Data centers powering artificial intelligence don't have to be environmental disasters. While they've become symbols of AI's carbon footprint, there are clear, actionable strategies to design, build, and operate them sustainably, from spreading them across new geographies to powering them with renewable energy and recycling their waste heat.

Why Are Data Centers Concentrated in Just a Few Places?

Currently, data centers cluster heavily in a handful of regions: Virginia, Ireland, Texas, and Singapore. This geographic concentration creates severe strain on local energy grids and water supplies. The rapid buildout of data centers demands enormous amounts of energy on compressed timelines, forcing developers to turn to "behind the meter" energy generation, which often relies on fossil fuel-powered turbines and generators that intensely pollute local communities.

Meanwhile, grid operators scramble to upgrade infrastructure while existing customers foot the bill. This pattern has sparked valid community pushback against new data center projects, but the solution isn't to reject data centers outright. Instead, experts argue for smarter design strategies that account for local constraints and spread facilities across diverse geographies.

What Are the Most Promising Solutions for Greener Data Centers?

There are multiple pathways to powering data centers sustainably. Google is developing a geothermal power project in Nevada designed to provide its data centers with renewable energy around the clock. Ireland has taken a regulatory approach, mandating that 80% of energy for new data centers come from renewable sources. These examples show that both corporate commitments and policy frameworks can accelerate the transition.

Beyond energy sourcing, data centers can become more sustainable throughout their entire lifecycle. During design and construction, developers can choose sustainable materials like timber and low-carbon concrete to reduce overall emissions. Rather than clearing land for new facilities, reusing vacant factories and industrial buildings preserves ecosystems and takes advantage of existing energy grid and water infrastructure connections, which also speeds up permitting.

One of the most innovative approaches involves capturing and reusing the massive amounts of heat data centers generate. In West London, the Old Oak and Park Royal Development Corporation is recovering 17 megawatts of waste heat from local data centers to warm up to 10,000 homes and businesses. In Norway, similar waste heat is being used to heat a trout farm. These examples demonstrate how data centers can become assets to their host communities rather than just resource consumers.

How to Make AI Models More Efficient and Sustainable

• Model Distillation: Training smaller AI models to mimic the behavior of larger ones, reducing the computing power and energy required to run them effectively.
• Quantization: Reducing the precision of the numbers used in computation, which decreases the amount of processing power needed without significantly sacrificing performance.
• Right-Sizing Models: Choosing appropriately sized AI models for specific tasks instead of defaulting to the largest, most generic models available, which can reduce energy consumption by a factor of 33.

But addressing data center sustainability ultimately requires focusing on what's driving their explosive growth: AI itself. The tech industry's "bigger is better" philosophy needs to shift toward doing more with what already exists. Research has shown that selecting the right-size models for specific tasks, rather than the biggest, most generic models available, can reduce energy consumption by a factor of 33.

"By thinking more holistically about how data centers are built and operated, we can also make them more beneficial to the communities that host them and to the planet as a whole," explained Sasha Luccioni, an AI researcher studying the environmental impacts of the technology.

Sasha Luccioni, AI Researcher

Why Transparency About AI's Energy Use Matters

Transparency will play a crucial role in driving change. Individuals and companies want to choose AI models with sustainability in mind, but they currently lack the data to do so. Imagine a counter that shows how much energy each ChatGPT query uses and how much carbon it emits. That kind of visibility could help users and companies factor sustainability into their everyday AI decisions.

Developers need to expose energy and carbon footprint data directly in AI model interfaces and application programming interfaces (APIs), which are the technical bridges that allow different software systems to communicate. This would empower both users and developers to make ecologically informed choices about the AI they build and use. Code Carbon, a software package developed to measure energy and carbon for open-source models, demonstrates this approach, but widespread adoption by major technology companies remains limited.

The stakes are high. AI is a transformative technology with immense potential, which makes it critical to get its environmental impact right. Society cannot simply accept skyrocketing carbon emissions and rising energy prices as the unavoidable cost of technological progress. Nor should we blanket reject data centers or AI as inherently unsustainable. By centering the conversation around data centers on sustainability, we can make it more nuanced and actionable, ensuring that the infrastructure powering our future doesn't cost us the earth.