Transcript notes: Data center water use, AI energy demand, and policy trends

Data Center Water Use, AI Energy Demand, and Policy Trends

• Key players and context

  • QTS: a big data center operator that rents space for IT infrastructure. They are investing heavily in water-saving technologies.
  • Aurora water official: highlighted the water use in landscaping for a data center site, versus cooling the servers.
  • Nationwide scope: researchers tracked and discussed roughly 1,200+ data centers and their power/water footprints.

• Water usage breakdown in data center sites

  • Example claim: some data centers use more water to irrigate grass outside than they use to cool servers inside.
  • Specific landscaping water claim:
  • Landscaping at a site uses $1{,}100{,}000$ gallons of water per year, which is double the amount used by the building itself.
  • If cooling were accomplished with solely air conditioning water, the total power use would still rise substantially, yielding no real win in water efficiency here.
  • Implication: there is a significant water footprint associated with landscaping at data center campuses, not just the cooling load.

• Scale of data center power demand

  • The data centers tracked: $1{,}200^{+}$ data centers.
  • Combined power demand projected: they could soon consume more power than Poland did in 2023.
  • AI-driven data centers projection: data centers driven by AI could use as much as $600\,\mathrm{TWh}$ of power by $2028$.
  • Interpretive note: these figures illustrate a magnitude of energy demand that has not been seen before across the country and underscores the rapid growth associated with AI workloads.

• Energy mix and policy shifts amid AI growth

  • Trend: as states race to support the AI boom, some are reversing prior green energy promises and reverting to coal and natural gas for power generation.
  • Nebraska example: the two largest electricity utility companies in Nebraska committed several years ago to net zero targets.
  • Broader implication: the push for AI-enabled energy growth may be clashing with earlier decarbonization commitments, creating tensions between economic/technological goals and environmental goals.

• Key concepts and connections

  • Water-energy nexus: data center operations illustrate how cooling, landscaping, and other site needs contribute to both water and energy demands.
  • Resource trade-offs: even when water-saving measures are implemented, overall energy use can rise if cooling approaches change, revealing complex trade-offs between water efficiency and energy efficiency.
  • Externalities and policy signals: local water resources (e.g., landscaping water) and regional electricity generation mix (coal/natural gas vs. renewables) create externalities that affect communities and environments, informing regulatory and policy discussions.
  • AI demand impact: the projected 600 TWh by 2028 reflects how computational workloads associated with AI escalate electricity demand, with implications for grid reliability, electricity pricing, and infrastructure planning.

• Numerical references (summary in context)

  • Landscaping water use: $1{,}100{,}000$ gallons per year.
  • Comparison: landscaping water use is double the building’s own water use.
  • Data centers tracked: $1{,}200^{+}$ centers.
  • Cumulative power need projection for AI data centers: $600\,\mathrm{TWh}$ by $2028$.

• Examples and hypothetical scenarios

  • Example scenario: If a data center campus allocates water to irrigate extensive landscaping, the site’s water footprint can rival or exceed the cooling load, illustrating the importance of integrated site planning (water, energy, and land use).
  • Hypothetical contrast: a campus that substitutes water-intensive landscaping with drought-tolerant landscaping could reduce water use, but needs to ensure cooling efficiency remains high to avoid increasing energy consumption.

• Ethical, philosophical, and practical implications

  • Environmental equity: high water use in landscaping at data centers may strain local water resources, especially in water-scarce regions.
  • Climate and sustainability: growth in AI-driven workloads may push electricity demand up while policy pressures push for decarbonization; balancing reliability, affordability, and environmental goals is critical.
  • Policy tension: a push for AI-enabled growth could conflict with green energy commitments, raising questions about how to align industry development with long-term climate objectives.

• Connections to foundational principles

  • Energy efficiency vs. water efficiency: improvements in one domain can unintentionally affect another (e.g., air-conditioning water use vs. total power draw).
  • Systems thinking: data center campuses are complex systems where cooling, landscaping, energy supply, and policy environment interact.
  • Real-world relevance: these issues tie directly to national and regional energy planning, water resource management, and environmental policy amid rapid AI expansion.

• Key takeaways for exam-style understanding

  • Data center water and energy footprints extend beyond server cooling to landscaping and on-site infrastructure.
  • AI-driven workloads are expected to dramatically increase electricity demand by 2028 (
    $600\,\mathrm{TWh}$).
  • There is a potential policy tension between supporting AI-driven growth and maintaining green energy commitments, evidenced by examples such as Nebraska’s net-zero pledges in the context of rising energy demand.

• Quick recap of the most important figures

  • Landscaping water: $1{,}100{,}000$ gallons/year (double the building’s water use).
  • Data centers tracked: $1{,}200^{+}$ facilities.
  • AI-related energy projection: $600\,\mathrm{TWh}$ by $2028$.
  • US footprint for QTS: roughly $30$ facilities in the U.S. (contextual reference to scale).