The Power Bill Behind the AI Boom
While AI companies race to build bigger models, the electric grid underneath them is quietly becoming the industry's biggest constraint — and the bill is landing on regular households.
Quick Highlights
- Global data center power demand is projected to rise 27% in 2026 alone, reaching 132 gigawatts.
- US data center power demand is set to climb from 31 GW in 2025 to 41 GW in 2026, and 66 GW by 2027.
- Utilities requested over $29 billion in rate increases in just the first half of 2025 to fund grid upgrades.
- Some residential customers near major data center hubs have already seen bills rise 9-14% in a single year.
- Lawmakers have introduced legislation aiming to shift grid upgrade costs away from ordinary ratepayers.
For most of the last decade, power was a background line item for the tech industry — a cost to manage, not a constraint to plan around. That changed fast. In 2026, electricity has become the single biggest bottleneck standing between AI companies and their next generation of models, and the buildout required to feed that demand is starting to show up directly on household electric bills across the country.
The Numbers: A Genuinely Unprecedented Surge
Start with the scale of what's actually happening. Worldwide data center electricity demand is projected to rise 27% in 2026 alone, reaching 132 gigawatts, up from 104 gigawatts in 2025, according to Gartner. That's expected to climb further to roughly 290 gigawatts by 2030. AI-optimized servers are driving most of that growth, and Gartner projects their share of total data center power consumption will overtake conventional servers as soon as 2027.
In the US specifically, Goldman Sachs Research projects data center power demand will climb from 31 gigawatts in 2025 to 41 gigawatts in 2026, and then to 66 gigawatts in 2027 — more than doubling in just two years. Data centers' share of US peak summer power demand is expected to rise from about 4% in 2025 to over 8% by 2027.
Why the Grid Can't Just Keep Up
Utilities across the country received requests for at least 700 gigawatts of new data center power connections in 2025 alone — more electricity than the entire United States consumed in all of 2023. Most of those requests will never be built, but even a fraction of them represents a scale of new demand the grid was never designed to absorb quickly.
The core problem isn't generation capacity in the abstract — it's speed and location. Building new power plants, substations, and transmission lines takes years, while hyperscale AI campuses are being proposed and built on much shorter timelines. That mismatch is forcing utilities into expensive, rushed upgrades concentrated in specific regions, rather than spread gradually across the whole system.
Who's Actually Paying for This
This is where the story stops being an abstract infrastructure debate and starts showing up in mailboxes. The Edison Electric Institute estimated its member utilities would invest nearly $208 billion in 2025 just to strengthen and expand the grid, with tens of billions in requested rate increases on top of that. Those costs are largely recovered through customer electricity bills.
Real household impact
One Virginia homeowner featured in Consumer Reports saw his monthly electricity bill jump from around $100 to $281 in a single month, and he isn't alone — residential customers in parts of Ohio and Pennsylvania have seen rates rise 9% and 14% respectively over the past year, regions tied closely to rapid data center buildout.
Manufacturers are feeling it too
It isn't just households. A 141-year-old Ohio brick manufacturer reportedly saw its electricity costs jump 90% in a single year, driven largely by rising grid capacity charges tied to the same buildout.
A Contested Political Fight
Rising bills near data center hubs have become a live political issue, and the numbers being cited publicly don't always tell the same story. Senator Elizabeth Warren claimed residential bills near large data centers had risen as much as 267% over five years — a fact-check by Politifact found that figure actually referred to wholesale grid capacity costs, not the residential bills consumers actually pay, even though real regional residential increases have still been significant.
In response to rising public concern, lawmakers introduced the SHIELD Act in early 2026, aiming to shift the cost of grid upgrades needed for large data centers onto the companies building them rather than everyday ratepayers. Whether legislation like this moves forward, and how state regulators handle rate increase requests, will likely shape how much of this cost households continue to absorb directly.
Frequently Asked Questions
They're a significant contributing factor in regions with heavy data center concentration, though overall electricity price increases also reflect broader grid infrastructure costs, weather-driven demand, and other factors.
Global data center power demand is projected to rise 27% in 2026 alone, and US demand is expected to more than double between 2025 and 2027 according to Goldman Sachs Research.
Renewables currently cover only a portion of data center electricity use, and while their growth is accelerating, current projections suggest they'll meet less than half of the additional demand expected through 2030.
Yes — legislation like the proposed SHIELD Act aims to shift grid upgrade costs from large data center operators onto the companies themselves rather than ordinary ratepayers, though its passage remains uncertain.
No. Impact is highly concentrated in regions with heavy hyperscale data center development, such as parts of Virginia, Ohio, and Pennsylvania, while other areas have seen comparatively smaller effects so far.
Final Thoughts
The AI industry's growth story in 2026 isn't just about bigger models and faster chips anymore — it's increasingly a story about power plants, transmission lines, and who ultimately pays for them. The scale of demand is genuinely unprecedented, the grid wasn't built to absorb it this quickly, and the resulting costs are already landing on households and manufacturers far removed from any AI lab.
How that cost gets distributed between tech companies, utilities, and ratepayers over the next few years may end up shaping the AI industry's public reputation as much as any model release does.
Comments
Post a Comment