By David Wojick
False fears of a flood of data centers is warping the electric power policy debate. Here are three little numbers that tell the story. Well, actually, it is two little real numbers and one huge fantasy number (which is the problem).
My first source is “North America Data Center Trends H2 2025” by the real estate tracker CBRE.
This quote provides the two little numbers, which are MW of data centers under construction:
“The total amount of new capacity under construction in primary markets declined for the first time since 2020. There were 5,994.4 MW under construction at the end of 2025, down from 6,350.1 MW in 2024. Many planned projects remain delayed due to ongoing permitting, zoning, and power procurement hurdles, underscoring the complexities of scaling infrastructure.”
So, there is only about 6,000 MW of new data centers in the construction pipeline for the entire U.S. That is a tiny number when it comes to the national grid. PJM alone peaks around 150,000 MW, Texas over 80,000, and so on. Combined American peak is over 750,000 MW.
Moreover, installed data center capacity is just around 17,000 MW. While adding 6,000 MW will be a big percentage increase, it will still be an extremely small amount. Note that this is not 6,000 MW a year, because it takes several years to build a data center.
For the huge fantasy number, we go to “Data Centers in the United States” by Clearview.
They have a nifty map showing all the existing and proposed data center clusters.
Here is the big number quote:
“As of April 2026, the U.S. has 602 operating data centers with 16,914 MW of capacity and 889 planned projects that would add 278,842 MW of additional capacity, according to Cleanview’s project tracker.”
Adding 280,000 MW of new power demand would certainly fry the grid, but this is just a fantasy number. These are not “planned projects,” just proposed or merely mentioned by some bigwig.
Unfortunately, it is variations of this fantasy number that are driving the data center panic. Many press articles read as though we are already being swamped with data centers, when the reality is new construction is barely noticeable. CBRE has a nice listing of the latest year-on-year additions for each region. Almost all are tiny.
Note, too, that the tiny amount of demand growth in existing data centers cannot possibly account for the regional surges in electric power prices. This is why there is no correlation between data center growth and power price increases.
It is helpful to see the long-standing parallel fantasy on the power generation side. The queue of applications for wind and solar projects to connect to the grid is over two million MW, when our total generating capacity is just over one million. Almost none of this proposed generating capacity is going to be built, and neither is the fantasy list of proposed data centers.
The data center threat is way overblown, if it exists at all. This supposed threat is just an unfortunate offshoot from the massive AI hype. Untold billions of investor dollars are flowing into the hot AI bubble, so there is tremendous incentive to develop AI data center proposals.
It is possible that over the next 10 to 20 years there will be significant growth in data center capacity. Some states, regions, and localities are already vying for this development. But it will be constrained by available power — in fact, it already is. New dispatchable generation would be good for restoring reliability, and if the data centers want to help pay for that, fine by me.
On the other hand, if the AI bubble bursts there might be relatively little growth. Investment capital can disappear very quickly.
But in no case is there a data center threat to the grid. Data centers are too small and growing too slowly to be a threat.
