How data centers and natural disasters are pushing the grid to its limits

Let's Talk Energy: Data centers, extreme weather and the grid Host: Noah Brenner Guests: Arun Nimmala, global head of grid operational technology products and services, Oracle (Atlanta); Tom Eyford, global industry specialist, utility operations solutions, Oracle (Phoenix) Transcript lightly edited for readability. Timestamps appear every four minutes. [00:00] Tom Eyford: The number of events is going up, the severity of these events is going up, our reliance on the grid is going up. So the consequences are going up. And as Arun points out, the grid is actually operating closer to the margin than it ever has. That's where the real challenge is: we're asking more of the grid precisely at the time it's being stressed the most. Noah Brenner: This is Let's Talk Energy, your go-to podcast for smart energy insights. I'm Noah Brenner. Keeping the lights on during natural disasters has never been more important, as life-saving services and life in general require more power and computing capacity than ever before. And that task is getting more difficult. From 2018 to 2024, the number of major outage events in the US each year, meaning power cuts affecting 30,000 customers or 15% of the customers in a specific area, increased by 40% to more than 6,500, according to a March study from Oak Ridge National Laboratory. The average duration of those outages rose 23%, to just under 12 hours. Now grids are being asked to add thousands of new data centers globally. Data center power demand grew 50% just from 2023 to 2025, and it's expected to almost double, to just under 250 gigawatts globally by 2030, according to Rystad Energy forecasts, putting unprecedented strain on the power system. So can grids cope with the simultaneous stresses of natural disasters and rapid data center buildout? How should big tech, utilities and regulators manage data center demand, as well as supply from on-site generation, during these types of weather events? And finally, what's the outlook for improving the ways grids function during weather-related stress, perhaps even leveraging the same AI that's adding to the stress in the first place? To help us understand the intersection of soaring data center demand and natural disasters, I'm joined today from Atlanta by Arun Nimmala, global head of grid operational technology products and services for Oracle. Arun, welcome to the program. Arun Nimmala: Thank you, Noah. It's always great to speak on these topics. Noah Brenner: And from Phoenix we've got Tom Eyford, a global industry specialist in utility operations solutions with Oracle. Tom, welcome. Tom Eyford: Thanks for having me. I'm really excited about this conversation. It's something I've been very passionate about, so I'm looking forward to the next hour or so. Noah Brenner: Excellent. Well, gentlemen, let's talk energy. Before we get started, a lot of people are probably familiar with Oracle as a large technology company. It has its own data centers as well. But they may not be as familiar with Oracle's work with utilities and grid managers. Could you each give us a quick grounding in what you do and how you got to where you are today? Arun Nimmala: Oracle has a zillion things going on right now, but we also provide software within industry verticals, and Oracle Utilities is one of them. We provide software across multiple layers, whether that's the customer side, the asset side or the operations side. Where my role comes in is that I lead our Oracle ADMS and DMS products globally. That's the technology that helps our utility partners and clients see, operate and optimize the distribution grid in real time. That's where my focus is. We're helping large utilities make the shift from a mostly one-way grid to one that is more dynamic, one that can withstand weather events and the electrification we're seeing with so much energy demand. My background: 15 years with the utilities industry, and 34 years in the technology industry. The utilities industry has been a very challenging experience, and given what we're seeing right now, this is one place I would really recommend [04:00] most youngsters should be, to understand what's going on. It's the perfect intersection of technology and energy need. Tom Eyford: Global industry specialist is just a fancy way of saying I've been a utility guy my whole career. I've reached the point where I can say 30-plus years in the industry and I'm going to cap it there from now on. I started out in distribution engineering and operations on the electric side, then worked across a bunch of different areas within the utility: technology, asset management, strategic planning, customer service. I sat in a lot of chairs before coming to Oracle in 2011. The really cool part now is that I get to take all that experience and work with utilities around the world to understand how they're thinking about technology and how we can use that technology to solve the problems they're seeing. Noah Brenner: It's been a tough year for electrical grids globally. Here in the US we've seen blizzards in New York, tornadoes in the Midwest, heat waves in