Data center power demand is now one of the biggest forces reshaping the US electricity system. Gartner expects worldwide data center electricity demand to rise by roughly 26 to 27 percent in 2026, from about 105 gigawatts to about 133 gigawatts, with AI-optimized servers accounting for around 31 percent of the total. The AI boom is no longer limited by chips alone. It is increasingly limited by power.
How big is the jump?
- Worldwide data center demand: about 105 GW rising to about 133 GW in 2026, according to Gartner.
- AI’s share: AI-optimized servers are expected to use about 31% of data center power this year.
- US electricity use: analysts expect US consumption to reach record highs in 2026 and 2027 as data centers expand.
- Looking ahead: data centers’ share of US electricity is projected to roughly double by 2030.
Why power has become the bottleneck
Building a data center takes a year or two, but connecting it to the grid and adding new generation can take longer. A late-October report says capital markets are tightening on AI data centers while about 15 GW of projects wait for power. In plain terms, many planned facilities have the chips and the money but not yet the electricity.
Who pays for the grid upgrades?
This is the question regulators and customers care about most. Utilities in several states are planning costly grid upgrades, such as in Missouri, and analysts say regional electricity price pressure is rising. Some large technology companies have begun paying for grid expansion themselves, and states such as North Carolina are considering new rules for how data centers are charged. How these costs are shared between data centers and homes will vary by state and by utility.
What it means for households and businesses
- Bills: more demand can push up costs in some regions, but the effect depends on local rules and who funds the new infrastructure.
- Jobs and investment: data center projects bring construction work and tax income to the places that host them.
- Energy choices: expect more interest in nuclear, gas, solar, batteries and efficiency, because AI companies need reliable power.
- Developers: efficient models, smaller models and caching can lower costs as power gets more expensive.
What to watch next
- State regulators’ decisions on data center electricity rates.
- New power deals between AI companies and utilities or energy producers.
- Financing for AI data centers, since tighter funding can slow new projects.
- Quarterly updates from chipmakers on how much AI demand they expect.
Frequently asked questions
How much electricity do AI data centers use?
Gartner expects AI-optimized servers to use about 31% of data center power in 2026, within a total of about 133 GW worldwide.
Will AI data centers raise my electric bill?
It depends on where you live. Rising demand can put pressure on prices, but some companies and regulators are shifting more of the grid cost to the data centers themselves.
Why is power the limit for AI?
New data centers need a large, stable supply of electricity, and building grid connections and generation takes time.
Sources
- Network World: Gartner on data center electricity growth in 2026
- IO+: Funding for data centers is drying up as electricity demand skyrockets
- OilPrice.com: AI boom sends US electricity demand to a new high
- E&E News: Data centers’ share of US electricity seen doubling by 2030
Related reading: Nvidia’s earnings and the AI boom.
Published 12 November 2026. Figures come from the cited reports and may be revised.
