Power grid operators and datacenter developers in the United States are in a bind, and energy analysts can't see an easy way out. 

The American power grid is old, outdated, and in desperate need of upgrades. Add in a growing number of gigawatt-scale AI datacenters demanding stable access to power that doesn’t disrupt service-level uptime agreements and things start to look even worse. 

Energy costs have already skyrocketed in the nation's largest energy market mainly thanks to the bit barn bonanza, leading to a new chorus of calls for datacenters to bring their own power generation if they want reliable supplies not bound by grid constraints. 

According to energy analysts at Wood Mackenzie, datacenter operators have a choice to make, and neither option is great. 

They can wait the five to 10 years it’ll take for grid operators to upgrade their transmission and generation capabilities to account for their demands. 

Or they can accept deals with power companies that supply them with power but require curtailment during peak loads, and then install their own on-site power generation to make up the difference. This is the far riskier option, but it’s one that many operators are going with. 

“With more than 90 GW of collocated generation in US interconnection pipelines, it is clear that the need to scale up at speed has sent many data centre developers down the riskier path,” WoodMac explained in the report. “Collocating volatile AI workloads with power generation has scarce precedent, though, and is far more difficult than most in the industry understand.” 

What the grid wants, the grid gets

Many datacenter operators are only thinking about generator megawatts, say the analysts, leaving engineers frustrated at having to explain the technical complexities of building colocated power generators. Multiple grid operators have passed rules that, as key stakeholders understand them, could give utilities priority rights over colocated power generators during shortages, potentially forcing datacenters to reduce demand while supplying power back to the grid, WoodMac said.

“This effectively makes the model unworkable,” the analysts said. “Few data centre developers would invest in baseload generation if it could not be utilised when it was most needed.”

Near-instantaneous swings in AI power demand can damage reciprocating engines and gas turbines, while batteries may not respond fast enough to every spike and can degrade over time. The rapidly fluctuating power demanded by hyperscale AI datacenters could even damage the grid itself.

“These loads can also cause sub-synchronous oscillations, which pose fundamental stability risk to not only local generators but also to distant ones on the transmission system,” the analysts note. “Technology providers are only beginning to come to terms with this challenge, the mitigation of which is site specific, making solutions hard to scale.” 

In other words, on-site generation might solve one problem while introducing a whole host of new issues. 

“While hyperscalers are likely to successfully operationalise some projects with collocated resources, it will come at considerable cost,” the report concludes. “This cost, along with the technical risk and project-specific mitigation required, will prevent collocation from emerging as a scalable model.”

So, what’s the alternative? Well, building out the grid, obviously. WoodMac said that grid operators largely don’t consider conditional interconnection with curtailment requirements to be a long-term solution, and are all investing billions in modernization. But those modernization initiatives are going to hit everyone in the wallet. 

Improving grids to serve AI datacenters “has profound implications for affordability … regional network upgrades necessary to support local large-load connections will be spread among all ratepayers,” WoodMac said. “This may trigger a political outcry.” 

Rates are already increasing across the US, said Ben Hertz-Shargel, WoodMac's global head of grid transformation and large loads, and no presidential decree can slow that down.

"As the data center buildout eclipses existing utility capacity and more infrastructure must be built, we'll see that increasingly become a driver of customer rates," Hertz-Shargel told us in an email. 

The study doesn’t present a solution for the current situation, forecasting uncertainty as grid operators hurry to come up with some solution, as datacenter operators’ plans blow past their ability to meet demand. Hertz-Shargel didn't have much comfort to offer here, either. 

"With utilities and grid operators reforming their load interconnection processes, we're likely to see more capacity made available through utilities," Hertz-Shargel said. Not everyone is going to be able to secure that utility capacity, though, and the organizations that are left out will be stuck with the same choice: Wait for the grid to catch up, or rely on colocated generation and conditional grid connection deals, Hertz-Shargel explained. 

"When the current wave of transmission capacity completes, though, the industry will be in a position for quick acceleration," Hertz-Shargel added – as long as the numbers continue to work out. "That presumes that investors like what they see in terms of AI's return on investment by that time, however, and remain comfortable deploying massive capital into digital infrastructure."

That's far from a sure thing at this point. 

The only thing that the report sees for certain in the current situation is a quick division into winners and losers in the AI race. Big players who are able to absorb the costs will emerge as victors. 

“The challenges facing [energy] collocation are surmountable for the most experienced and deep-pocketed developers,” Wood Mackenzie said. “Companies capable of operating reliably without firm grid service will be able to scale their AI business faster than others, positioning them to outcompete.”

In other words, expect the big guns to further entrench their dominance as the little guys are starved to death or are gobbled up by the competition. There’s nothing like the American dream, eh? ®