the South, wildfires across the West. We heard some outage figures at the top of the show, but if we needed a reminder, as we tape this right now more than 70,000 people are without power around the Houston area from Tropical Storm Edward. From where you sit in the industry, how does it look? Are the stresses getting worse from extreme weather, or do we just hear a lot more about them? Arun Nimmala: We are seeing these weather events more frequently. If you look at NOAA data, it shows billions of dollars of investment and damage happening because of weather disasters, and they have nearly doubled in frequency over the last decade. So the short answer is that both statements are true. The weather itself is getting more volatile, and the grid has become more critical to daily life. Any given event now carries a higher stake, and it gets more and more visibility. If you look at our distribution grids, they were largely designed decades ago, when climate conditions were very different. Today's temperature swings, wind patterns, even something as small as changing population density across cities, have increasingly exceeded what we initially assumed for those grid designs. The grid is now operating closer to its limits in most places. Seventy percent of US transmission lines are over 25 years old, which tells you what the grid was designed for versus what we're experiencing right now. Coupled with that, look at society's reliance on continuous power. With electrification coming of age, everything from kids' toys to EVs needs electricity, and with IoT in the picture that need has only increased, right down to heat pumps and water heaters. Energy demand is growing, so these events are more common. And because of that dependency, the media glare is coming to these weather events too. Tom Eyford: The number of events is going up, the severity is going up, our reliance on the grid is going up, so the consequences are going up. And as Arun points out, the grid is operating closer to the margin than it ever has. That's the real challenge: we're asking more of the grid precisely at the time it's being stressed the most. [08:00] The good news is that we're compensating for some of this with technology. We have significantly better visibility and forecasting than we ever did, much faster controls, much better automation, and increasingly we're using autonomous systems to identify conditions and respond before a human operator ever could. But we need all of this because of those factors coming into play. We have all this technology, but frankly we have much slimmer margins to deal with. That's why we're still seeing issues, still seeing outages across the United States and across the world hitting the news. Without all this technology it would be a whole lot worse. Noah Brenner: What happens in a utility control room, or at an ISO, during these types of events? For those who aren't grid wonks, an ISO is an independent system operator, the body that makes sure the grid works as it should. What's going on in a control room right now in Houston as people try to get power back online? Tom Eyford: The more preparation we can do ahead of time, the better we're going to be. So what they're spending their time on beforehand is: what type of storm is it, where is it going to hit us, how much damage do we think it's going to do, what type of damage, and what material, equipment and resources are we going to need to restore it? Am I going to need to bring in neighboring crews, or maybe not neighboring crews, because they may be hit with the exact same thing we are? So I need to look to utilities outside my area and say, "Hey, can you help me?" Those are the questions they're already starting to ask. They start staging material, staging people, just to get ready. Then when the storm actually hits, a lot of the time all we can do is wait. There's no point going out to restore something immediately when it's just going to get knocked down 20 minutes later. So in a lot of cases you let the storm blow through, and then as soon as it's safe, because we don't want to put people in harm's way when we don't have to, you get crews out in the field to start assessing. The next questions are: what just happened, where did we get hit, how bad is it? Then we triage. First, get the backbone back: if we had a transmission issue, get that back on first, then substations, then the mainline feeders, then critical facilities, hospitals, and even things like schools, because we can use gymnasiums as public resource centers if we need to. Then we can start working on the smaller problems. The other piece we can't forget is making sure the public stays informed the whole way. In a lot of ways it doesn't matter how good your performance is if the public doesn't think you're doing a good job. Noah Brenner: How is the use of AI changing the processes you've just mentioned? How are utilities able to leverage AI to predict, manage and, if necessary, actually fix? Arun Nimmala: AI can substantially help in these situations. It can help utilities become more prepared and make decisions faster. One thing I'd add to what Tom was saying is that modern