As AI adoption accelerates across every department, its energy, water, and carbon footprint is becoming a governance issue, not just an engineering one.

Every AI-generated email draft, every chatbot query, every model your organization fine-tunes runs on a physical data center somewhere, pulling electricity off a real grid and, in many cases, water out of a real reservoir. As AI adoption accelerates across every department, from marketing to finance to operations, the environmental footprint of that adoption is no longer a topic reserved for data center engineers. It is becoming a boardroom issue, a compliance issue, and increasingly, a career issue for the professionals expected to use AI responsibly.

This is the conversation behind sustainable AI: understanding what AI actually costs the planet, and what that means for how professionals across every function should think about the tools they use every day.

AI’s environmental cost is not hypothetical anymore. It is showing up in utility bills, water permits, and corporate sustainability reports. Global data center electricity consumption reached roughly 415 TWh in 2024 and is projected by the International Energy Agency to climb toward 945 to 1,050 TWh by 2030, with AI accelerator workloads accounting for the largest share of that growth. In some regions, the strain is already visible at the grid level: data centers now account for a substantial share of total electricity use in markets like Ireland and the U.S. state of Virginia, where local utilities have had to expand generation capacity specifically to keep up with AI demand.

What has changed is the shape of the workload. A generative AI training cluster can draw several times more power than a typical enterprise computing workload, and the newest AI server racks require liquid cooling because they run far hotter than the air-cooled infrastructure most data centers were built for. That shift is forcing an infrastructure retrofit across the entire industry, and it is happening faster than most organizations’ sustainability strategies were designed to handle.

It is tempting to think of a single AI query as environmentally negligible, and on its own, it largely is. The issue is scale. Training a large language model like GPT-3 has been estimated to consume over 1,000 megawatt-hours of electricity and produce several hundred tons of CO₂ equivalent, before a single end user ever types a prompt. Once a model is deployed, inference (the actual day-to-day use of the model) becomes the dominant, ongoing cost, multiplied across billions of queries a day across the industry.

For professionals evaluating AI vendors or tools, this matters because energy intensity varies enormously between models, providers, and use cases. A well-optimized, purpose-built model answering a narrow business question can be dramatically more efficient than a massive general-purpose model used for the same task. That is a procurement and governance question as much as a technical one.

Electricity gets most of the attention, but water is the environmental cost most professionals have not heard about. Data centers use water in two ways: directly, through evaporative cooling systems that keep servers from overheating, and indirectly, through the water consumed by power plants generating their electricity.

Estimates for AI’s direct water use vary depending on model size and data center design, but researchers have put the cost of a single AI-generated response somewhere in the range of a few milliliters to tens of milliliters of water, and training a large model can evaporate hundreds of thousands of liters of freshwater on site. Projected forward, some researchers estimate global AI-related water withdrawals could reach several billion cubic meters annually within the next few years, a volume comparable to the annual water consumption of a mid-sized European country.

The bigger concern for professionals working in ESG, procurement, or site selection is where this water is being drawn. A significant share of data centers built in recent years sit in regions already facing water stress, and local communities and regulators are pushing back. Any organization signing a new cloud contract, expanding on-premises AI infrastructure, or opening a data center lease should be asking vendors for water usage effectiveness (WUE) figures, not just power usage effectiveness (PUE).

The carbon math behind AI has two layers most people miss. The first is operational: the emissions tied directly to the electricity a data center consumes, which vary enormously depending on whether that grid runs on coal, gas, or renewables. The second is embodied: the emissions baked into manufacturing the GPUs, servers, and cooling infrastructure themselves, plus the electronic waste generated as AI hardware, which cycles through generations far faster than typical enterprise IT, reaches end of life.

Researchers project that AI-driven data center growth in the U.S. alone could add tens of millions of metric tons of CO₂-equivalent emissions annually by 2030, depending on how quickly the grid supplying that growth decarbonizes.

For a long time, AI’s environmental footprint was a voluntary disclosure, something companies mentioned in a sustainability report if they felt like it. That is changing quickly, and professionals in compliance, procurement, and IT governance roles need to be paying attention.

In the European Union, the Energy Efficiency Directive already requires data centers above a certain power threshold to report annually on energy consumption, renewable energy use, and water usage to a public EU database. A 2026 draft regulation goes further, establishing a mandatory sustainability rating and label for data centers, similar in spirit to an energy-efficiency label on a household appliance, that is expected to influence everything from green financing eligibility to public procurement decisions. Separately, the EU AI Act requires providers of general-purpose AI models to report on the energy consumption used to train those models, and tasks EU regulators with developing standards for energy-efficient AI deployment going forward.

This regulatory direction matters even outside the EU, because multinational organizations sourcing AI infrastructure or vendors that operate in Europe will increasingly need to meet these disclosure standards regardless of where their own headquarters sit. Professionals responsible for vendor risk, ESG reporting, or IT governance should expect “AI sustainability disclosure” to become a standard line item in procurement and audit checklists over the next few years.

You do not need to be a data center engineer to influence AI’s environmental footprint. The decisions that matter most are often made by the people choosing, deploying, and governing AI tools, not the people building the underlying infrastructure.

The organizations that get ahead of this shift will not be the ones scrambling to respond to a regulation or a headline. They will be the ones who built sustainable digital practice into how their teams already work with data and AI. That is precisely the gap DASCIN’s Green IT Framework was built to close: a vendor-neutral methodology for understanding and reducing the environmental footprint of enterprise technology, including AI.

If this topic is new to your team, the Green IT Awareness credential is a practical starting point for non-technical professionals in ESG, operations, and procurement roles who need a working understanding of digital sustainability without deep technical training. Teams ready to go further can pursue Green IT Foundation to build the skills to actually implement a Green IT strategy, or have their organization’s current practices independently reviewed through a Green IT Audit.

For a closer look at these ideas in practice, watch our previous session, Making AI More Sustainable: What Every Professional Should Know, where DASCIN President Jan Willem Middelburg walks through how non-technical teams can meaningfully contribute to sustainability goals through everyday digital habits and decisions.

For more in-depth learning, read the related article: What Is the Carbon Footprint of AI? A Data Center Energy, Water, and Emissions Primer.