Technology · Analysis
AI's Power Hunger Rewrites Energy Rules
Data centers are driving the first sustained rise in U.S. electricity demand in a decade, forcing utilities to rethink infrastructure, regulators to fast-track pipelines, and battery makers to chase a market that didn't exist three years ago.
Stake & Paper Editorial TeamJuly 30, 2026
American Electric Power raised its full-year earnings guidance Wednesday morning after reporting second-quarter results that missed analyst expectations. The miss didn't matter. What mattered was the number buried in the release: 69 gigawatts of new load additions through 2030, up from previous projections, according to the company's earnings statement. Nearly all of it is data centers running artificial intelligence workloads.
AEP lifted its 2026 operating earnings guidance to $6.25 to $6.55 per share from a prior range of $6.15 to $6.45 per share, and reaffirmed its annual operating earnings growth rate of 7% to 9% through 2030
, the company reported. The revision reflects what utilities across the country are confronting:
U.S. power demand is projected to rise from a record 4,195 billion kilowatt-hours in 2025 to 4,269 billion kWh in 2026 and 4,399 billion kWh in 2027, according to the Energy Information Administration
.
For the first time on record, commercial sector electricity demand is expected to outpace residential demand in 2026
, the EIA said.
The shift is structural, not cyclical. After more than a decade of flat consumption, electricity use is climbing again -- and the driver is compute, not households.
Can the Grid Keep Pace With AI's Appetite?
The infrastructure required to meet this demand is staggering.
Construction on NextEra and Brookfield's $100 billion data center campus in Kentucky is expected to be completed in 2031
, Bloomberg reported.
The project will include up to 2 GW of natural gas-fired power and 2.6 GW of battery storage capacity
, according to the companies.
The campus will be built on the Department of Energy's Paducah Site, which was constructed in the 1950s to produce enriched uranium for nuclear weapons and later for reactors before closing in 2013
, Reuters reported.
The site already has transmission capacity, water, fiber, and other infrastructure in place -- a rare advantage in an industry where lead times for new power projects can stretch beyond a decade.
NextEra said the project fulfills the Trump administration's "Ratepayer Protection Pledge," which seeks to ensure that companies building and using data centers do not pass energy costs onto residential customers
, according to Reuters.
Not every project is moving as smoothly. Energy Transfer's Green Chile pipeline, designed to supply natural gas to Oracle and OpenAI's Project Jupiter data center in New Mexico, has hit regulatory delays.
The 17-mile pipeline extension was expected to deliver up to 400 million cubic feet of natural gas per day to power 2.5 GW of Bloom fuel cells deployed on site
, Data Center Dynamics reported. But federal regulators moved the project out of fast-track review, Natural Gas Intel reported, potentially delaying a decision until December.
State officials cited large-scale water use, increased CO2 emissions, and a lack of state revenue generation as major factors in their decision
, according to Data Center Dynamics.
Where Does the Power Come From?
Nuclear is emerging as a preferred option for hyperscale operators seeking carbon-free baseload power.
Rolls-Royce CEO Tufan Erginbilgic said the company has been in discussions with major data center operators about power systems and small modular reactors, and told CNBC that "early next week, in [the] investment committee, we are going to sign another big hyperscaler sort of deal"
.
In the first half of 2026, Rolls-Royce's defense unit delivered organic revenue gains of 17% and a 57% jump in underlying profits, while its power systems unit, which includes engines for data centers, grew organic revenue 28% and profits 72%
, Yahoo Finance reported.
Battery storage is also seeing a surge in interest.
InoBat, a European battery energy storage systems manufacturer, announced a business combination with Cartesian Growth Corporation II that values the company at $1.265 billion on a pre-money basis, with $77.5 million in committed financing
, according to a company statement.
InoBat claims to have delivered or contracted 875 MWh of utility-scale battery energy storage capacity to date and is now targeting the data center power market
, Energy-Storage.News reported.
Meanwhile, Chinese battery cell manufacturer Highstar unveiled a full-chain battery solution for AI data centers at an industry summit in late July, according to a company press release. The portfolio spans UPS systems, battery backup units, and grid-side energy storage -- all designed to handle the high rack density and load volatility that AI workloads create.
What Changed This Week
American Electric Power's guidance revision signals that utilities are no longer treating data center demand as speculative. The 69 GW pipeline AEP disclosed represents firm commitments, not forecasts. Rolls-Royce's imminent hyperscaler deal, if finalized, would mark another step toward nuclear power becoming a standard option for AI infrastructure. And the regulatory pushback on Energy Transfer's Green Chile pipeline shows that not all projects will sail through -- even when backed by OpenAI and Oracle.
What to Watch
American Electric Power reports second-quarter earnings on July 30 before market open. Watch for details on how the company plans to finance its expanded capital program and whether regulators in its service territories are approving cost recovery for data center-related infrastructure. The Federal Energy Regulatory Commission's decision on Energy Transfer's Green Chile pipeline, now expected in December, will test whether federal agencies prioritize AI infrastructure development over state-level environmental concerns. And Rolls-Royce's next hyperscaler announcement, expected in early August, could clarify whether small modular reactors are moving from concept to commercial deployment faster than the market expects.