storms, and any event, play out in multiple phases, and there are places where we cannot react, where we cannot put humans in an unsafe place. That's where the software comes in. During the event we don't want to put crews in the middle of the storm, but the software can still do its job. [12:00] For example, the Oracle ADMS solution gives us telemetry from SCADA and from smart meters. That information becomes an input into how control room operators, or the field crews, react to storm events. Coming back to AI, it can make utility operators more anticipatory. I'll divide this into two phases: prediction and management. On prediction, the models have improved so much that when you combine them with weather forecasts, vegetation data, asset age and historical patterns, we can generate a score that helps us understand the risk of each asset. We can understand the risk to an individual pole or asset, not just to a region. We've seen deep learning frameworks deliver up to a 35% reduction in load shedding during extreme weather events. On management, AI-assisted restoration estimates drawn from historical repair patterns improve accuracy, which means we can inform customers properly. As Tom said, communication is key. We can manage customer expectations and we can also manage the load on the crew. In practical terms, it's a decision support mechanism for those crews: making them aware of the data so they can make decisions the right way. Getting the right information to the right person faster. That's where we provide AI as a solution to control room operators. Noah Brenner: How do the operations you've discussed change when you have this proliferation of large loads, particularly from data centers? I know there are others out there as the US re-industrializes. With a mix of grid-connected and behind-the-meter power, how does that dynamic affect how you deal with these situations? Tom Eyford: I'd separate it out a little. From a traditional storm restoration perspective, a large data center is probably not the thing keeping distribution operators up at night, because typically they're so big that they aren't served by the distribution system at all. Generally speaking they'll have their own facilities served directly on the transmission system. In a lot of ways they're a separate problem. They'll even have managers on the utility side working with them directly, letting them know what's happening upfront. It's very much a personal conversation at that point. Where it really becomes an issue for system operations is simply the fact that they're tremendously big loads. A single large data center can represent hundreds of megawatts or more, and increasingly we're talking about campuses of them. That showing up, and going away, can have major impacts on the grid. We've spent more than a century thinking about what happens when we lose a large generator. Well, a data center represents that size of impact to the grid as well. So it's an event that can create its own stability issues, just from the fact that it goes on or off. What we have to think about is not only the scale but also the rate of change. If it suddenly drops off, that's a problem. You'd think, hey, that's just load going away, we should be fine. No, because we have to match generation and load, so whatever we just lost we have to immediately remove from the generation mix. [16:00] The more warning we have, the more easily we can ramp that down, the better. That's why those types of facilities have at least one backup, and why we make sure we're working with data centers so they don't lose large chunks of load immediately. Arun Nimmala: Most utilities and grid operators look at data centers as a different kind of load. They are different, not just from the energy consumption point of view, but because they also require finer-grained visibility. That's why we always say data centers, and any energy demand coming from data centers, should be looked at as grid participants, not just as passive load. Noah Brenner: Is this an ISO or a utility saying, in the interest of everyone's stability, we'd like you to act in this way in response to this event, this weather, wildfire or hurricane? Tom Eyford: I'd say that's the grid operator's job, typically the ISO, because data centers are connected into the transmission system. Their job is to communicate the constraint, and to do it as early as possible, to give them enough warning that they can start taking action more circumspectly. The more warning we give them, the more options they have. Data centers have surprisingly sophisticated ways of manipulating the load they actually require. They can defer workloads that aren't time-sensitive, prioritize certain customer service classes, shift workloads geographically to other data centers. They have all sorts of ways to prioritize the tasks they're doing, which can have a really large effect on the load they're drawing at any given time. So rather than the ISO saying "you must do this," they say, "look, I need 200 megawatts, and I need it by here," and the data centers can start to ramp down. Again, we don't want it all to go away in a snap. We want to say, can