the coalition a mandate and a reach private initiatives cannot match, linking technical work to the regulatory, procurement, and investment levers governments actually control4. Current public sector members include—from France—ADEME, Arcep, Bpifrance, CNNum, IGN, Shom, Cerema, and the French Ministry of Ecological Transition; from Thailand, NECTEC; and from Serbia, the Office for Information Technologies and eGovernment (NITRA). The French agencies alone cover ecological transition, telecom regulation, public investment, digital policy, geospatial data, hydrography, and urban mobility, giving the coalition real public-sector use cases to test against. The presence of Thai and Serbian agencies signals the coalition is not designed only around G7 priorities; emerging deployment contexts are in scope from the start.
NGOs and International Organizations
Civil society members include Climate Change AI, Climate Policy Radar, Impact AI, Hub France IA, Mercator Ocean International, Nature Finance, BSR, Libraries Without Borders, Clean Tech Open France, Ad Net Zero, AFNOR, AI Cargo Foundation, Blockchain & Climate Institute, and Climate Collective4, alongside others focused on environmental action, digital rights, and inclusive access. They bring fieldwork in climate-affected regions, legal analysis of AI governance, and frameworks for equity and long-term public interest—and their presence functions as a check, flagging greenwashing, pushing for measurable commitments, and insisting the benefits of AI for climate flow to communities most exposed to environmental harm.
Investors
Investor members include Mirova, Eurazeo, Tikehau Capital, Ardian, Crédit Agricole, Daphni, GIRA Digital Power, Growthfund, Princeville Capital, Raise Ventures, and Revaia4. Their involvement tracks a broader shift: environmental performance is now a material factor in diligence, valuation, and portfolio construction, not a reporting line item. Putting investors in the same coalition as operators, researchers, and governments closes a feedback loop—the metrics and disclosures built inside the coalition become signals investors can use, which gives portfolio companies a clear incentive to adopt them. That is what makes sustainable AI economically durable, rather than dependent on any single company's goodwill.
International Momentum and Upcoming Milestones
The coalition maintains an active presence at major international forums, keeping sustainable AI on the agenda at venues where technology policy, climate action, and development priorities converge. Its 2025–2026 calendar has been built around the moments where AI standards, climate negotiations, and industrial policy actually meet.
At COP30 in Belém (2025), the ITU ran AI for Climate startup contests that showcased practical applications of AI for environmental monitoring and resource optimization3. A few weeks later, the ADOPT AI Conference (Paris, 25–26 November 2025), held at the Grand Palais, extended the AI Action Summit into a business-focused format, drawing CEOs and C-level executives from CAC40 corporations, startups, unicorns, and global tech partners across tracks including AI for Finance, AI for Health, and AI for the Planet. A dedicated coalition session on "Greening AI and Greening with AI: From Climate Footprint to Climate Action" was organized by UNESCO on 26 November3. The AI Standards Summit (Paris, 2–3 December 2025) then focused on developing the common frameworks needed to measure and report the environmental impacts of AI3, and the India AI Summit (February 2026) extended those conversations toward resilience and efficiency in resource-constrained deployment environments3.
Looking further ahead, the coalition's roadmap includes continued engagement at COP conferences, regional AI summits, and specialized gatherings focused on specific environmental domains such as oceans, agriculture, energy systems, and disaster response. France's sustained leadership signals that sustainable AI will remain a priority in both European and international technology policy through 2026 and beyond.
The Coalition's Initiative Hub and Working Groups
Beyond convening stakeholders, the coalition operates an active Initiatives Hub that catalogs and connects cross-border efforts in sustainable AI5. It is organized around seven areas: Green AI research and deployment; AI for Green use cases from smart grids to biodiversity monitoring; shared reports and publications; capacity-building through online courses and webinars; hackathons and challenges such as the 2026 Tideline Startup Challenge; AI and environmental knowledge; and participation in the global AI governance framework that integrates environmental considerations into AI policy from the outset5.
Members participate in workstreams at their discretion, contribute initiatives for listing in the hub, and access resources developed by other members. This structure avoids duplication while letting each organization focus on areas aligned with its capabilities and priorities—which is how a coalition this wide stays useful without becoming bureaucratic.
What Skytells Is Committing To
Joining the Coalition for Sustainable AI formalizes work we were already doing and binds it to international standards. Our commitments cover four areas—measurement, architecture, standards collaboration, and applied AI for the environment—and each one is anchored to something already running in production, not a roadmap slide.
