AI’s Power Problem: What It Means for a Responsible Investor

While much of the public focus has been on what AI can do, less attention has been given to the energy needed to make it all possible. For investors focused on sustainability, this emerging dynamic is becoming increasingly relevant, both operationally and in how we assess and engage with companies.

From Algorithms to Gigawatts
Forecasts from the International Energy Agency, Goldman Sachs, and RAND paint a clear picture: the infrastructure underpinning AI is on track to become a major consumer of electricity. At present, data centers account for roughly 1–2% of global electricity use. The IEA expects this figure to more than double by 2030 as AI workloads expand (IEA, 2024).

Goldman Sachs estimates that global data center power demand will rise by 29 gigawatts (GW) by 2027 and by 67 GW by 2030 compared to 2023 levels (Goldman Sachs, 2024). RAND outlines even more aggressive scenarios, where power requirements could reach 327 GW by the end of the decade (RAND Corporation, 2024).

To put these numbers into context: 1 GW is approximately the output of a large nuclear power plant. Building such a plant typically takes between seven and ten years (U.S. Department of Energy, 2021). The scale of energy infrastructure that would be needed to support projected AI growth is, in a word, vast.
This demand is not abstract. It stems from specific technical needs: training ever-larger machine learning models, running inference across billions of queries, and keeping data accessible around the clock. Each innovation in AI capability tends to require more computing power, not less.

AI and Our Operational Emissions
For firms like PGGM, AI is not just an external trend. Internally, we are increasingly using AI to support investment research, automate reporting, and enhance data analysis. These tools are often cloud-based and computationally intensive. While they may offer efficiency gains, they also contribute to our Scope 2 emissions through increased electricity use.

This shift is subtle but important. As the carbon footprint of digital services grows, organizations with climate targets will need to pay closer attention to how AI adoption intersects with their operational emissions. The emissions may be indirect, but they are real—and likely to grow over time.

A Shifting Landscape for Stewardship
At the same time, the rise of AI is beginning to reshape expectations around active ownership. Investors are starting to question how companies account for the energy demands of AI in their broader climate strategies. For some sectors, such as cloud infrastructure, semiconductors, or real estate, this issue is already material.

Questions are emerging around how firms monitor AI-related emissions, optimize efficiency, and secure low-carbon energy sources to power expanding compute loads.

There is no universal framework for these disclosures yet, but the landscape is evolving. Engagements are gradually moving from general climate ambitions to more specific dialogues about digital infrastructure, compute intensity, and energy sourcing. These conversations are still in their early stages, but they hint at a broader shift: as AI becomes embedded in business models, its environmental implications will become a more regular part of stewardship and ESG analysis.

References
Goldman Sachs. (2024). *AI to drive 165% increase in data center power demand by 2030*. https://www.goldmansachs.com/insights/articles/ai-to-drive-165-increase-in-data-center-power-demand-by-2030
International Energy Agency. (2024). *AI is set to drive surging electricity demand from data centres*. https://www.iea.org/news/ai-is-set-to-drive-surging-electricity-demand-from-data-centres-while-offering-the-potential-to-transform-how-the-energy-sector-works
RAND Corporation. (2024). *The Energy and Environmental Implications of Artificial Intelligence*. https://www.rand.org/content/dam/rand/pubs/research_reports/RRA3500/RRA3572-1/RAND_RRA3572-1.pdf
U.S. Department of Energy. (2021). *Nuclear power is the most reliable energy source and it’s not even close*. https://www.energy.gov/ne/articles/nuclear-power-most-reliable-energy-source-and-its-not-even-close