you do that over a stage of an hour or two, so we can start adjusting the generation mix appropriately, if we're going to take a particular unit out of service, for example. The more coordination those two entities can have, the better. Let the ISO handle the constraint side, and let the data centers handle how they meet it. They can bring on other generation, maybe they have battery storage. They have all sorts of options on the data center side. Noah Brenner: Could these types of facilities, the ones that have their own power, help support the stability of the grid as well? Is there bidirectionality there? Tom Eyford: It absolutely is. We hear people talk about behind the meter, but it doesn't necessarily mean disconnected from the grid. In fact, most of the distributed generation being added today is interconnected with the grid. It just happens to sit on the customer's side of the meter. That wasn't always the case. I think people still have in their head the concept of a plug-in portable generator that can't be connected to the grid, that has to be isolated and have protection around it. That's really not the case for anything new, and certainly not for anything permanently installed. The biggest value we can get out of those assets, both from the grid perspective and from the owner's perspective, is to have them interconnected full-time. That unlocks the value for the grid, because I can leverage them to enhance grid stability. [20:00] When there are issues I can potentially call on them, and the customer then has the ability to provide that service, sell their excess energy and offer it as needed to the grid. That's value for both sides. That's the key: we want those connected to the grid. The standards communities have been working for close to 20 years on how to do this safely and appropriately, and how to handle all of the issues. Arun Nimmala: If you look at energy utilities and customers, they were always looked at in silos. Utilities provided a solution to the customer. This is the first time we're seeing customers become part of the solution, with behind-the-meter assets coming into the picture, because utilities are asking customers whether their devices can feed back into the grid, and there's an incentive going back to the customer. So now customers are not just the problem statement, they are part of the solution. That's what behind-the-meter assets are providing. They are solving the problem. They are not just consuming energy, they are giving energy back at the right time. That balance is helping utilities manage energy needs across critical assets, weather events, and whatever demand is coming from data centers. This is an exciting time, with terms like DERMS coming into the picture, where the customer becomes part of the solution and not just the receiving end of the spectrum. Tom Eyford: The caveat here is that just because these resources can help the grid doesn't mean they're going to. They may all respond in the exact same way, and that itself can be an issue. We need to orchestrate them centrally, to say, don't overrespond, don't respond too quickly. We've seen similar things in financial markets, with auto traders where suddenly everybody sells because they all see the same price signal, and things collapse. The same thing can happen on the grid. We can overcorrect if everybody drops off and then everybody comes back, and you get these oscillating events. So we still need to make sure everyone is coordinated and orchestrated. But the resources are there and we can use them. Noah Brenner: How much of what we're talking about is reality today, and how much is more of a story over the next two to three years, as we see data center buildout adding load but also the capabilities AI can bring? Arun Nimmala: I would say it is a reality today. Maybe not every piece of it, but a lot of the building blocks are already there. Today we have tools like switching, voltage optimization and FLISR, all of which are already part of the ADMS application. That's the primary building block. We also have AMI-based outage detection tools, large-scale commercial demand response programs, and DERMS deployments for large load curtailment contracts. These are already being implemented by utilities today. That's why I say it's part of the story now. Is everything ready? No. Obviously there's a learning curve, and the technology is evolving. Building for the next two to three years, I'd point to standardizing large-load interconnection tariffs, which is part of the regulation, and tying that directly into real-time distribution operations rather than just running demand response programs. Then there's deeper grid edge visibility, down to the low-voltage network, and AI providing proper recommendations, which closes the loop of automated control. [24:00] That part is future-looking. Utilities are always preparing with a phased approach. We ourselves are implementing DERMS solutions for multiple utilities right now, and they are preparing for that kind of future evolution. Real-time visibility, automated responses, predictive maintenance and prescriptive AI solutions are being implemented