Measuring the Environmental Impact of Our AI Workloads
You cannot manage what you do not measure. We track the environmental impact of AI workloads across our platform through three disciplines:
- Energy accounting for agent orchestration, model inference, and data processing—per-workload, not per-account averages.
- Model efficiency analysis that documents which tasks run well on smaller, cheaper models and which genuinely need larger ones, so model selection is justified rather than defaulted.
- Infrastructure metrics covering data center efficiency, renewable energy sourcing, and hardware utilization across our regions.
As the coalition publishes standardized measurement frameworks2, we will adopt them directly and expose the relevant data to customers who need to report on the environmental footprint of their own AI stack—so the numbers our customers report to regulators and their own boards come out of a framework built in public with international partners, not out of internal spreadsheets.
Efficient Architecture by Design
Skytells Cloud Agents was built around one assumption: most AI tasks do not need the biggest available model running continuously. Our orchestration platform activates specialized agents only when a task actually requires them, which keeps compute overhead down without hurting quality. Compared to monolithic, always-on systems, this architecture uses less energy by default. Routing tasks to the right-sized model and scaling agent deployment to match real workload keeps provisioned capacity from sitting idle—and that design choice compounds into real savings at scale, because the inefficiency of oversized always-on inference is one of the largest hidden costs in production AI today.
We will keep refining that approach, informed by the work other coalition members are doing on efficient multi-agent systems, and publish architectural patterns, trade-offs, and measurement methods so other teams can apply them without starting from scratch.
Contributing to Shared Standards
Sustainable AI is not something any one company can standardize by itself. The coalition's model of "multi-stakeholder collaboration—bringing together governments, academia, civil society, and the private sector"2 matches how we think this work should happen, and it is the reason we chose to contribute inside a formal coalition rather than publish parallel frameworks alone.
We plan to contribute to coalition workstreams on lifecycle impact assessment that accounts for both hardware production and software energy consumption; reporting frameworks that let organizations compare different AI systems and deployment scenarios on equal terms; and multi-agent architecture best practices, where our production experience with orchestration efficiency is most useful.
Applying AI to Environmental Problems
Our core platform today is focused on software engineering workflows, but the same agent orchestration patterns are useful well beyond that—environmental monitoring, resource optimization, climate adaptation. The coalition's initiatives hub includes active workstreams on AI for climate, ocean protection, sustainable agriculture, and disaster response5. As those workstreams mature, we plan to look at where our technology can contribute and, through the Skytells Climate program, continue directing 2% of our annual revenue to verified carbon removal projects that offset the footprint we cannot yet eliminate.
The Scale of the Challenge and Why Coordination Matters
The AI industry is at a decision point. The choices being made now—about architecture standards, measurement frameworks, disclosure requirements, and which applications get prioritized—will decide whether AI ends up as a climate liability or a tool that helps with environmental stewardship. The trajectory is the reason this work is urgent: AI workloads in 2026 are a fraction of what is projected for 2030, every major technology company has announced large AI infrastructure investments, governments are launching national AI strategies, and applications keep moving into new sectors.
On current growth, data center electricity demand starts to rival that of small countries. Cooling water use pressures municipal supplies in regions that were already water-stressed. Semiconductor production adds strain to supply chains that are already tight. E-waste from short hardware refresh cycles becomes its own disposal problem. None of this is inevitable, but none of it fixes itself either—and the window in which the industry can influence its own trajectory is narrow, because today's infrastructure decisions lock in energy and water use for a decade or more.
Each company optimizing its own systems helps, but it is not enough on its own. Without shared frameworks, "efficiency" is hard to define or verify. Without shared research, every team repeats work others have already done. Without transparency, customers and regulators cannot tell real progress from marketing. Better algorithms lower compute requirements, more efficient hardware does more work per watt, smarter orchestration reduces idle capacity, clean energy contracts decouple compute from carbon, and circular economy practices extend hardware life—but making those improvements happen at scale requires coordination across the supply chain, from chip designers to cloud providers to application teams.
The Coalition for Sustainable AI is designed to close that coordination gap. It gives companies that often compete in the market a neutral place to solve problems together—problems that none of them can solve alone and that matter well beyond any single product roadmap.
How This Membership Shows Up in Our Product
Coalition membership is not a press statement. It changes the way we prioritize internal work. Environmental impact is now an explicit input to roadmap decisions, documentation priorities, and customer-facing reporting, sitting alongside performance, reliability, and cost as a dimension we are actively managing rather than passively observing.
