Data center moratoriums are sweeping the nation—and IT pros need to be prepared

In a staunchly divided country, Americans are pretty aligned on one thing: opposition to data centers. IT pros, though, are in a trickier spot because data center moratoriums could spell trouble for orgs that have woven AI tools into various workflows, if the compute power to keep them running doesn’t get built.

Thus far, 26 states have passed or are considering data center moratoriums and guardrails. Most recently, Pennsylvania Governor Josh Shapiro signed an executive order to ensure anything built there complies with specific environmental and affordability requirements.

“If you can’t agree to our strict requirements and get the community where you want to build to say ’yes,’ you’re not going to have the Commonwealth’s support either,” Shapiro said in a press release. “These are some of the biggest companies in the world—they can afford to be good neighbors, follow the rules, and do this right.”

As similar—and stricter—data center bans and limits take effect in states and counties across the country, the legislation could start to affect IT departments’ workflows and budgets. Chiefly, IT professionals will face uncertainty as the availability of compute power shifts. But that doesn’t mean there aren’t ways to get ahead of it.

The scope of it all. The effect moratoriums will have on IT pros will depend on the scope of the legislation, which Constellation Research VP and Principal Analyst Holger Mueller said will be out of step with AI development, because “technology has always evolved faster than legislative speeds.” Even so, Mueller suggested, anywhere a data center moratorium isn’t, a data center will be.

“If the data centers are not going to be built in states A, B, C, they will be in states X, Y, Z. Or they might be in Canada, or they might be in Mexico,” Mueller said. “The data centers will get built.”

Should there be any interval of time in which there aren’t enough data centers to accommodate the compute power needed, though, IT departments could experience something akin to an AI brownout or even blackout. Due to insufficient GPU capacity, AI-supported IT work will be slower and IT pros will need to prioritize which workflows they want to keep automating and which they can do by hand.

“Cloud vendors who offer AI will not be able to make good on their commitments to provide people’s GPU capacity,” Mueller said. “That means if people banked for those GPU capacities, they will not be able to get the automation that they want.”

Cybersecurity would also be a major concern. Mueller said if IT departments don’t have enough GPUs to maintain an adequate cyberdefense, their data could be vulnerable to hackers with stronger AI—and shutting down operations for safety purposes might be an option.

The price (might not be) right. As is the case in any market where demand outpaces supply, prices will go up. Mueller said he predicts that the (relative) scarcity of data centers will drive up the price of AI compute—and workarounds for data center scarcity aren’t cheap, either. Wannie Park, the founder and CEO of data center energy orchestration company Pado AI, told Morning Brew that in solving for inadequate vertical assets, “IT budgets are gonna get hit hard.”

One possible solution, Park said, is using neoclouds like CoreWeave if a data center isn’t available. On one hand, neoclouds are a “good short-term stopgap” that can buy IT departments time to secure longer-term infrastructure. On the other hand, they’re not very cost effective.

“You’re probably not going to get the best price, but you’ll get the supply,” Park said.

There are ways to save on AI compute should prices skyrocket, though. Park said he recommends IT departments not make any changes to their vertical asset(s) until the legislation and its effects stabilize. It’s also worth investing in workload optimization software to ensure that GPUs are being used efficiently.

“We’re not in a market or a situation right now where the demand is dropping off. It’s the inverse. The supply is dropping off, and the demand continues to outpace the growth of supply,” Park said. “When you can synthetically—through software and services—increase your productivity 40%, that’s pretty good.”

Generally speaking, Park said, IT departments can approach the uncertainty around data center and GPU availability similar to sustainability and disaster recovery protocols: Make sure hardware is working efficiently, and diversify “cloud topology” in case legislation affects assets.

“It’s going back to first principles of just managing a better P&L,” Park said. “No one’s just going to write you a blank check to open up a new data center.”