currently. Tom Eyford: To build off Arun, I totally agree. The building blocks are absolutely there. From a technology perspective I don't want to minimize the complexity, but I think we understand those problems, and utilities, technology providers and research laboratories are all working on them. The problem we have yet to solve is more the people side and the regulatory side: how do we make sure we're compensating customers appropriately, making sure the benefits are equitable, making sure customers are going to want to allow utilities to use these resources for their benefit. There's still a trust issue there. We have to work together and build that cooperation, and I think that's going to be more of a challenge than the technical one. Noah Brenner: What about weather that isn't quite as extreme but is no less taxing on the grid? You've got 10 days of temperatures above 100 degrees, which in Phoenix is probably a cold day, or a cold snap in West Texas. Could you lay out some of the less appreciated impacts of these types of events, and how you deal with that slow burn? Tom Eyford: There's certainly more going on than just more or less load. During a heat wave the clear bet is more air conditioning, during a cold snap you see more electric heating. But there are a lot of second-order effects that make these more challenging than simply seeing higher demand. Heat is probably the bigger of the two problems. Air conditioning is one of the fastest growing residential sources of load, so we're progressively putting more load on the grid at exactly the times it can handle it the least. One of the limiting factors on the grid is excess heat, thermal limits. The closer the ambient temperature gets to that limit, the less headroom there is, and that actually lowers the capacity of the grid. I can't heat the equipment as high because I'm already starting out at a higher temperature. So right when we want the grid to operate the most, during those really hot periods, it has the least capacity available. Cold isn't exactly a panacea either. It's wonderful that we have much more capacity on the grid during cold temperatures, but equipment doesn't like cold. Anyone who's tried to start their car in 20 below understands that this stuff doesn't just work the way we think. Fuel gets viscous, equipment can freeze. We saw that in Texas several years back, where a lot of equipment was frozen and we couldn't use it. So it becomes a problem on both ends. The important thing to remember is that extreme temperature doesn't just change how much electricity we use. It changes the condition of almost everything we rely on to produce and deliver that electricity. [28:00] Arun Nimmala: And it isn't only during the event. Post-event, as Tom said, asset maintenance becomes a bigger challenge, because now you have to get those assets back to working condition. In these extreme events, peak demand is just the visible effect. Extreme temperatures change equipment performance and resource availability, and the operators have to absorb all that contingency and prepare for the next heat wave or the next cold snap. Noah Brenner: What's different, or similar, about wildfires, in terms of preparing and preventing as well as dealing with them? Tom Eyford: The response to them can be pretty similar, but there's a huge difference these days, and it's really about liability. This is pretty much the one major event that utilities could potentially be held responsible for, and that changes everything in terms of how they prepare, how they operate and how they interact with the public. It's almost unique within the electric industry, in that we operate thousands of miles of energized equipment outdoors, through forests and communities, exposed to wind and vegetation. It's almost waiting to happen, and we can't practically engineer that risk to zero. What society used to accept as a community risk, because wildfires have been happening, and even utility-caused wildfires have been happening, since we've had utilities, has changed. Society has moved from "yes, it's something that happens" to "utility, you should have prevented this, you should have fixed this." We've now seen several billion-dollar lawsuits, to the point where this is potentially an existential risk to the utility. We've even seen bankruptcies as a result. Inside the utility we don't want this to happen. We want to prevent it and mitigate it as much as possible. But you'll see utilities starting to operate extremely conservatively. As soon as there's a credible risk of wildfire, we see preemptive shutoffs and very conservative behavior during the event. This is not fun for people. People don't like having their power shut off, they don't like being off supply for hours, and then nothing happens and they ask, "well, why was I shut off?" Because the risk is really asymmetric. All these shutoffs upset people, we get commission complaints, we see reliability impacts, all the stuff utilities have been trying to avoid for a hundred years. But the flip side is somebody dying, or billions of dollars of damage. Those are very asymmetric risk profiles, so utilities are being very conservative