TOPIC 3.4

Energy, Power & Sustainability

⏱️30 min read

🌿Sustainability

Data centers consumed approximately 460 TWh globally in 2022— roughly 2% of worldwide electricity. Projections suggest this will double to 1,000 TWh by 2030, driven primarily by AI workloads. This creates a sustainability paradox: remarkable efficiency gains at the facility and chip level are being overwhelmed by exponential demand growth, resulting in rising absolute consumption of energy and water.

The Scale of Energy Consumption

Current Power Demand

Global data center electricity consumption breakdown (2022):

Hyperscale facilities: 200 TWh (43%)
Enterprise data centers: 150 TWh (33%)
Colocation facilities: 70 TWh (15%)
Edge computing: 40 TWh (9%)

For context, 460 TWh equals the entire electricity consumption of the United Kingdom. A single large hyperscale facility consuming 100 MW uses as much power as 80,000 homes.

The AI Acceleration

AI workloads are dramatically more power-intensive than traditional computing. Training a single large language model can consume 10-50 GWh— equivalent to the annual electricity use of 1,000-5,000 homes. The International Energy Agency projects AI and cryptocurrency could drive data center electricity demand to 1,000 TWh by 2030, representing 3-4% of global electricity.

⚡ Data Center Energy Consumption Projection

2020

400 TWh

2022

460 TWh

2025

650 TWh

2030

1,000 TWh

Projected doubling of energy consumption by 2030, driven primarily by AI workloads

Efficiency Metrics: PUE and Beyond

Power Usage Effectiveness (PUE)

PUE measures total facility power divided by IT equipment power. A PUE of 2.0 means for every watt powering servers, another watt is consumed by cooling, lighting, and other overhead. Industry progress:

2007 average: PUE 2.5 (60% overhead)
2022 average: PUE 1.55 (35% overhead)
Best-in-class: PUE 1.1-1.2 (10-20% overhead)
Theoretical minimum: PUE 1.0 (0% overhead)

Google's hyperscale facilities average PUE 1.10, while Meta achieves 1.09 through aggressive free air cooling and custom designs. However, PUE improvements are slowing as facilities approach theoretical limits.

Water Usage Effectiveness (WUE)

Data centers consume massive amounts of water for cooling— approximately 1.8 liters per kWh in facilities using evaporative cooling. A 100 MW facility can consume 5-10 million liters per day, equivalent to a city of 30,000-60,000 people.

Water stress is becoming a critical constraint. Facilities in Virginia, Arizona, and Singapore face increasing scrutiny over water consumption during droughts. This is driving adoption of closed-loop liquid cooling systems that recycle water.

Renewable Energy and Carbon Neutrality

Corporate Commitments

Major cloud providers have made ambitious sustainability pledges:

Google: Carbon-free energy 24/7 by 2030 (currently 64% carbon-free)
Microsoft: Carbon negative by 2030, remove all historical emissions by 2050
Amazon: Net-zero carbon by 2040, 100% renewable energy by 2025
Meta: Net-zero emissions across value chain by 2030

The Reality of Renewable Procurement

Cloud providers are the world's largest corporate buyers of renewable energy, having contracted over 50 GW of wind and solar capacity. However, challenges remain:

Intermittency: Solar/wind don't match 24/7 data center demand
Grid constraints: Renewable energy often generated far from data centers
Accounting methods: Renewable Energy Credits (RECs) allow "virtual" matching rather than physical delivery
Additionality: Questions about whether purchases actually add new renewable capacity

🌱 Renewable Energy Mix (2025)

☀️

Solar

45% of renewable contracts

💨

Wind

40% of renewable contracts

💧

Hydro

10% of renewable contracts

⚛️

Nuclear

5% emerging interest

The Nuclear Renaissance

Growing recognition that intermittent renewables cannot meet 24/7 AI data center demands is driving renewed interest in nuclear power. Microsoft signed a 20-year deal to restart Three Mile Island Unit 1, providing 835 MW of carbon-free baseload power starting 2028.

Small Modular Reactors (SMRs) are being explored as on-site power sources for data centers. These 50-300 MW reactors could provide reliable, carbon-free power without grid dependence. However, regulatory approval, public acceptance, and economics remain significant hurdles.

The Sustainability Paradox

Despite dramatic efficiency improvements— PUE declining from 2.5 to 1.55, chip performance-per-watt improving 100x— absolute energy consumption continues rising due to exponential demand growth. This "rebound effect" means efficiency gains are overwhelmed by scale increases.

The fundamental tension: AI promises solutions to climate change (optimization, materials discovery, grid management) while simultaneously driving massive energy consumption growth. Resolving this paradox requires not just efficiency improvements but fundamental questions about which AI applications justify their environmental cost.

🎯 Key Takeaways

Data centers consumed 460 TWh in 2022 (2% of global electricity), projected to double to 1,000 TWh by 2030 driven by AI workloads— equivalent to adding another United Kingdom's worth of demand
PUE improved from 2.5 (2007) to 1.55 (2022) with best-in-class at 1.1, but efficiency gains are overwhelmed by exponential demand growth creating a "rebound effect"
Major cloud providers contracted 50+ GW of renewable energy targeting carbon neutrality by 2030-2040, but intermittency and grid constraints limit 24/7 matching
Nuclear power renaissance emerging with Microsoft's Three Mile Island restart (835 MW, 2028) and Small Modular Reactor exploration for reliable carbon-free baseload power

[

← Previous Topic AI Infrastructure & LLM Economics

](topic-3.html)[

Next Topic → Cloud Economics & Resilience

](topic-5.html)