Powering Intelligence: Energy’s Critical Role in AI’s Rise

Disclaimer

This podcast is for general information purposes only and is not advice. Outside of the United States, this episode is intended for investment professional use only, not for further distribution. Please listen to the end for complete information.

Vineet Khanna
You know, while it's great we pay AI researchers hundreds of millions of dollars, we should probably allocate some of that to the trade schools so that we have enough electricians.

Jackie Fortner
Welcome back to “The Angle” from T. Rowe Price, a podcast for curious investors. I'm your host, Jackie Fortner, a portfolio specialist here at T. Rowe Price Associates. This season, we've been exploring the rapidly evolving global energy landscape. And we can't do that without addressing the explosive growth in power demand that's already underway, thanks in overwhelming part to the massive buildout of AI.

The capital we've seen being plowed into the AI infrastructure is astounding. But, as we'll discuss, computing power is only one aspect of the global AI arms race. Ultimately, the AI revolution is going to run on electrons.

So what could the U.S. push for AI leadership mean for the power industry and, for all of us listening today, what could it mean for our electricity bills?

To dig deeper into all of these topics and more, I'm joined today by our T. Rowe Price utilities analyst, Vineet Khanna. Welcome to “The Angle,” Vineet.

Vineet Khanna
Thanks, Jackie. Great to be here.

Jackie Fortner
Glad to have you here. We have a lot of ground to cover, so let's just go ahead and get started.

Big changes are under way in the utility sector. Power demand has started to accelerate after growing at a modest pace for much of the last two decades. And, you know, a formerly sleepy sector is capturing the imagination of growth investors.

So does the prospect of an AI-driven surge in electricity consumption maybe mark the start of a golden era for U.S. utilities?

Vineet Khanna
Yeah. You know, the magnitude of demand, demand growth here is just immense, especially after 15 to 20 years of no or low demand growth. However, I'm just not sure that the market fully appreciates sort of the challenges that utilities face in meeting this upsurge in electricity demand while growing earnings and keeping customer bills affordable. Some will be up to the task, but, you know, others might struggle.

Jackie Fortner
So it sounds like there could be winners and losers in this.

Vineet Khanna
Yeah, yeah. And, you know, this will depend on a slew of factors, but, you know, it kind of presents the opportunity for us to bet on the winners while avoiding the losers. And, you know, in the utility sector, return dispersion has been quite narrow over the past 10 to 20 years.

Jackie Fortner
So the last few decades were perhaps, you know, quote unquote easier for utilities. Is the talk about a golden era of utilities going forward warranted then—or maybe at least for the, you know, kind of winners, quote unquote, depending on how they handle the transition?

Vineet Khanna
Yeah. You know, it's a great question. And maybe before we dive into the future, if we take a step back and kind of think about the last 20 or so years—over that time period, utilities roughly quadrupled their net earnings and, meanwhile, customer bills remained affordable.

And so, you know, most, if not all, of the utilities in that 20-year period benefited from three tailwinds that probably accounted for 80% of that dynamic where bills were flat, but earnings went up. And those three dynamics are sort of falling interest rates, falling tax rates, and falling natural gas prices. And you know, these trends, along with two decades of limited demand growth, permitted utilities to invest in rate base while minimizing bill inflation.

Affordability, amongst other factors, can really help foster good relationships with customers, regulators, and politicians—and good relationships with these key stakeholders help drive rate base growth that directly drives earnings per share and dividend growth.

Jackie Fortner
So you mentioned the three tailwinds that have helped give utilities that nice backdrop over the last decade or two. But those have started to reverse or at least are showing signs of petering out. You know, the extended era of near-zero interest rates has come to an end. The U.S. fiscal situation probably constrains the potential for additional corporate tax cuts—I guess we’ll see. And then natural gas prices certainly have moved higher over the last couple of years, post the Russian invasion of Ukraine, and they have become more volatile.

Vineet Khanna
Yeah. No, that's exactly right. All three of those tailwinds are now headwinds. And then, you know, in addition to that, we've certainly seen a rise in sort of the frequency and magnitude of extreme weather events. And then on top of all that, you know, we're beginning to see a tightening in supply-demand of power fundamentals. These five headwinds, unmitigated, will certainly put upward pressure on customer bills.

But mitigating these pressures will certainly require collaboration and trust between utilities, customers, regulators, and investors. And the utilities that outperform are the ones that’ll be able to leverage the unprecedented level of AI-driven demand growth that we're seeing and translating that into improved affordability for existing customers. This will not come naturally to all management teams and regulators, given it is a counter to much of what they've experienced over the past 20-plus years.

Jackie Fortner
So I've seen estimates from the Department of Energy suggesting that data center energy use could grow from, you know, somewhere around 4.4% of U.S. electricity consumption to, you know, somewhere between 6.7 and up, you know, even 12% by 2028. So how big of a game changer is AI for the power industry?

