📨 AI for Social Impact Deep Dive: Environmental Impact

✍🏼 A Note From the Editor

Aside from AI ethics, the environmental impacts of AI seems to be top of mind for most social impact professionals. Every choice we make (technological and otherwise!) shapes the world we leave behind. Let's dive into the complex relationship between AI and our planet. 🌱

😵‍💫 TL;DR

New! Check out the AI-generated podcast of this issue, created in NotebookLM.

🏭 Resource Usage: The Ugly

“The environmental impacts of AI technologies are undeniable - from the energy to power model training, and deployment, to the water needed to cool data centers and the rare earth metals to manufacture the hardware.” Dr. Sasha Luccioni, Climate Lead at Hugging Face.

💎 Minerals: AI depends on a complex ecosystem of minerals, from the silicon that forms microchips to rare earth elements that power high-performance semiconductors. Mineral supply chains, in addition to the extractive practices of mining and the resulting environmental destruction, have also become a strategic priority as nations race for digital sovereignty, often at the expense of communities where mining occurs. For example, the Democratic Republic of the Congo produces 70% of global cobalt supplies, and the vast majority of Congolese cobalt mines are Chinese-owned, underscoring how access to resources is as much about geopolitics and power as it is about the environment.

💦 Water: AI data centers' water use comes in two forms: direct and indirect use. Directly, data centers are cooled by water evaporation that results in water being lost to the atmosphere. Indirectly, water is used to power the electricity that powers data centers, often making up 80% or more of overall water use. And as AI data centers expand, their water usage can strain communities. For example, in the US state of Georgia, data centers is a booming industry, but their water infrastructure was not built for the resulting surge in consumption.

⚡️ Electricity: According to the International Energy Agency, data centers are responsible for just 1% of global energy consumption and 0.5% of CO2 emissions. Currently, fossil fuels provide nearly 60% of power to data centers globally, with coal being the biggest single electricity source, largely due to the numerous facilities in China. Projections of AI-driven expansion vary dramatically because there is no global requirement for data center energy reporting. With few governments tracking usage, the true scale of carbon and grid impacts remains uncertain.

☢️ Nuclear Power: As AI infrastructure demands escalate, nuclear energy is re-emerging as an option to provide reliable, low-carbon power. For example, Microsoft recently signed a 20-year power-purchase agreement to relaunch Three Mile Island to supply its AI data centers. That said, obstacles remain. A Reuters analysis notes that meeting AI-driven electricity demand with new or resurrected nuclear plants is not easy given the regulatory and permitting complexity. Even so, projections indicate that in the U.S. and China, nuclear could account for 16–18% of AI data center electricity supply by the mid-2030s, reducing reliance on fossil-fuel generation but creating other environmental vulnerabilities.

 📄 Lack of Regulation: The Bad

Here's where it gets frustrating: unlike other industries with clear environmental reporting requirements, Big Tech operates in a regulatory wild west when it comes to AI's environmental impact.

While the EU's AI Act—the world's first comprehensive AI regulation—requires that AI systems be "environmentally friendly," the actual environmental provisions are lacking. According to a legal analysis published in Minds and Machines, the Act's environmental protections are among its “less green pieces” of the whole legislation. As AI’s environmental footprint grows, the absence of strong environmental regulation leaves major questions about accountability.

In the United States, progress remains stalled. The Artificial Intelligence Environmental Impacts Act of 2024, introduced by Senators Ed Markey and Martin Heinrich, would require the EPA to study AI's environmental impacts and create voluntary reporting guidelines. But the bill hasn’t advanced since 2024, an example of how legislative action often lags behind technological deployment. As Senator Markey noted when introducing the bill, "There is a Dickensian quality to the use of AI when it comes to our environment: It can make our planet better, and it can make our planet worse."

To see the latest on other AI legislation globally, check out OECD.AI, a central resource for tracking public AI policies.

Finally, there is also a huge lack of transparency when it comes to the actual data on the environmental impacts of AI. While companies like Google and Microsoft publish sustainability reports, they rarely break down energy consumption by specific AI operations. OpenAI, despite ChatGPT's massive usage, provides virtually no public data on their environmental footprint. This lack of disclosure makes it challenging for policymakers, researchers, or the public to assess AI’s true environmental costs. Interestingly, a recent Cornell University research paper introduced a novel benchmarking framework for quantifying the environmental footprint of LLMs.

🌳 AI for Conservation: The Good

Now for the hope! AI is also becoming a powerful ally in environmental conservation.

This Stanford Social Innovation Review article highlights some use cases and case studies of how AI is already being leveraged for environmental conservation. Some other examples include improving biodiversity protection, protecting and monitoring wildlife with smart camera traps, and mapping deforestation and poaching risks via satellite imagery. Beyond wildlife, AI can analyze water quality, track pollution, and guide sustainable farming practices to reduce environmental harm. And, as noted in our last edition, Climate Policy Radar uses AI models to organize and search through thousands climate law and policy documents from countries all over the world, making climate legislation searchable for policymakers worldwide.

While we are still at the beginning of development, these innovations show how AI, when guided responsibly, can become a powerful tool in safeguarding our planet. 🌎

🌟 Final Reflection

The relationship between AI and our environment clearly isn't black and white… but isn’t that where the social sector thrives? We are practiced in navigating complexity, balancing trade-offs, and shaping solutions in gray areas. So we can move forward with our eyes wide open, be intentional about our AI usage, and acknowledge that powerful tools come with significant responsibilities.

And because AI is here to stay, the future isn't about choosing between AI innovation and environmental sustainability—it's about demanding both. And if anyone can hold that tension and push for better, it's us. 🌱

👋🏼 About AI for Social Impact

I’m Joanna, and I’m on a mission to help folks in the social impact sector understand, experiment with, and responsibly adopt AI. We don’t have time to waste, but we also can’t get left behind.

Let’s move the sector forward together. 💫

♥️ Spread the Love

Spread the love and forward this newsletter to anyone who might benefit from a dose of AI inspo!