Practically, that shows up in five places:
- Roadmap. As frameworks for measuring agent orchestration efficiency mature, we will build native support for those metrics into the platform so customers can track environmental impact next to performance and cost.
- Efficiency features first. Work that reduces compute waste—smarter model selection, better caching, tighter agent lifecycle management—moves up the queue, informed by research shared inside the coalition.
- Documentation. We will publish what we learn about running multi-agent systems efficiently, including architectural patterns, trade-offs, and measurement methods, so other teams can use them.
- Research. Where our production experience is useful—particularly on multi-agent efficiency—we will contribute. Where we lack expertise, we will learn from members that have been working on environmental science and climate policy for longer.
- Reporting. As coalition reporting standards come online, we will adopt them and keep our methodology open, so customers can verify environmental claims instead of taking them on trust.
The direction of travel is simple: we want "sustainable by default" to stop being a feature Skytells markets and start being a property of the platform that customers can take for granted—verified by standardized metrics developed with international partners, rather than asserted in marketing copy.
Why We Joined
We joined the Coalition for Sustainable AI because the challenge is too big for any one company to solve, and the opportunity is too important to sit out. Making AI sustainable means changing how hardware is built, how data centers are run, how models are designed, how applications are deployed, and how all of that is disclosed—work that requires governments, researchers, NGOs, investors, and companies that normally only meet at conferences to work on the same problems together.
We build tools for engineering teams, and engineering teams are the ones who will implement whatever standards come out of this work. Our role inside the coalition is to help make sure those standards are practical, implementable, and compatible with how modern software is actually built and operated—so that environmental responsibility does not arrive as a compliance tax bolted onto finished systems, but as a property designed in from the first commit.
There is also a more self-interested reason, and it is worth being honest about it. Customers, regulators, and investors are asking hard questions about the environmental impact of AI, and the companies that have not engaged with those questions will be the ones scrambling to answer them. Joining the coalition early lets us help shape the standards instead of inheriting them—and lets us tell customers, with receipts, that the environmental claims we make about our platform are rooted in frameworks built in public with international partners.
Sustainable AI is not a quarterly initiative. The decisions made over the next few years will shape AI's environmental trajectory for decades. Infrastructure built now will be in service for a long time, architectural patterns set today will show up in thousands of downstream systems, and the norms the industry adopts for measurement and transparency will decide what becomes visible—and therefore what becomes possible to govern. That is the horizon we are signing up for. Standards development is slow, behavior change across an industry is slower, and efficiency gains compound gradually rather than arriving as dramatic breakthroughs. The coalition's value is in direction more than speed: it is aligning the AI industry with environmental limits from the start, rather than trying to retrofit sustainability once systems are already deployed at scale.
Building AI That Is Both Useful and Sustainable
Innovation and environmental responsibility are not opposites. With honest measurement, deliberate architecture, and coordinated industry action, AI can be powerful and efficient at the same time. That is the working assumption of the Coalition for Sustainable AI, and it is the one we are signing up to.
The work ahead is clear enough. Build standardized measurement frameworks that hold up across different AI systems. Create transparent reporting that informs rather than confuses. Design architectures that deliver capability without waste. Apply AI to environmental problems at a scale that actually moves the numbers.
Skytells' membership in this coalition—alongside Nvidia, Cisco, IBM, Hugging Face, Mistral AI, SAP, French governmental agencies, UNEP, ITU, UNESCO, and leading research institutions across six continents—reflects our conviction that this is the right path forward.
Build on a Sustainable AI Platform
If you are evaluating AI infrastructure against environmental, procurement, or ESG criteria—or preparing to report on the footprint of your own AI stack—Skytells is built to meet that bar from the first commit.
- Talk to our team about sustainability and enterprise AI reporting: contact Skytells.
- Learn about the Skytells Climate program, including how 2% of our revenue is directed to verified carbon removal.
- Explore Skytells Cloud Agents, our efficiency-first multi-agent orchestration platform.
- Read the coalition's public materials at sustainableaicoalition.org.
About the Coalition for Sustainable AI: The Coalition for Sustainable AI is a multistakeholder global initiative co-initiated by France in collaboration with the UN Environment Programme (UNEP) and the International Telecommunication Union (ITU). Launched at the AI Action Summit in Paris in February 2025, the coalition brings together corporations, research institutions, governments, NGOs, and investors committed to ensuring AI contributes positively to environmental and climate objectives. Learn more at sustainableaicoalition.org.