Where Technology Meets Humanity.
AI • Technology • Ethics • Society
Where Technology Meets Humanity.
AI • Technology • Ethics • Society
The Latest
World Summit AI
The AI Industry Is Asking to Slow Down. But Who Gets to Decide How Fast Is Too Fast?
OpenAI Calls for National AI Safety Rules
AI Leaders, Researchers and Policymakers Gather to Shape the Future of Artificial Intelligence
OpenAI Calls for National AI Safety Rules
The AI Industry Is Asking to Slow Down. But Who Gets to Decide How Fast Is Too Fast?
OpenAI Calls for National AI Safety Rules
OpenAI is urging U.S. lawmakers to establish mandatory safety requirements for increasingly capable AI systems, adding momentum to a growing debate over government oversight, independent testing, and corporate responsibility.
The AI Industry Is Asking to Slow Down. But Who Gets to Decide How Fast Is Too Fast?
The AI Industry Is Asking to Slow Down. But Who Gets to Decide How Fast Is Too Fast?
The AI Industry Is Asking to Slow Down. But Who Gets to Decide How Fast Is Too Fast?
As AI systems become increasingly capable, some of the industry's biggest leaders are calling for stronger safety measures and a slower approach to frontier development. But the debate raises a difficult question: can the companies building these systems also be trusted to regulate them?
Because the future shouldn't be a one-way conversation.
THE TALK TREE CONVERSATION
Interviews, perspectives and discussions exploring the people behind technology—and the people affected by it.
The Focus
WE WANT AI. BUT DO WE WANT THE INFRASTRUCTURE NEEDED TO POWER IT?
The technology may be digital. The resources it requires are not.
Artificial intelligence is often presented as something that exists in the cloud. But behind every AI model, chatbot, image generator, and autonomous system is a very physical infrastructure: data centers, electricity generation, transmission lines, cooling systems, water, land, and enormous amounts of hardware.
As AI adoption accelerates, that infrastructure is becoming one of the most important and least visible parts of the AI conversation.
The question is no longer simply how powerful AI can become.
It is increasingly becoming:
How much infrastructure are we willing to build to make it possible?
AI has an energy problem
The International Energy Agency estimates that electricity consumption from data centers could roughly double from about 485 terawatt-hours in 2025 to 950 TWh by 2030. Electricity use by AI-focused data centers is expected to grow even faster, potentially tripling over the same period.
The impact is particularly significant in the United States.
The IEA estimates that data centers could account for approximately half of U.S. electricity-demand growth through 2030.
The U.S. Energy Information Administration is also forecasting record U.S. electricity consumption in 2026 and 2027. Data-center development and increased manufacturing activity are among the factors contributing to that growth.
That does not mean AI is solely responsible for rising electricity use. Manufacturing, electrification, cooling, and other industries are also contributing.
But AI is changing the scale and location of electricity demand.
Unlike many forms of electricity consumption, data centers can concentrate enormous loads within particular communities, placing new demands on local grids and infrastructure.
Then there is water.
Electricity is only part of the story.
Data centers generate tremendous amounts of heat, and cooling that equipment can require significant amounts of water depending on the technology and location.
Rystad Energy estimates that data centers consumed approximately 222 billion liters of water directly for cooling in 2025. Without additional water-saving measures, its central forecast puts that figure at nearly 644 billion liters annually by 2030. Rystad also emphasizes that the actual amount varies considerably depending on cooling technology, geography, and whether water withdrawal or actual consumption is being measured.
That distinction matters.
The environmental footprint of an AI data center is not identical everywhere. A facility using closed-loop or dry-cooling technology in one region can have a very different water footprint from a facility using more water-intensive cooling in another.
The U.S. Department of Energy has also highlighted the potential for modern cooling technologies to reduce water consumption compared with traditional evaporative systems.
So the question should not simply be:
“How much water does AI use?”
It should be:
“Where is the data center being built, how is it being cooled, where does its electricity come from, and who bears the environmental cost?”
The community question
This is where the AI infrastructure conversation becomes an issue of society, not simply technology.
Data centers need power. Power requires generation and transmission infrastructure. They also require land, construction, cooling systems, roads, and other supporting infrastructure.
And communities are beginning to ask who should pay for it.
In Texas, regulators have taken steps to examine rapidly growing requests for electricity connections from data centers amid concerns about projects that may not ultimately materialize. Reuters has reported on the issue of so-called “ghost demand,” referring to electricity requests associated with projects that may never come online.
Meanwhile, lawmakers in Washington are examining legislation aimed at preventing data-center expansion from shifting additional electricity costs onto ordinary consumers.
That creates a question that deserves considerably more attention:
If the infrastructure is being built to support an AI economy, who should pay for it?
The technology companies?
Utilities?
Investors?
Local governments?
Or the communities whose electricity systems, water resources, and land are being used?
There is no simple answer.
We shouldn't reject the technology. We should examine the cost.
There is an important distinction between questioning AI's infrastructure and opposing AI itself.
AI has legitimate potential to improve medicine, scientific research, accessibility, education, energy systems, and countless other areas.
And the infrastructure supporting AI can bring economic investment, jobs, and technological development to communities.
The issue is what happens when enthusiasm for technological progress moves faster than the systems responsible for managing its consequences.
The IEA has identified electricity grids, transformers, chips, permitting, and other infrastructure as potential bottlenecks to continued data-center expansion.
That means the future of AI may ultimately depend on something much less futuristic than another breakthrough model.
It may depend on whether we can build the physical infrastructure to support it responsibly.
The question we should be asking
The AI conversation has spent years asking:
What can AI do?
We should also be asking:
What does AI require?
How much electricity?
How much water?
How much land?
How much new infrastructure?
How much investment?
And perhaps most importantly:
Who benefits, and who bears the cost?
AI may live in the cloud.
But the resources required to keep that cloud running exist very much on the ground.
Sources
International Energy Agency (IEA)
Key Questions on Energy and AI
https://www.iea.org/reports/key-questions-on-energy-and-ai/executive-summary
U.S. Energy Information Administration (EIA)
Short-Term Energy Outlook
https://www.eia.gov/outlooks/steo/
Rystad Energy
Tech Thirst: Data Center Water Consumption Could Triple by 2030 Without Efficiency Gains
https://www.rystadenergy.com/news/data-center-water-consumption-could-triple-by-2030
U.S. Department of Energy
Cooling Water Efficiency Opportunities for Federal Data Centers
https://www.energy.gov/cmei/femp/cooling-water-efficiency-opportunities-federal-data-centers
Reuters
Texas' halt on powering data centers reflects U.S. reckoning over “ghost” demand
Reuters
U.S. House to take up bill aimed at curbing data center-driven electricity costs