Vineet Khanna
Yeah, these, these numbers are just astonishing. You know, even at the lower end of that range that the DoE gave, this is an immense amount of growth in electricity consumption. We probably haven't seen levels of consumption growth like this since the ‘50s, ‘60s, and ‘70s. And I would also add that just the nature of this growth is something we've never seen before. The likes of Microsoft, Google, Meta, and Amazon have announced data center campuses that are 1 to 3 gigawatts in size. This is an unprecedented concentration of electricity demand, showing up at a pace that we've never seen.

Jackie Fortner
So for those of us that are, you know, casually saying, oh, just add a gig here at a gig there, it sounds like we're missing how extremely difficult that is.

Vineet Khanna
Yeah. You know, throwing around gigs is tough. And, you know, just for context, 1 gigawatt is the generating capacity of a nuclear power plant. And 3 gigawatts is roughly half of the greater Baltimore area’s summer peak. You know, so the AI data centers have much higher power densities than the cloud data centers we're used to. And they also have the similar reliability requirements to ensure that they can run 24-7, 365 days of the year.

Just based on company and news reports, the large-load [project] pipelines amongst the U.S. utilities are roughly sitting at 70 gigawatts. And that's through 2030. A good portion—roughly three-quarters of that—is data centers.

It took roughly 10 to 15 years for data centers to get up to that 4.4% of U.S. consumption that the DoE report referenced. And the DoE is now forecasting a 2X in that over the next 4 to 5 years. This sort of growth is certainly not typical of the utility industry, at least for the past few decades.

And, you know, lastly, I’d just mentioned that data centers run 24-7, and so, the increase in the average levels of demand, but also the peak demand is really important. And that's why, you know, they require just a lot more generation and delivery infrastructure than other sources of electricity.

Jackie Fortner
So reliability of service has always been paramount for electric utilities. Nothing, you know, sours a utility's relationship with its customers or regulators like a string of unplanned outages. So, what are the biggest kind of grid-related constraints to meeting the AI-driven upsurge in power demand while still ensuring we have that consistency of service?

Vineet Khanna
Yeah, I mean, you know, rule No. 1 in the utility management playbook is always keep the lights on. And doing so reliably and safely is, is paramount. You know, nobody wants the bad press and the political blowback that comes from a major outage. And, you know, layering on higher bills is just a recipe for disaster.

Data centers, cloud or AI, contribute to a utility system peak, which typically means the system has to have excess capacity or add capacity to reliably meet the incremental data center demand. Just Grid 101, you know: Supply and demand must match to prevent disruptions. The entire grid—generation, transmission, and distribution—needs to be sized to meet the peak demand, even if under-utilized for much of the year.

This is further exacerbated by, you know, large changes occurring in the generating fleet. That really refers to the retirement of coal that we're seeing and sort of the increasing role of solar, wind, and storage on the grid. And grid planning and operations are just fundamentally different when we have those changes in the generating fleet.

So, you know, the challenge is really just understanding how supply and demand characteristics are changing over time and ensuring that the utilities can maintain reliability. You know, somewhat ironically, AI could be used to ensure grid stability and reliability persist in the future. But, you know, obviously, we need to power the AI to do so.

Jackie Fortner
So maybe, you know, back up a second, pull it up a level, because you touched on this earlier—about how the U.S. hasn't experienced annual load growth of this magnitude since the, you know, ‘50s, ‘60s, ‘70s.  We haven't built like this since the ‘80s. As we look across the backdrop today, there's also less skilled labor to do the work, and there are a lot more rules and regulations that could potentially get in the way. So, will it maybe take a combination of private and public sector to pull this off? Is that really even an option going forward?

Vineet Khanna
Yeah. You know, like you said, it's just a daunting task. We as a country haven't really built like this in a long time. And it's not really just about the grid. We're building a whole host of things, you know—data centers, semiconductor fabs, liquefied natural gas export facilities. And then on top of that, we have an aging workforce, in particular in the skilled trades.

And so, you know, I like to half-jokingly say that you know, while it's great [that] we pay AI researchers hundreds of millions of dollars, we should probably allocate some of that to the trade schools so that we have enough electricians and other skilled trades to make sure we can get all of these things done.

But, you know, in some ways the AI revolution may sort of be a forcing mechanism that drives the changes needed to get us there.

Investing more in trade schools and ensuring that we're doing the right thing to allow folks to realize that it's a great sort of long-term career is really important.

And then, similarly, permitting reform, which seems to be bipartisan for the most part, is critical to really getting a lot of these projects over the finish line in sort of a timely manner.

And then, you know, finally, some certainty from the political and legislative side of things just to ensure that utilities feel comfortable making the necessary large-scale investments here.

