Data centers are now measured in megawatts, not CPUs
Daniel YerginDaniel Yergin — Oil destroyed Hitler, fracking destroyed Putinat 54:00
From the conversation
Now we talk about them as 300 megawatt data centers. The sense is that you have electric cars and energy transition demand. Then you're bringing back chip manufacturers and smart manufacturing to the US. That's electricity demand. Then you have AI and data centers. Suddenly this industry that had been very flat is now looking at growth. How you are going to meet the growth is very much on the agenda right now. Data centers are looking at where they can position themselves so that we have access to the electricity that they need: reliable 24-hour electricity. Now there's energy security in terms of oil and gas. Actually it's also energy security in terms of electricity. There's your potential constraint on economic activity. Some will say the answer to that is innovation. Chips will become less electricity dependent or data centers will operate differently. So the demand will not grow as much. There are those who say that will happen, but it hasn't happened yet. Others are saying, “How are we going to meet that demand?” AI is going to demand a lot more electricity than we had thought about a year or a year and a half ago. It's potentially even worse than the 10% number implies, because it's not widely distributed like households would be. In many cases, they have to be one gigawatt to one specific campus or location. Right. You look at developing data centers.…
Summary
The scale of data centers is now measured in megawatts of power consumption rather than CPU counts, reflecting a fundamental shift in how AI infrastructure is understood. Converging demands—electric vehicles, reshored chip manufacturing, and AI data centers—are pushing an electricity industry that had been stagnant into a new growth phase. Data centers are actively choosing locations based on access to reliable, around-the-clock electricity supply.
Watch the clip on YouTubeStarts at 54:00