Sir Martin Frobisher is the explorer that most of us outside of Canada never get to learn about. In 1577, after sailing past Greenland and getting stuck in the icebound northern reaches of what would become Canada, Frobisher came across a mountain of gold, and excitedly filled his holds to ship the riches back to England. Of course, had Frobisher or the assayers on his ship not slept through Geo 101, they would have known that the glittering hard, brittle crystals embedded in the shale ores weighing down his ships were pyrite - more commonly known as fools gold - a mineral formed from iron and sulfur. While lovely to look at, pyrite is mostly known for its hazardous explosive properties in coal mines and its insidious role in acid mine drainage. Frobisher’s haul was deemed useless, and ended up as the gravel bed to an English road.
Frobisher got off relatively easy. All that was lost were two ore-laden ships, a bit of Queen Elizabeth’s goodwill, and some of Frobisher’s seemingly boundless enthusiasm for misadventure.
The supersizing of data centers and artificial intelligence (AI) has put us in the midst of a modern rush for the hills, and while nobody knows if the hills are filled with proverbial gold or pyrite, the Queen’s wealth is once again going to build an armada in heady anticipation. Unlike Frobisher’s folly, however, the data center gold rush is impinging itself on a common good - the utilities and electric system upon which we all rely and pay for (and increasingly, our climate). The rush to build, and power, data centers offers an unprecedented opportunity for utilities, and an extraordinary risk - and ratepayers are unwittingly in the middle.
For many utilities, the data center boom came out of seemingly nowhere. As of late 2022, most utilities predicted steady, but slow electricity growth driven by new manufacturing, electric vehicles, and building electrification. The pace of forecasted growth had changed radically by 2024. Propelled by shock value of ChatGPT and other AI models, proposals for new data centers started cropping up in nearly every state, driven by both the name brand hyperscalers – like Amazon, Google, Meta, and Microsoft – and newly emergent developers.
The potential scale of data center growth is extraordinary. Observers estimate there could be well over 100 gigawatts (GW) of data centers online by 2030, consuming more than 15 percent of US electricity (and well above 25 percent in some states). Around 8 GW of data centers were built in 2024 alone, and there are an estimated 10 GW coming online in 2025, which will consume enough energy to power 8 million homes. Adding 100 GW of new demand by the end of the decade would be enough energy to power 80 million homes, or more than half US housing units.
While the enormity of the current data center sector is deeply impactful for the utilities that have to react to this system shock, it pales in comparison to what utilities are being told to expect by data center developers moving forward. Based on utility filings and investor presentations, the Sierra Club uncovered two dozen utilities that collectively claim they’ve been approached with over 700 GW of data center development (non-cryptocurrency mining, a whole other ballgame), or enough energy to consume every unit of electricity produced in North America in 2024 - and then more.
For example, Oncor, a Texas-based electricity delivery company, claimed it had been approached by 156 GW of data center interest (equivalent to the entire US coal fleet), while FirstEnergy, an Ohio-based utility, claimed it had been approached by 80 GW of data center developers (just under the capacity of all US nuclear plants). PPL, which operates in Pennsylvania and Kentucky, claimed that it had 56 GW data centers come knocking, and Southern Company, which covers Georgia, Alabama, and Mississippi, claims that it has over 50 GW of prospective large load customers, many of which are data centers. The scale of these requests is staggering - and these just represent the utilities that have been willing to disclose their pipeline.
So what’s actually represented by 700 GW of data center inquiries? It’s actually remarkably difficult to know, because utilities hold their cards close, and data center developers hold them even closer, subjecting both utilities and local officials to non-disclosure agreements. Utilities in potentially fast-growing spaces, like Duke (NC, SC, IN, OH, and FL), TVA (TN), MidAmerican (IA), and APS (AZ) have been cagey about the inquiries they’ve received from data center developers.
Unfortunately, it is remarkably hard to nail down real data center electric demand projections from either utilities or data center developers: even top analysts aren’t clear about the size of the market, or if it will be financially viable, the folks who just build data centers aren’t sure if they’ll have takers over the long run, and utilities really just don’t know what to think.
There are a huge number of data center speculators emerging: Organizations that track data center development show a rocketing number of real estate developers working to get in on the data center action. According to data from 451 Research, there are at least 50 developers in the US looking to build significant (>50 MW) data centers that have little or no operational data centers in their portfolio. Indeed, those speculators account for nearly one-third of the data center capacity (by energy) in the database, and likely far more where they’re untracked.
Utilities tend to disclose only name-brand data center customers: While utilities may claim huge pipelines of data center customers, the fraction that they do discuss tend to be recognizable names. The hyperscalers like Amazon, Google, Meta, and Microsoft feature prominently, as well as wholesalers like Digital Realty, Equinix, Aligned, and QTS. The new speculators don’t tend to make the cut, and for good reason; nobody’s established a need for real estate developers to get in on the game.
Disconcertingly, speculators may be the largest fraction of the pipeline: Some utilities appear to take many of these inquiries with a grain of salt (or in some cases, a lot of salt), significantly downgrading data center customer inquiries in their load forecasts. For example, NV Energy disclosed that it only assumes 15 percent of data center customers without firm contracts will actually get online, while PacifiCorp makes no allowances without a firm contract. Other utilities appear to be more eager to see if they can get in on the action.
So what’s the big deal with a bit of good old speculation?
Well, first, we need to recall that utilities aren’t really just private enterprises, we grant them monopolies to - theoretically - attend to the common good, namely services we rely on like electricity and water. So when a utility succumbs to speculation, it’s betting with our money. And betting they are. Kentucky utility KU/LG&E is on the record pursuing nearly 2 GW of new gas, an investment of $3 billion of new gas power and retrofits to old coal plants to chase after data center customers without signed contracts; Duke Carolinas wants to install 3.6 GW of new gas, largely in the name of new data center customers; and Georgia Power is positioning itself to develop 13 GW of generation to serve new large loads in the next decade.
Second, utilities and policymakers are using the speculative rush for data centers to not only prop up fossil fuels, but roll back climate and environmental protections. President Trump has issued no less than five executive orders citing the rush for data centers as the excuse for rolling back environmental and consumer protections (example). North Carolina lawmakers overrode a veto by Governor Stein, allowing the state to roll back mandated climate targets, and Virginia may follow suit. And the three largest Regional Transmission Operators (PJM, MISO, and SPP) have new plans to fast-track gas projects – at the expense of both time and money for renewable energy projects – to meet burgeoning data center demands.
And finally, utilities are using the economic development pipeline as an excuse to skip good planning principles. Normally, we expect utilities to look out two decades or more to figure out what demand is going to look like, make sure that they can get cost-effective new generation on line, and get rid of the stuff that doesn’t make sense anymore, like old coal plants. And while many utilities are far from perfect, the last decade saw significant advances in planning sophistication, clean energy integration, and increasingly thoughtful approaches towards technologies like battery storage, grid modernization, efficiency, and demand management. But today, we’re seeing utilities issue “emergency” resource plans, fast-tracking new gas, and quietly decommit coal retirements, all in service to bring on data centers.
If we’re not careful, we may all end up paying for this expensive quest. And while Sir Frobisher simply lost his shirt, a boat, and a bit of pride, the rest of us have a lot more to lose if utilities all go chasing after that seductively sparkling pyrite.
Sierra Club has developed a series of no regrets policy measures that large customers, utilities, and regulators can adopt to reduce speculation, and drive demand towards clean, sustainable options. In addition, we’re working to ensure that large load tariffs protect ratepayers and utility systems.