Just as an example, you know, nuclear power has been getting a lot of attention. But I know the next energy podcast is going to cover the nuclear renaissance with Rick de los Reyes, our head of commodities, so I won’t steal his thunder.

Jackie Fortner
Actually, in our last episode, Maria Drew and I talked about how both clean energy sources and traditional fossil fuels are likely to have a role together, going forward, in power generation. So it's no longer this kind of either/or dichotomy that maybe we had thought it could become years ago. So, in our conversation, while the topic of affordability came up as a priority, I'm going to—you're the lucky one—I'm going to be more explicit and ask you a pretty direct question. Who's going to end up paying for this massive grid buildout?

Vineet Khanna
For better or for worse, the answer is that it depends. And, you know, this is a great question and, frankly, one I spend a lot of time thinking about. But the response is nuanced, right? The AI revolution could usher in sort of a paradigm shift in how electricity rates are set and how system costs are allocated across different customer class.

Like in the past, large industrial customers—think, factories, manufacturers (the power users of the day, really)—would receive a sweetheart deal on their electricity rates while more of the system costs would be allocated to residential and other commercial customers.

And really, the thinking here for this trade-off was that the lower rates would attract the factories, which would then stimulate the local economy, create jobs, create folks opening new businesses.

But today, you know, we sort of have to flip the script there. Today's power users in the AI space—the mega-cap companies leveraging this technology and building out a lot of this infrastructure—are really pressed to procure the electricity they need to support their growth plans.

From this perspective, the opportunity cost is multiples of any power costs. And so the reality is that the rate needed to charge to an AI data center is still minimal relative to all the capital spending and operating expenses involved in building and running these data centers. An extra cent per kilowatt-hour would go a long way in covering a lot of these upgrades to the broader grid and generating capacity.

The nuance comes in because each power market has different features and is regulated by a state regulator.

In instances where a utility is vertically integrated—meaning they own the transmission, distribution, and generation—odds are that the grid buildout will be appropriately allocated to the AI data centers.

In deregulated markets, where the utilities only own transmission and distribution, and not the generation, it could be a little bit more of a gray area.

And, you know, outside of data centers, the trend toward increasing electrification is likely to continue because we're getting to a point where the cheapest way to do certain things ends up being with the electron.

And, you know, the need for utilities to invest, in a way, was sort of inevitable. And AI data centers are accelerating that process and are likely to provide sort of foundational revenue to support some of the spending.

Unfortunately, I know I haven't probably fairly answered your question—but, you know, this is the best I can do for our listeners out there. So, just say it depends.

Jackie Fortner
So earlier in the conversation you did highlight the potential for, you know, a wider dispersion of outcomes in the utility sector and the power sector, depending on how companies individually handle this transition. So let's talk about your day-to-day job. How do you, from your seat as a utilities analyst, assess which companies can benefit from AI-driven growth while keeping their customers’ bills at or below inflation? It seems like that would be important.

Vineet Khanna
Yeah. You know, and a lot of it is just delving into the numbers.

The utility sector is heavily regulated at both the state and federal level, and that provides us with access to just an immense amount of information with regards to past investments that have been made, future investment plans, as well as the rates that customers are being charged, their monthly bills, and their usage.

And so what we try and do is consider the capex that they've said they need to make for AI and reliability and other, you know, purposes, and we’ll look at what it can mean for affordability if the allocation of these costs among different customer classes doesn't change.

And I think of it in two buckets.

·       The first bucket: Some utilities have excess generation, transmission and distribution capacity. In that case, these embedded costs are divided across more megawatt-hours of demand. In other words, customers will pay a lower share of that fixed cost.

·       The second bucket: To the extent that AI-related demand requires a step change in investment, including new generation and as well as transmission and distribution infrastructure, you know, we want to understand the mechanisms that utilities have to ensure that these costs are allocated appropriately to the data centers. This just varies meaningfully state by state.

And given the size of the capex we're talking about, I just also want to be mindful of stranded asset risk. In other words, if the data center demand doesn't appear, what structures do the utilities have in place to ensure that existing customers don't end up footing the bill and that the data center customers are sort of held liable?

Jackie Fortner
So we've covered a lot of ground in this conversation. But I'm going to ask you one last question. Just with the market, you know, so focused on AI and the innovations coming from that technology field, you know, what emerging technologies do you see in your coverage space? You know, could they help? Is there something coming up that could help power the AI revolution and maybe alleviate some of the grid constraints that we've talked about?

Vineet Khanna
Yeah. You know, innovation on the nuclear front, such as small modular reactors, has certainly garnered a lot of attention. But, you know, I'll let Rick de los Reyes and Maria cover that on the episode later this season.

From my perspective, there are probably a handful of technologies to just be mindful of—and they fall into a couple of categories.

The first one is just better utilizing the grid that we have today.

