Has Google Figured Out The 'Right Way' To Power An AI Data Center?

[00:00:00] Speaker 1: This field will soon become another node in the growing network behind America's AI boom. Google plans to build a data center here in Pine Island, a small town of about 4,000 in southeastern Minnesota. It's slated for a capacity of 780 megawatts. That alone is enough to power more than half a million homes. [00:00:22] Speaker 2: My biggest concern is my electric bill is going to go up. [00:00:27] Speaker 3: I'm concerned that they're going to use all of our electricity. We're not going to have the air conditioning that we need, the heat that we need. [00:00:35] Speaker 1: And all around the country, tech companies are scrambling to secure enough power for their data centers. The grid in the United States is aging and requires major upgrades to handle the increasing demands of the AI boom. [00:00:49] Speaker 4: 70% of our transmission grid in the U.S. is over 25 years old. And it's starting to reach the end of its natural life. And that means we need to invest more capital as a country to update these grids. [00:01:02] Speaker 1: So Google is taking matters into its own hands. [00:01:05] Jagar Shah: The truth of the matter is, is that data centers can be designed the wrong way and they can be designed the right way. And if you design them the right way, they can reduce electricity rates for everybody. [00:01:16] Speaker 1: Google and local utility Xcel Energy say that Pine Island is an example of the right way. [00:01:22] Speaker 6: I think there's a real opportunity there that we're just scratching the surface on. [00:01:25] Speaker 1: The tech giant will pay the costs of connecting and serving its facility. And Xcel projects that the agreement will provide more than a billion dollars in benefits to its existing customers. [00:01:34] Speaker 4: The timing of this industrial revolution that we're really living through on the energy side that's being brought on by demand for AI couldn't be more perfect because it's right at the time when we needed capital to fix aging infrastructure in the United States. [00:01:48] Speaker 1: And if it works, it could become a model for how some of the world's biggest tech companies plug into an aging American power grid without pushing the cost onto everyone else. AI has changed the scale of the data center business. The largest campuses can require power on the scale of small cities, which many local utilities can't provide without making massive changes to their infrastructure. [00:02:25] Speaker 7: What's happened in the last few years is we've gone from data centers that are 25, 50 megawatts to data centers that are now 1,000 megawatts or 5,000 megawatts, or we're even talking about 10,000 megawatts. So all of a sudden, it's a couple orders of magnitude larger, and that's a huge amount of power. [00:02:46] Speaker 1: So some data center developers are turning to off-grid energy sources, also known as behind-the-meter power. They essentially build their own power plants, sometimes hauling in gas turbines from companies like GE Vernova and Caterpillar, which has seen their stocks skyrocket over the last year. [00:03:02] Speaker 8: Off-grid gas plants, short of a coal plant, are basically the most polluting way you could power the computing function of a data center. [00:03:11] Speaker 1: Now that's part of the reason data centers have become so controversial. Elon Musk's SpaceX AI built a gas power plant with dozens of turbines to power his Colossus data center complex in Memphis, which then drew intense scrutiny over air and noise pollution to nearby communities. [00:03:29] Speaker 9: Some nights you can smell the fumes from the turbines, and it's really, really noisy at night. You just have to endure this. [00:03:38] Speaker 1: But beyond harms to human health and the environment, the gas turbines proved to be unreliable. [00:03:44] Jagar Shah: So they brought in basically natural gas turbines on skids and said, we're just going to run them off of these things. They learned very quickly that that wasn't going to work because the data center was blowing out a lot of the natural gas turbines because they're so hard on turbines. That's when we learned that building stuff off grid didn't really work. [00:04:04] Speaker 1: US gas fired capacity to power data centers is nearly doubled in the first half of 2026. But the grid remains the destination because it still gives data centers something that an isolated power plant can't easily match. Backup. [00:04:20] Jagar Shah: The grid is the most complicated engineering marvel of the 20th century, right? [00:04:25] Speaker 1: Jagar Shah is a partner at Multiplier, a firm that helps clean energy firms go public. [00:04:30] Jagar Shah: It is thousands of generators and batteries and other things that back each other up. So if any one thing on the grid fails, then the rest of the grid is there to back it up. [00:04:41] Speaker 1: Google, Meta, Microsoft and Amazon are all spending millions on local infrastructure upgrades to power their data centers. [00:04:48] Speaker 4: Any infrastructure you put in the ground, whether it's a data center, a new power plan or a transmission line, is an asset that's going to be there for decades. [00:05:01] Speaker 1: Pine Island is one of more than 2,600 data center projects planned across the U.S., which could add nearly 400 gigawatts of capacity on top of the more than 1,300 data centers already in operation. [00:05:13] Speaker 4: When you have customer demand like we've seen for a Gemini product, we had 400 million users in May and that grew to a billion in August. That's our fastest growing Google product ever. But what's important is that in the long run, we have a stable energy supply that's reliable for our customers and it's built the right way. [00:05:30] Speaker 1: Google is backing more than twice the power it needs with new clean energy resources, which could make the local grid cleaner and more reliable. The company is financing the addition of 1,400 megawatts of new wind power, 200 megawatts of solar and a 300 megawatt iron air battery capable of