The sustainability crisis in the AI industry: How change can happen
The rapid growth of AI has transformed many industries and led to amazing new technology, but it also brings with it a major problem: a massive increase in energy consumption. This increase in energy consumption is not just a technical problem; it is a major environmental problem that everyone in the AI field must address. As AI continues to develop, it is not just about making smarter models and solving more problems for users; it is about ensuring that these developments do not harm our planet. The question is not only what AI can do for us, but how we can ensure that these developments are sustainable for the planet.
Head of Impact and Sustainability at Photoroom.
The current landscape
Gartner predicts that without sustainable AI practices, AI will consume more energy than human labor by 2025, significantly offsetting gains toward carbon neutrality.
According to a recent report from the Federal Energy Regulatory Commission, demand for data centers in the U.S. is expected to reach 35 gigawatts by 2030. That’s equivalent to powering about 26 million homes. (For comparison, 1 GW is enough energy to power about 750,000 homes.)
In regions like Salt Lake City, where energy giants like Meta and Google are building data centers, there’s a noticeable shift back to coal as more data centers are needed to support AI workloads. Plans to close coal plants early are being abandoned, pushing back the data until as late as 2042 and scaling back clean energy sources.
This is a worrying shift that underscores the complex trade-off between technological advancement and sustainability, especially as AI is on track to drive a 160% increase in data center energy demand by 2030.
While some tech giants like Google, Amazon, and Microsoft have pledged to power their data centers with 100% renewable energy by 2030, AI operations still have a significant carbon footprint.
According to public data from Meta, one of its data centers in Iowa uses as much power annually as 7 million laptops left on for eight hours a day.
According to a study by Hugging Face and Carnegie Mellon University, creating an image using generative AI uses as much energy as fully charging your smartphone.
ChatGPT queries consume nearly 10 times as much electricity as a Google search. For one startup, training its AI models in the US uses about 1,000 tons of CO2 per year, which is the equivalent of 1,000 trips from Paris to NYC.
AI needs an energy breakthrough. Industry is exploring solutions like nuclear fusion to accelerate the energy transition from fossil fuels, but until this energy breakthrough happens, people and companies in AI must take individual steps toward change.
Why the AI Industry Has Been Slow to Embrace Sustainable Practices
Companies using AI face challenges in technology, financial investments, and stakeholder engagement as they seek to implement sustainable AI practices.
Transitioning to sustainable AI solutions often requires a substantial upfront investment in energy-efficient technologies and renewable energy sources. According to an IBM sustainability study, the majority of executives (76%) agree that sustainability is central to their business, but nearly half (47%) struggle to finance sustainability investments.
Furthermore, only 31% of organizations report that they are comprehensively integrating sustainability data into their operational improvements, highlighting a gap between sustainability goals and actionable steps.
The shift to green data centers and sustainable hardware requires not only capital, but also a strategic rethink of existing infrastructures. Companies building AI must make complex decisions about upgrading to more efficient systems while managing ongoing operational costs.
This, plus the rapid pace of technological change, can make it difficult for companies to keep up. Many AI companies in the early stages of development may deprioritize sustainability due to the immediate pressures of competition, technological development, and finding product-market fit.
But as demand for AI increases, it becomes increasingly important for companies to integrate sustainability into their decision-making processes to meet environmental goals and drive innovation.
How we can all shape the future of AI sustainability
The entire industry has a role to play in influencing a more sustainable future for AI. The readiness, adoption, and development of green AI practices depend on market maturity and stakeholder engagement.
Venture capitalists can evaluate the environmental impact of their portfolio, request impact statements from companies, and share sustainability best practices to inspire more companies to take action.
Companies and SMEs using AI can request environmental impact reports from suppliers to evaluate sustainability efforts and commitment.
Companies developing AI products can be selective about the type of AI model they use. Recent studies show that specialized AI models consume less energy than general AI models. The more energy efficient it is, the faster it can run, improving the user experience and reducing energy consumption.
Companies developing AI models can partner with green data centers like Genesis Cloud to leverage renewable energy sources and minimize their environmental impact. They can query cloud providers for their data centers’ Power Usage Effectiveness (PUE) scores and even use an open-source tool to measure their cloud carbon footprint. Internally, they can develop more energy-efficient specialized AI models to reduce carbon emissions. Externally, they can publish their model carbon emissions, as Meta did for Llama 3.1.
Cloud providers like Amazon Web Services (AWS), Google Cloud, Scaleway, and Genesis can help reduce the carbon footprint of AI by building infrastructures that maximize energy efficiency and by being transparent. This means sharing their PUE scores, including the energy used to cool the data center and the carbon emissions to build the data center, and potentially offering green pricing options. Data centers can also pass on demand for energy-efficient chips to hardware providers.
Hardware vendors can develop power-efficient chips, such as NVIDIA, which claims its new “superchip” can increase performance for generative AI tasks by a factor of 30 while consuming 25 times less power using new chip cooling techniques.
Government funding can play a role by integrating carbon footprint assessments into decision-making processes.
Regulators can develop into an organization that holds all players in the AI ecosystem accountable.
As an industry, we must take a systematic approach with shared responsibility to reduce the environmental impact of AI at all levels of the ecosystem.
Change can happen now
To stay competitive, it’s critical for businesses to keep up with today’s rapid AI innovation. But as the market matures, sustainability should play a bigger role in the decision-making process.
Take the first step by choosing an AI vendor that is already taking steps to reduce energy consumption. For example, ask about their environmental impact assessments or whether they measure their company’s carbon footprint.
It is up to all of us to take the lead and commit to a more sustainable future for AI.
We have highlighted the best green web hosting for you.
This article was produced as part of TechRadarPro’s Expert Insights channel, where we showcase the best and brightest minds in the technology sector today. The views expressed here are those of the author and do not necessarily represent those of TechRadarPro or Future plc. If you’re interested in contributing, you can read more here:
