AI (Alarming Insights): How Machine Learning Affects Environmental Goals

By Anne Krieger, University of Virginia

To the average American, AI is just a search away, whether you use Chat-GPT, Gemini, Meta AI, or another chatbot. Your phone or computer battery may drain faster than you expected, but there is little downside to this new and rapidly improving way of surfing the internet. Upstream from this convenient new tool, the companies that create, train, and deploy these AI models are expending massive amounts of energy to bring in revenue and power the AI boom that has exploded worldwide. Billions of dollars are being poured into the construction of new data centers and tech giants are contracting for more and longer with energy companies to guarantee they have electricity to meet their growing demand. While many say that AI will help expedite environmental goals, there are still large concerns about the vast amount of energy that AI is consuming and whether Earth should, or even can support it.

One of the sector’s investors hope AI will be essential is environmental initiatives, and they have used that hope to try to offset potential fears of AI’s damage to Earth’s fragile and ailing climate. According to the Harvard Business Review, “Using AI can help companies incorporate renewable power sources, manage better decisions about EV charging infrastructure, and reach their sustainability objectives, all while cutting energy costs.” AI’s ability to process large amounts of data and find areas of improvement is seen as essential to managing the energy transition from non-renewables to renewable energy, as the power grid will need to be fundamentally reworked to provide energy across America. Additionally, AI can help meet other environmental harm reduction goals as corporations integrate AI into their business model. From the S&P Global Corporate Sustainability Assessment, “a Latin American bank us[es] AI to identify native forest areas with high carbon capture potential” and “a chemicals company us[es] AI to help track compliance with environmental rules.” The question arises: Can the current strides in environmental goals and hopes for AI aiding the energy transition outweigh the massive energy cost of creating and maintaining these models?

There is no denying that tech giants have greatly increased their emissions in the past few years, mostly due to AI. The influx of power into data centers and the construction of new data centers has increased CO2 emissions and reliance on natural gas and oil. According to the International Energy Agency, in 2022 data centers already accounted for 2% of global electricity usage. Google’s CO2 emissions increased by 30% since 2020 and its GHG emissions are up 50% since 2019. As data centers proliferate to help fuel AI models, these effects will only increase unless major technological leaps are made. Currently, 40% of data centers’ energy expenditure is spent just on cooling, which also demands massive amounts of water. Building data centers in cooler environments could potentially offset some of these cooling costs, but many choose to build data centers where the infrastructure to build, maintain, and access them already exists. Therefore, the US data center capital is concentrated in northern Virginia, where summer temperatures consistently roam between the 90s and low 100s Fahrenheit, as convenience and profitability take precedence over reducing environmental harm. Power demands on these data centers are only expected to grow, says S&P: “Projections from S&P Global’s 451 Research show that hyperscaler and leased data center power demand doubled from 2020 to 2024. This demand is set to grow even faster (137%) through 2029 as AI-driven computing grows.” The damage of this energy demand growth could be partially offset by using renewable sources, but the infrastructure for that is not yet in place and tech giants are yet to prove full commitment to renewables, despite promising steps.

A flashy headline came across newspapers in September this year—Three Mile Island announced that it would reopen following its infamous stint providing nuclear finally ended in 2019. The plant was found to simply be too expensive. Behind this reopening sits one of the world’s largest hyper-scalers, Microsoft. The power plant is expected to open its doors again in 2028 and has a twenty-year-long contract with Microsoft to power its data centers. This is one of the largest and most aggressive commitments powering the AI boom yet. Nuclear has long been seen as an idealistic, but expensive and problematic solution to our energy crisis. New investments into other plants by tech companies also show that the AI demand for energy is pushing companies to seek riskier alternatives and potentially proves that the capital invested into AI is paying off in a major way as companies are willing to now invest in more expensive clean energy sources. However, the reliance on natural gas and oil stays strong as new data centers are being built in the “sweet spot for gas delivery” according to COO Will Brown of Kinder Morgan Inc., one of the midstream giants of American natural gas. Natural gas is expected to expand with the data center building and some natural gas execs are projecting as much as triple the growth in demand by 2040, as said by an executive VP of Williams. However, data centers are some of the largest buyers of PPA agreements, which allow them to buy energy long-term at a pre-set price, exempting them from the unpredictable fluctuations of the commodity market. According to Forbes, “In 2021, Amazon and Microsoft were the two largest corporate buyers of renewable energy through PPA” and this extreme demand and power that tech giants hold is helping switch the tide from non-renewable to renewable energy, as long as they stay committed to moving towards renewables.

We cannot know what the long-term effects of AI will be on the environment, and some perspectives are “rosier” than others. However, what is clear is that how hyper-scalers buy their energy is key to mitigating the environmental damage done by AI. The development and management of data centers cannot be understated. Currently, tech companies are fighting to meet their energy needs and cannot be picky about where the energy is coming from. This high demand can help lead to investment in new, better, cleaner energy now as companies are happy to throw capital into promising projects. If clean energy can start to meet these high demands in the coming years, AI might have helped spur the energy transition based on necessity alone. If clean energy fails to meet its targets, tech companies may double down on traditional and more reliable non-renewables. This is a rapidly changing industry with huge potential for growth and progress, but also very real environmental concerns as the planet struggles towards the daunting goal of net zero.

Cover photo taken from StockCake.