Big Tech data center problem: rising energy costs and community backlash

Show summary Hide summary

• How much energy do AI data centers actually need?
• Why Texas and energy-rich hubs are prime locations
• Desert data centers: the Gulf’s fast track—and its limits
• Space-based data centers: a moonshot with big caveats
• When conflict turns servers into targets
• AI versus households: a competition for scarce energy
• Could new energy sources change the calculus?
• Has technology affordability reached an inflection point?

The next wave of artificial intelligence depends on a hidden—but enormous—supply chain: electricity, cooling water and sprawling racks of silicon. As Alphabet, Amazon, Meta and Microsoft announced plans this spring to pour roughly $725 billion into AI infrastructure, the world’s data-center footprint is set to balloon. That growth raises a blunt question: where will the power come from, and what will it cost people and the planet?

The answer is not just technical. It’s geographic, economic and geopolitical. Choices about where to site data centers—from Texas plains to Gulf deserts, from orbit to offshore islands—carry trade-offs that will shape how fast AI can scale and who benefits when it does.

How much energy do AI data centers actually need?

Modern AI isn’t a lightweight app running on a laptop. It’s trained and hosted in massive warehouses filled with specialized processors that chew through petabytes of data. Those chips demand continuous power and aggressive cooling. To put the scale in context:

• Major new projects can require gigawatts of continuous power—for example, a proposed Texas facility known as Stargate is expected to need about 10 gigawatts when fully operational.
• By comparison, some planned nuclear reactors in the U.K. expect combined outputs in the single-digit gigawatt range, showing how concentrated the demand from AI can be.

Cooling alone often requires enormous volumes of water or advanced liquid-cooling systems, adding another layer to the resource equation. These operational needs make data centers not just digital infrastructure but heavy industrial consumers.

Why Texas and energy-rich hubs are prime locations

Data-center builders follow reliable power, permissive regulation and cheap land. Texas has become a magnet for large deployments for those exact reasons. The state offers a mix of renewable generation and fossil fuels, high-capacity transmission corridors and local incentives.

• Pros for sites like Texas: abundant generation capacity, favorable business climate, large grids and proximity to U.S. markets.
• Cons: building transmission and balancing intermittent renewables at large scale is not trivial; localized environmental and water impacts can rise as the footprint expands.

Energy availability helps explain why corporate investment has clustered in certain U.S. regions rather than being evenly distributed worldwide.

Desert data centers: the Gulf’s fast track—and its limits

Another clear trend is the rush to energy-rich Gulf states. Oil-producing monarchies are eager to diversify and host the next generation of compute, and they can rapidly deploy the electricity that hungry data centers want. But the desert presents its own set of problems.

• Water scarcity: many cooling approaches require large amounts of water—scarce in arid regions.
• Environmental hazards: wind-driven sand and dust can degrade sensitive electronics and increase maintenance costs.
• Strategic risks: concentrating critical infrastructure in politically volatile areas increases vulnerability to regional conflict.

These trade-offs mean that while deserts can provide raw power, they may impose long-term operational and resilience costs that temper initial enthusiasm.

Space-based data centers: a moonshot with big caveats

Elon Musk and partners have floated a radical alternative: put data centers into orbit and run them on unfiltered solar energy. Conceptually, space-based solar could deliver continuous high-density power and free centers from many terrestrial constraints.

But the technical hurdles are formidable. Engineers would have to solve radiation exposure, thermal management in vacuum, and the logistics and cost of launching enormous hardware. Even if a prototype is attempted by late 2027, deploying and maintaining fleets of orbital data centers would require breakthroughs in launch economics and in-space reliability.

When conflict turns servers into targets

Physical infrastructure can be—and has been—weaponized. Recent attacks on commercial data centers in Bahrain and Abu Dhabi shut down banking, ride-hailing and food delivery systems in parts of the Gulf for hours or days. Those outages showed how dependent daily life has become on on-the-ground server farms.

The incident highlighted two uncomfortable realities:

1. Cloud services are only as resilient as their physical locations and the security of those sites.
2. Siting critical infrastructure in volatile regions magnifies systemic risk, affecting civilians who rely on digital services for basic transactions.

AI versus households: a competition for scarce energy

Energy is finite at any given moment. In many places, the same grids that would power a sprawling AI campus also heat homes, run hospitals and keep supply chains moving. Where demand spikes or prices surge, ordinary consumers can lose out.

Recent global shocks to energy markets—such as geopolitical tensions in the Middle East—have already pushed prices higher and exposed how fragile some supply arrangements are. That means AI’s rapid expansion could create direct competition between data centers and people for electricity and cooling resources.

Could new energy sources change the calculus?

The most straightforward escape hatch from these constraints is a dramatic shift in power supply: commercially viable nuclear fusion, scalable space solar, or some other high-density, low-footprint energy source. If such technologies arrive at scale, they could untangle energy limits from AI growth.

But those options are still uncertain and years or decades away at best. Until then, decisions about where to locate compute—balanced against water scarcity, grid capacity and geopolitical risk—will determine how fast AI can expand and who bears its costs.

Has technology affordability reached an inflection point?

For decades, consumer technologies got faster, cheaper and more accessible: personal phones, budget air travel and compact computing spread benefits across societies. Today’s AI build-out looks different: it concentrates costs in massive, energy-intensive facilities owned by a handful of corporations.

If infrastructure and energy intensity remain as central as they are now, the path toward broadly diffused benefits from AI may be less automatic than past technological waves suggested. That raises political and economic questions about who pays, who profits and what limits governments and communities will impose on an AI-driven future.

You might also like:

• Offshore wind turbine prototype powers onboard data center while ocean disperses heat
• Tennessee requires data center owners to pay full electricity and infrastructure costs
• Seine in Paris cools 800 buildings in summer: river cooling draws public pride
• No electricity, no complex tech: this new cooling system drops 25°C to 3.6°C in 20 minutes
• 86 MW solar farm: Singapore to build on its largest reservoir