discharging for up to 100 hours. Google and Excel also say that they've structured their agreement in a way that protects local rate payers from paying for those upgrades. [00:05:58] Speaker 6: Nobody wants to have a big build out and then have nowhere for that power to go and nobody to pay for that power. [00:06:05] Speaker 1: Minimum demand charges and exit fees are also baked into the contract, which are intended to keep Google on the hook even if the AI boom cools or the facility uses less power than forecasted. And Google, it is promising flexibility. [00:06:18] Speaker 4: So when the infrastructure is built, we commit to pay for it even if our load doesn't show up. [00:06:23] Speaker 8: A shockingly high percentage of the costs of running the electricity grid are for like the four hours of the year on the hottest day in the summer when everybody's running their air conditioning. You're talking about the peaks of the earth. A lot of the engineering that utilities do are to cover those peak hours. [00:06:41] Speaker 1: That's independent watchdog David Pomerantz, who calls Google the high road operator in the tech sector. If a data center can reduce demand during those rare peaks, the utility may be able to avoid building new equipment that would only be used a few hours each year. [00:06:56] Speaker 4: The benefit that we see the most in being able to be flexible is that we can connect to the grid faster. So it allows us to have what we call a non-firm interconnect. If you need me to drop off, I will drop off. And that is a much smarter and more durable solution while they build out the infrastructure. But Excel also told CNBC that customers shouldn't expect a lower bill. [00:07:19] Speaker 6: You're not going to see your bills lower. There's inflation everywhere else that we're all experiencing. But when we bring these large users onto the system and do it right, like we have with Google, it can actually help drive down that inflation, even though bills are going to increase. [00:07:36] Speaker 1: Now, in other words, Excel's argument is that bills may still rise, but by less than they would have without Google's project coming onto the system. On top of those great improvements, Google's promised $100 million in infrastructure upgrades to the area, pledged $24 million for Pine Island public schools and $3 million to the city of Pine Island. But it's still not enough for many locals. [00:08:01] Speaker 10: And I understand they're offering the city any number of incentives. From what I understand, they don't have a very good reputation in other small communities that they've built in. [00:08:12] Speaker 3: I don't think it's worth it. I'm concerned that they're going to use all of our electricity. Where's all that energy coming from? [00:08:19] Speaker 1: Data centers can bring tax revenue and investment, but all of the Pine Island residents at CNBC spoke with still object to this one. [00:08:26] Speaker 11: We're told that our energy bills will not go up. I mean, they're just so vague on the information they give, but yet they accept us to just trust them. It's not going to be good. It's, I mean, how can it be? [00:08:38] Speaker 4: You can't be given trust. It has to be earned. People want transparency and they want to be heard. We're absolutely committed to building the most environmentally responsible data center we can. [00:08:49] Speaker 1: Google's power agreement with Xcel Energy has not yet been approved by the Minnesota Public Utilities Commission, while the transmission lines and other facilities are still going through the permitting process. However, Shah from consulting firm Multiplier tells CNBC that data centers could be taking advantage of excess capacity that already exists on the grid without the need to fund new infrastructure projects at all. [00:09:16] Jagar Shah: If you fit the load growth that the data centers want into the existing grid, everybody's rates go down by a lot. [00:09:23] Speaker 1: Think of the grid as a system with big fixed costs, power lines, substations, transformers and existing power plants. If a data center can tap unused room in that existing system, it brings in a lot of new revenue while adding relatively little new cost. [00:09:39] Jagar Shah: The question really is why are people not doing these 25 megawatt data centers? Why are they doing 1000 megawatt data centers? And it turns out people are just lazy, right? They're like, well, it's easy for me to contract one data center that's one gigawatt in size than figuring out how to do something on the order of 40 data centers that are 25 megawatts each. [00:10:01] Speaker 1: But Pomerant says that utilities aren't necessarily eager to embrace a model that could require them to build less and infringe on their margins. [00:10:08] Speaker 8: The thesis behind how part of how data centers can save money for customers is that they can operate in ways that, at least on a relative basis, requires less stuff to be built. Well, that is a bit threatening to the utilities profit models. Utilities get to profit from the equity share of the capital investments they make in the grid. When they build something new, they fund part of that investment with dollars from their investors on Wall Street and they enjoy incredibly high exorbitant returns on those investments. [00:10:42] Jagar Shah: Energy bills are going up and they have been going up because electric utility companies get paid to invest capital, right? And so the more money they invest, the more money they make for shareholders. [00:10:55] Speaker 1: Pomerantz also says Google and Excel's promises shouldn't be accepted on faith alone. [00:11:00] Speaker 8: I buy that data centers could lower costs for rate payers in theory. There are a lot of things that have to happen for that promise to be made true. If those things don't happen, it's very possible for data centers to raise customers costs. [00:11:15] Speaker 3: I like Pine Island. I like it the way it is. [00:11:18] Speaker 2: You want to bring in this big thing and yeah, maybe it's good for the world. Maybe it's good for America. I don't know. But is it good for me?