Signal · AI Infrastructure
AI Data Centers Drive Surge in Power Demand Amid 18% Global Cooling Market Growth
AI Data Center Energy Consumption and Power Demand Growth in the US and Global Markets Recent OSINT indicates a significant increase in datacenter energy use driven by hyperscaler AI buildouts, with US demand projected to reach 260 TWh by 2026 and global…
Thesis
AI Data Center Energy Consumption and Power Demand Growth in the US and Global Markets
Key findings
- Recent OSINT indicates a significant increase in datacenter energy use driven by hyperscaler AI buildouts, with US demand projected to reach 260 TWh by 2026 and global cooling markets expanding at 18% YoY. These developments reflect heightened power consumption and infrastructure scaling within AI and digital asset sectors.
- Over the past 72 hours, data points reveal major hyperscale expansions by Microsoft, AWS, and Meta, alongside rising grid strain in regions like Virginia and Texas, driven by AI workload growth and crypto load integration.
- US datacenter electricity demand has been revised upward to project a 70% increase from 2022 to 2026, with most of the growth attributed to AI workloads according to the EIA. Microsoft’s AI datacenter capacity under construction has increased to 6.5 GW, representing a fivefold rise since 2021, primarily in the US and Ireland.
Analysis
Meta’s new AI datacenter in Kansas City, valued at $800 million and fully powered by renewables, aims to support Meta’s plan to add 2 GW of AI-optimized datacenter load by 2026. Meanwhile, Dominion Energy reports a 250% YoY increase in AI datacenter load requests in Virginia, warning of potential grid constraints in the “Data Center Alley.”
AWS confirmed a $10 billion expansion in Mississippi, with an estimated 1 GW power draw, primarily driven by AI training clusters. ERCOT forecasts a 135% rise in data center power demand from 2023 to 2028, including AI and crypto loads, with grid capacity additions lagging behind.
The global datacenter cooling market is projected to reach $12.6 billion in 2024, growing 18% YoY, driven by increased rack densities exceeding 50 kW and accelerating liquid cooling adoption.
AI compute intensity remains high, with models like GPT-4 consuming 5 to 10 times more energy per inference than previous generations, indicating continued infrastructure and energy demand growth.
These signals collectively indicate a substantial increase in datacenter power consumption driven by hyperscaler AI expansion, with regional grid constraints emerging in key markets and cooling infrastructure scaling to meet higher density demands.
Such developments suggest ongoing capital flows into AI infrastructure scaling and digital asset energy demand, with potential implications for grid capacity and energy supply resilience in major data center regions.
The dataset does not specify margin levels for energy supply, nor does it include detailed liquidity breakdowns for infrastructure investments; it also lacks forward guidance beyond the current growth projections.