The 2025 Data Center World conference in Washington DC attracted close to 5,000 people. And a lot wanted to know how they could bring more power to their data centers.
They came to the right place. Developers of “AI factories” laid out how they find power to run massive data centers being set up to run AI applications.
Mr. Wonderful Kevin O’Leary from ABC’s “Shark Tank” is building the world’s largest data center in Alberta, Canada. “Data centers are today’s gold rush,” said O’Leary during his keynote. “The capital cost of an AI data center is so high that you have to build big.”
His power plant project, Wonder Valley in the Municipal District of Greenview, is now under construction. Phase one, providing 1.5 GW, is to be completed by the end of 2027 for $2 billion. Subsequent phases will ramp up capacity to almost 8 GW.
Why Alberta? It’s all about stranded natural gas in large quantities. Instead of bringing power to the data center, O’Leary has brought the data center to a source of cheap power.
He’s buying 10 gas turbines for the site.“Everything comes from the availability of power in abundance,” said O’Leary. “Stranded natural gas is inexpensive and clean.”
Grid delays spur BYOP
There are over 12,000 active projects in the U.S. seeking grid interconnection representing 1,570 GW of generator capacity and 1,030 GW of storage.
With data center power consumption expected to grow from 2.5% of U.S. electricity to 7.5% over the next five years due to AI, the shortage of power for data centers could amount to over 45 GW, according to Boston Consulting Group.
No wonder 62% of data centers are exploring on-site power generation. 19% have started it. AI factory developers want power now and many are committed to Bring Your Own Power (BYOP).
Some find success by locating stranded wind assets. Data center developer Applied Digital found a large wind farm in North Dakota looking for customers. Over 400 MW of power will be fed to the company’s AI factory under construction.
“The AI race favors builders who can energize fast –computer demand isn’t going to slow down and the biggest bottleneck is power,” said Laura Laltrello, COO at Applied Digital. Laltrello. “Major metros are tapped out, grid constrained, and slow to scale.”
Nuclear energy from small modular reactors (SMRs) is another popular option. Google has partnered with Kairos Power to develop SMRs that will run Google data centers.
“We aim to have an initial SMR deployment by 2030, with further expansion of the site through 2035,” said Lucia Tian, head of clean energy & decarbonization at Google. “Nuclear energy will be more cost effective and dependable than variable wind and solar.”
SMRs, however, are not a route to immediate power. It might take five years to develop, permit and open an SMR – perhaps longer. Natural gas offers the promise of fast power at the level of availability an AI data center demands.
“Between 2024 and 2030, natural gas consumption by data centers will rise by almost three times,” said Dave Bell, Vice President of Data Center and Microgrid Development at VoltaGrid. “By then, data centers will be using up 4.5% of U.S. gas consumption for electricity generation.”
He recommended that data centers find locations close to existing natural gas basins and near intrastate and interstate pipelines. This includes West Virginia, Pennsylvania, Ohio, Virgina, Texas, and Oklahoma.
Gas reserves are one part of BYOP.; obtaining gas turbines is the other vital element. Some OEMs are saying it might be three or more years to deliver a gas turbine ordered today. That applies especially to large advanced-class H and J-technology turbines.
But there are plenty of other small and mid-sized turbines available that also can do the job. Further, there are companies around who offer trailer-mounted gas turbine packages that can be onsite, within a few months of order in many cases.