You know, as I previously mentioned, the grid is not fully utilized most of the time, right? The grid is built to accommodate the summer and winter peaks. And so we can use AI to better optimize the grid to add AI data centers. And we're just really seeing a lot of innovation here in terms of where to site data centers relative to the transmission and where there may be congestion or, you know, we'll say trapped electrons. So that's definitely an area to watch.

And then, kind of more on the physical infrastructure side, there's also some dynamics to be mindful of, you know, including sensors, cameras, drones, and dynamic line rating technology that really all they do, sort of in a nutshell, is allow more power to flow through the existing infrastructure as we have it. And so improving the utilization just goes a long way.

And then, separate from that, another area that we spend a lot of time thinking about is just energy storage. And the way to think about energy storage paired, with, you know, any form of generation, is that now, for the first time ever, we can sort of store energy in sort of a controlled manner.

We've already seen in Texas and California that lithium-ion battery storage, which can provide 2 to 4 hours of backup, can help to stabilize the grid in times of peak demand and weaker output from intermittent renewables. This technology could be just deployed a lot more all over the country. So that's certainly one to follow and, frankly, I think we will see a lot more of that in the coming years.

And then, taking a step back and just thinking further out, long-duration energy storage—so, really something in excess of 48 hours—could really just be a gamechanger.

And so, you know, where did I come up with this 48 hours?

Really, it's when you look at the demand peaks that we're building the grid for, they typically aggregate into about 500 hours of any given year but are sliced up into sort of smaller pieces of 24 to 72 hours at a time.

Jackie Fortner
And we've seen, you know, maybe even some high-profile examples of where you'd like to be having that backup storage capacity in recent years—thinking, you know, winter storms, heat waves, times when the utility or the electricity goes offline. We've seen some of those prolonged power outages and maybe this could help with those events.

Vineet Khanna
Yeah. No, that's exactly right.

And you know, a really good example is winter storm Uri, which led to power outages in Texas. And we've also seen it during summer heat waves. And the long-duration energy storage could be a huge way of addressing this.

And frankly, you know, in both the case of winter storm Uri [and] some of the outages that we've seen in the summers, one of the issues was getting gas supply to the gas plants, right? So, while gas plays an important role, it clearly is not always enough.

And so having kind of something like long-duration energy storage alongside some of these other generating assets would go a long way in allowing us to not necessarily overbuild the grid as much as we have in the past.

Jackie Fortner

Well Vineet, I think that's actually a perfect note to end on. I really appreciate your time and your insights, walking us through how the rise of AI is expanding the range of outcomes in the utility sector in ways that maybe are both challenging but also present kind of a rare opportunity to drive innovation in a sector that maybe hasn't always been associated with innovation.

So meeting the upsurge in electricity demand is definitely going to be a daunting task, as we've talked about. But, you know, I think you've laid out some reasons for optimism for us, and I thank you again for guiding us, you know, through how the AI revolution is really changing the game for utilities and the power industry as a whole.

Vineet Khanna
Yeah. And thanks for the invite and the great conversation—and hope we can do it again soon.

Jackie Fortner

Yeah, definitely hope so.

And to our listeners, thank you again for listening to “The Angle.” We look forward to your company on future episodes. You can find out more information about this and other topics on our website. Please rate and subscribe wherever you get your podcasts. “The Angle” —better questions, better insights only from T. Rowe Price.

Disclaimer
This podcast episode was recorded in November of 2025 and is for general information and educational purposes only. Outside the United States, it is for investment professional use only. It is not intended to be used by persons in jurisdictions which prohibit or restrict distribution of the material herein. The podcast does not give advice or recommendations of any nature or constitute an offer or solicitation to buy or sell any security in any jurisdiction. Prospective investors should seek independent legal, financial, and tax advice before making any investment decision. Past performance is not a guarantee or a reliable indicator of future results. All investments are subject to risk, including the possible loss of principal. Discussions relating to specific securities are informational only, and are not recommendations, and may or may not have been held in any T. Rowe Price portfolio. There should be no assumption that the securities were or will be profitable. T. Rowe Price is not affiliated with any companies discussed. The views contained herein are of the speakers as of the date of the recording and are subject to change without notice. These views may differ from those of other T. Rowe Price associates and/or affiliates. Information is from sources deemed reliable but not guaranteed.

Capex is capital expenditures, or investments in production plants or other physical infrastructure.

Rate base is the net value of a utility’s plant, property, and equipment on which it can earn a return.

DoE is the U.S. Department of Energy

Publications referenced in the podcast include:

“2024 United States Data Center Energy Usage Report,” Lawrence Berkeley National Laboratory, 2024.

Please visit http://www.troweprice.com/theanglepodcast for full global issuer disclosures. This podcast is copyright by T. Rowe Price, 2026.