Big Tech spent the last few years telling everyone they’d save the planet. Now shareholders want to know why the planet-saving companies need enough water to fill a small sea just to run their AI models.

Activist investors are ramping up pressure on Amazon, Alphabet, and Meta to come clean about the electricity and water demands of their rapidly expanding AI data center operations. The push centers on shareholder resolutions demanding transparency about how these massive infrastructure buildouts square with the companies’ own climate commitments.

The numbers behind the noise

Here’s the thing about training and running large AI models: it requires an almost comical amount of resources. North American data centers alone are projected to consume nearly 1 trillion liters of water by 2025. That’s not a typo. One trillion liters.

Google’s water consumption tells a particularly striking story. The company’s water use is estimated to have increased by 51% between 2020 and 2024, reaching 5,637 megaliters. For context, that’s enough to supply over 13,000 homes.

On the electricity side, the figures are equally eye-popping. Advocacy group Green America claims AI currently consumes enough electricity to power 7 million US homes. And the trajectory is steep: the group projects AI’s electricity appetite could rise to serve the equivalent of 22% of US households by 2028.

Microsoft, which isn’t named in this particular shareholder push but operates in the same arena, projects its data center electricity usage may more than double by 2030. The company simultaneously maintains a commitment to being carbon negative by that same year. Squaring those two ambitions is, to put it gently, a challenging math problem.

What shareholders actually want

The activist investors aren’t necessarily demanding these companies shut down their AI operations. That would be like asking Netflix to stop streaming. Instead, they want disclosure.

The shareholder resolutions focus on getting Amazon, Alphabet, and Meta to provide clear, detailed reporting on the water and energy footprint of their AI data centers. The underlying question is straightforward: if you made public climate commitments, show us how building dozens of new power-hungry facilities fits into that plan.

This isn’t a fringe movement. It reflects a growing institutional concern that the economic gold rush around AI is outpacing any serious environmental accounting. Investors want to understand the operational risks, too. When a data center is competing with local communities for water during a drought, that’s not just an environmental problem. It’s a business risk.

Local communities near planned or existing data center sites have increasingly raised alarms about strain on power grids and water systems. The tension between Big Tech’s expansion plans and local resource availability is becoming a recurring storyline across the US.

Big Tech’s defense

Tech executives aren’t exactly sitting quietly through this. The standard response involves a two-part argument: first, that they’re investing heavily in cleaner energy sources and efficiency improvements for their facilities; and second, that AI itself can help mitigate emissions in other sectors of the economy.

Google and Microsoft have both pointed to AI’s potential applications in optimizing energy grids, improving agricultural efficiency, and accelerating climate research. The pitch is essentially that the energy AI consumes will be more than offset by the emissions reductions it enables elsewhere.

It’s a compelling argument in theory. In practice, those offsets are speculative and future-tense, while the resource consumption is very much present-tense and measurable. Shareholders, understandably, want the receipts.

Companies have also highlighted investments in renewable energy procurement and next-generation cooling technologies that reduce water dependence. But the pace of AI infrastructure buildout has consistently outstripped the pace of these mitigation efforts, which is precisely why the gap between promise and performance is widening.

What this means for investors

The intersection of AI and environmental, social, and governance (ESG) concerns is becoming one of the more consequential themes in tech investing. For shareholders in Amazon, Alphabet, and Meta, the question isn’t whether AI is a good business. It obviously is. The question is whether the environmental liabilities accumulating alongside AI growth are being adequately priced and managed.

If regulatory bodies start imposing stricter requirements on data center water usage or carbon emissions, companies that haven’t planned for those constraints could face significant operational disruptions. The activist investor push is, in some ways, a preview of the regulatory pressure that may follow.

This dynamic also has ripple effects beyond Big Tech. Data center energy demand influences the broader computing environment, including cloud infrastructure pricing and, notably, the crypto mining sector, which competes for many of the same energy resources. As AI data centers consume an ever-larger share of available power capacity, that competition is likely to intensify.

Look, nobody expects Amazon or Google to voluntarily slow their AI buildouts. The competitive incentives are too strong. But the growing chorus of shareholder activism signals that the era of unchecked AI expansion, where companies could spend billions on infrastructure without detailed environmental accounting, is starting to close. Investors who ignore the resource cost side of the AI equation are pricing in the upside while ignoring a material and growing category of risk.