Infrastructure Economics & Reliability
Selected industry measures of capital investment, computational demand, and outage consequences. Updated October 16, 2026. Values below are not a measure of AI-caused losses.
U.S. investor-owned utility capital expenditures
Industry-wide investment, billions of U.S. dollars. The 2024 value is reported as actual; 2025 and 2026 values are projections from the cited reporting series, not final audited outcomes.
These amounts cover broad utility infrastructure programs, not only AI or data centers. The three-year sum is not an estimate of economic damage.
What the figures can—and cannot—tell us
Growing computational demand is one of several drivers of electrical infrastructure planning and investment. Industry-wide utility capital spending is not an estimate of AI-related losses, and data-center outage surveys do not isolate incidents caused by AI workloads or grid overload. These measures should not be added together as a total cost of AI infrastructure.
Global data-center electricity consumption
Selected IEA estimates and outlook, in terawatt-hours per year. The 2030 value is a forecast, not an observed result.
Consumption growth indicates planning pressure; it does not by itself establish inadequate capacity or the cause of a particular disturbance.
Cost of significant data-center outages
Uptime Institute's 2024 survey reports that 54% of respondents put the cost of their most recent significant outage above $100,000; approximately 20% reported costs above $1 million. These are overlapping thresholds, not additive categories.
Survey responses are not a census of all incidents. These costs cannot be attributed to AI-driven grid problems without incident-specific evidence.
Sources, interpretation & limitations
These are third-party estimates and survey findings, not DEOB-generated operational measurements. Reporting periods and forecast vintages differ; figures should be rechecked against the latest original publications before external citation.
- Edison Electric Institute — Industry data and capital investment reporting
- International Energy Agency — Energy and AI
- Uptime Institute — Research and annual surveys
Interpretive boundary: Capital spending is not an outage loss. Data-center electricity demand does not identify the cause of grid disturbances. A survey of outage costs does not establish the amount of AI-caused damage.
What this means for technical planning
As infrastructure becomes more interconnected, facility operators may benefit from defined technical examinations, coordination with qualified specialists, review of operating characteristics, and clear documentation of findings. DEOB supports these activities within the agreed scope of each engagement.
Computational Load Readiness → · Technical Services → · Contact DEOB →