in how they operate through these events. Noah Brenner: These issues have been rapidly evolving and changing. If we were to sit down five years from now and have a similar discussion, how do you think the conversation might be different, whether that's grid development, new loads coming on, or new weather patterns? Tom Eyford: I'm going to be cautious about predicting five years out, because if you'd asked me five years ago what we'd be talking about today, I think I'd have been totally wrong. In fact I might have answered this question a little differently even a few months ago. [32:00] Certainly the reality was that data centers were inevitable and they were going to be everywhere. Well, just in the last couple of months we've started to see meaningful pushback from communities, to the point where it's front and center in the midterm elections here. As an aside, how cool is it that energy and utilities is actually an issue in the midterms? We're used to being invisible, so it's always exciting to see the national discourse take up these issues. But the landscape is changing. I still think data centers and AI are here to stay. It's just a question of how we get there from here. We may not see the explosive growth, or maybe there's more, who knows how this changes. Five years ago it was all about connecting DERs and how we decarbonize the grid, shifting from traditional fuel sources to renewables. Now we're saying we can't get rid of the old ones because we can't build renewables fast enough. Resource adequacy is a thing, and on the transmission system it's probably the most important issue operators are dealing with at this point. So you're seeing plants that were planning to shut down stay open, new gas-fired plants coming online, and nuclear going through something of a renaissance. In five years, I don't see this AI slowing down, so it's probably going to be more of the same: how do we build the right resources? We're going to be talking a lot about what "the right resources" means, what that mix is, and what we as a society want our resource mix to be. Certainly as battery technology improves, energy storage is going to expand a lot, and that makes wind, solar and other intermittent renewables a whole lot more viable, which can change the game as well. I'm not going to put a pin in it and say this is where we're going, but those are the issues we're going to see. I'm really excited to see what the next five years brings. Arun Nimmala: The problem is not just for the United States. Yes, there is a lot of pushback on data centers in the US, but we are shifting the problem to different countries, because these data centers are being built globally. Data centers are a global problem, so energy becomes a global problem. I strongly think that in the next few years we will not be talking about whether the grid can handle this, or whether energy demand is going to grow, or how we are going to supply it. We will be talking more about how autonomously the grid can handle this, and what tools we are going to give the grid, whether that's AI, AI-orchestrated systems or other grid edge devices. How are we going to make the grid more sustainable to help with this demand? I see it dividing into three layers. First, look 10 years back, to what happened in Japan, where they completely shut down a nuclear power plant, and our nuclear plants were shut down too because of safety. From that extreme, we now have so many companies coming up with small modular reactors. We never heard the term SMR in the past. Demand is driving the need for technology right now. That's where I see grid operations becoming more and more autonomous, with closed-loop AI making real-time switching and optimization decisions. We're going to see more and more of that. [36:00] And don't discount the use of bots in the control room to execute these things. DERMS is becoming more and more apparent, and it is going to increase a lot, with more utilities adopting it, because distributed resources are going to increase. Right now we're just talking about EVs and water heaters. I'm pretty sure almost every electric device will be a distributed device going forward. The last piece is how we manage weather events and wildfires. That's a wild card, and that's where some technology improvement needs to happen. Generating the energy is one thing, but sustaining it, minimizing the risk of losing energy we've already generated, is something we need to focus on. So it's not about whether the grid can handle it or not. It's about what tools we can provide so the grid can autonomously manage itself and reduce energy losses. Noah Brenner: Gentlemen, thank you for a fantastic and timely discussion. Tom, thanks for joining us. Tom Eyford: Absolutely. Thank you. Appreciate it. Noah Brenner: And Arun, thank you as well for joining. Arun Nimmala: Thank you, Noah. It's great to talk to you. Noah Brenner: Thanks for listening to Let's Talk Energy. This podcast is a production of Rystad Energy. Check out the show notes for further analysis on the topics we've discussed in the episode, and find us on social media. We're Rystad Energy on all your major platforms.