Science Friday

As Companies Build Data Centers For AI, Communities Push Back

December 5, 2025

Key Takeaways Copied to clipboard!

  • The massive buildout of data centers to fuel the AI boom is causing significant strain on global electricity consumption, projected to double by 2030, with 2025 AI/data center investment ($580 billion) already surpassing global oil supply development spending. 
  • Data centers, particularly those using evaporative cooling, consume substantial amounts of water, often requiring drinking-quality water, leading to local grid stress in water-stressed areas like Arizona and Nevada, despite ongoing research into less water-intensive cooling methods like direct liquid or immersion cooling. 
  • Community pushback against rising utility rates driven by data center energy demands is leading to project delays and cancellations, shifting the conversation from individual AI usage choices to a necessary systems-level discussion about digital infrastructure growth versus energy and water sustainability. 

Segments

Data Center Scale and Cost
Copied to clipboard!
(00:01:16)
  • Key Takeaway: Data center construction for AI is consuming resources at a scale exceeding previous drains like crypto mining, exemplified by a single Meta facility requiring three times New Orleans’ annual power.
  • Summary: The buildout of data centers for the AI boom involves hundreds of billions in spending, with trillions more anticipated. A new Meta data center in Louisiana is projected to require nearly three times the power consumed annually by New Orleans. This rapid expansion is already causing public concern, evidenced by rising utility rates becoming an election issue in some areas.
AI Query Energy Consumption
Copied to clipboard!
(00:03:14)
  • Key Takeaway: Individual AI queries consume measurable energy, with Google’s Gemini averaging 0.24 watt-hours and OpenAI’s ChatGPT around 0.34 watt-hours per query.
  • Summary: While individual AI queries use a relatively small amount of energy (comparable to a second in a microwave), the sheer volume of billions of daily queries results in significant total energy usage. Companies are now providing estimates for single query consumption, though the total energy footprint for all AI activities remains a larger, less clear figure.
Water Use in Data Centers
Copied to clipboard!
(00:04:39)
  • Key Takeaway: Over 60% of water consumption related to AI often comes indirectly from the power plants running the data centers, but direct cooling systems frequently require drinking-quality water to prevent pipe clogging via evaporation.
  • Summary: Data centers are often sited in water-stressed regions like Arizona and Nevada due to the search for cheap energy and land. While direct data center water use is currently about 0.3% of the nation’s total, this is expected to double by 2030, potentially impacting local grids severely. Companies are exploring alternatives to evaporative cooling, such as direct liquid or immersion cooling, to mitigate water loss, though these methods can sometimes increase electricity intensity.
Political Pressure and Climate Pledges
Copied to clipboard!
(00:08:26)
  • Key Takeaway: Political pressure regarding high utility rates is causing significant project delays, with $93 billion in data center projects halted between March and June due to community pushback.
  • Summary: Elected officials in states with high data center concentrations, like Virginia, are campaigning on lowering utility rates, reflecting public sensitivity to rising costs. Some tech companies have reduced reporting on climate pledges as their energy demands increase sharply due to the AI buildout. The rapid growth of AI deployment outpaces the slow, multi-year timelines required to build new energy infrastructure, including nuclear power.
Systems Conversation Needed
Copied to clipboard!
(00:11:11)
  • Key Takeaway: The impact of AI energy and water use requires a systems-level conversation about digital infrastructure rather than focusing on personal choices, as AI is rapidly becoming integrated into essential digital services.
  • Summary: The dependency on AI is quickly becoming non-optional, as it influences Google searches, ad serving, and other digital infrastructure elements. Listeners should view this as a systemic challenge involving trade-offs between water/climate impact and the new lifestyle AI enables. The speed of AI adoption far exceeds the speed at which the complex energy system can adapt and build capacity.