AI is becoming an energy story
The artificial-intelligence boom is no longer only about software, chips and algorithms. It is becoming a test of whether our energy, infrastructure and governance systems can keep pace with the technologies we are building.
WHAT WE ARE SEEING
Around the world, technology companies are racing to build enormous data centers to support artificial intelligence. Today, Anthropic announced its first Australian data-center agreement, involving a planned 2.16-gigawatt campus near Brisbane. Last week, Google announced at least €13 billion in AI infrastructure investment in Finland, including nuclear-power supply. Meanwhile, the U.S. Energy Information Administration expects American electricity consumption to reach record highs in both 2026 and 2027, with AI-intensive data centers among the major drivers.
What looks like a technology race is therefore becoming something much larger:
a race for electricity, transmission capacity, land, water, cooling equipment, capital and public permission.
WHAT IS REALLY GOING ON
For most of the digital era, people experienced computing as something almost weightless.
Search engines, streaming services, cloud storage and social networks appeared on screens. The physical infrastructure behind them remained largely invisible.
Artificial intelligence is making that hidden infrastructure impossible to ignore.
Advanced AI requires enormous computing capacity. Computing requires data centers. Data centers require electricity, cooling systems, transmission lines, transformers, land and communications infrastructure.
That means AI is colliding with systems that were built for a very different economy.
Electric grids were designed around relatively predictable growth in households, factories and businesses. They were not necessarily designed for clusters of facilities that can each demand the electricity of a small city.
The result is a fundamental systems question:
Who should adapt to whom?
Should communities and power systems reorganize themselves around rapidly expanding AI infrastructure?
Or should AI infrastructure be designed around the ecological, economic and social limits of the communities hosting it?
That debate is only beginning.
HOW THE SYSTEM WORKS
The current system encourages speed.
AI companies compete to develop more capable models. Cloud providers compete to supply the computing power. Investors finance new infrastructure. Utilities receive requests for enormous amounts of electricity.
Each participant has an incentive to move quickly.
But the consequences are distributed across systems.
A developer requesting power for a data center may be thinking about computing capacity.
A utility must think about transmission lines, generation and reliability.
A city must think about land and water.
Residents may be thinking about electricity prices.
Governments may be thinking about economic competitiveness.
Climate planners may be thinking about emissions.
The problem is not simply that AI uses electricity.
The deeper problem is that different institutions make decisions about different pieces of the same system.
That fragmentation can produce distorted signals.
In the United States, requested data-center connections have reportedly exceeded 700 gigawatts—far beyond current actual consumption. Some projects may never be built, yet utilities may still have to plan infrastructure around those requests. Texas has begun scrutinizing proposed projects more closely, while other states have introduced requirements intended to distinguish serious projects from speculative demand.
This is what systems failure often looks like.
Not one bad decision.
Many rational decisions interacting badly.
THE CONNECTIONS
Energy. AI increasingly depends on access to reliable electricity. That makes grid capacity, renewable generation, nuclear power, natural gas and transmission infrastructure part of the AI economy.
Water and resources. Data centers can require significant cooling infrastructure. In water-stressed regions, computing may therefore compete indirectly with agriculture, households and ecosystems.
Finance. Building AI infrastructure requires enormous capital. Data-center companies are increasingly tapping banks, institutional investors and infrastructure financing markets. Decisions about AI development are therefore becoming decisions about where global capital flows.
Cities and communities. Local governments decide zoning, infrastructure and development approvals. Communities increasingly want to know whether large data centers create enough jobs, tax revenue and public benefit to justify their demands on local resources.
Climate. AI can help optimize energy systems, scientific research and industrial efficiency. But research also warns that productivity gains enabled by AI could stimulate additional economic activity and energy consumption, potentially offsetting some environmental benefits.
The AI debate therefore cannot remain inside the technology sector.
AI is becoming a whole-system infrastructure question.
THE HUMAN-RIGHTS CONNECTION
Several principles of the United Nations Universal Declaration of Human Rights become relevant.
Article 25 — an adequate standard of living. Energy affordability, water availability and environmental conditions directly affect people’s ability to maintain healthy lives.
Article 27 — participation in scientific advancement. People should benefit from scientific and technological progress.
But there is an important systems question embedded inside that principle:
Who receives the benefits of technological progress, and who absorbs its costs?
If communities supply land, water, electricity and infrastructure while the economic gains concentrate elsewhere, technological progress can deepen inequality rather than reduce it.
Human rights therefore provide a useful design principle:
Technology should expand human capability without undermining the basic conditions people need to live.
WHAT IS CHANGING
Governments are beginning to move from simply attracting data centers toward setting conditions for them.
Authorities in several places have considered restrictions, moratoriums or new requirements because of concerns about electricity prices, water, land and grid capacity.
At the same time, companies are experimenting with different models: locating facilities near abundant renewable energy, contracting directly for nuclear power, improving cooling efficiency and building infrastructure in regions with favorable climates.
The next phase may therefore shift from:
Build the biggest data center possible
to:
Build computing infrastructure that fits within the energy, ecological and community system around it.
That would represent an important change in thinking.
CHANGE IN ACTION
Australia: designing around resource constraints.
Anthropic’s planned Australian facility is expected to use renewable electricity and closed-loop air cooling designed to reduce water consumption. The project still requires regulatory approval, but the design illustrates how computing infrastructure can begin responding to local environmental constraints rather than treating energy and water as unlimited inputs.
Texas: questioning “ghost demand.”
Texas has temporarily slowed some new data-center grid connections while examining whether proposed projects are financially and technically credible. Requiring developers to demonstrate that projects are real before utilities build infrastructure around them is a relatively simple systems intervention: improve the information entering the system, and planning becomes more rational.
WHAT TO WATCH
Watch electricity before you watch algorithms.
The most important signals may increasingly be grid connections, transmission construction, power-generation agreements, water restrictions and community-benefit requirements.
Also watch who pays.
If utilities must build billions of dollars of infrastructure to serve data centers, regulators will have to determine whether those costs belong primarily to technology companies or should be spread across ordinary electricity customers.
Finally, watch geography.
Regions with abundant clean electricity, strong transmission systems, cooler climates and reliable communications infrastructure could become strategic centers of the AI economy.
The geography of computing may increasingly follow the geography of energy.
WHAT YOU CAN DO
Follow the infrastructure. When you hear about a major AI investment, ask where its electricity, water, land and financing will come from.
Ask who benefits and who pays. Economic-development announcements should include community costs as well as promised investment.
Support transparent planning. Communities should be able to see projected electricity and water requirements before major infrastructure projects are approved.
Connect the conversations. Energy planners, technologists, environmental experts, local governments and communities should not be discussing these projects separately.
THE BIG PICTURE
The defining question of artificial intelligence may ultimately be bigger than whether machines become more intelligent.
It may be whether our systems become more intelligent.
AI is revealing something we often forget: technology does not exist outside society. Every digital system ultimately rests on physical systems—energy, water, materials, finance, communities and governance.
The opportunity is not simply to build more computing power.
It is to redesign the systems surrounding that computing power so technological progress improves the quality of life rather than competing with it.
The real AI revolution may be learning to design technology and society as one interconnected system.