Will The Power Industry Meet The Moment? At Dtech Data Ce…

The question landed early on day two of DTECH Data Centers and AIin Scottsdale, and it set the tone for nearly everything that ed.

“Our next generation will look at us and ask whether we were the industry and society that handled it right, or the industry and society that didn’t,” said Arshad Mansoor, president and CEO of the Electric Power Research Institute (EPRI), during the conference opening keynote panel.

Mansoor framed the data center buildout as the kind of structural shift that comes once a century, drawing historic parallels to the public opposition that electricity, railroads, and cars each faced before becoming foundational to modern life.

The framing carried through three days of sessions last week that drew various utility engineers and planners, hyperscaler representatives, OEMs, EPCs, and policy voices to the Valley of the Sun.

Demand and Ye Shall Receive?

EPRI’s most recent forecast, published earlier this year, projected data center demand reaching as high as 95 gigawatts (GW) in the United States by 2030 under its high case. Mansoor said the high case is now tracking as the base case.

“Energy is at the foundational level of everything we do,” said Marc Spieler, senior managing director at NVIDIA, the chipmaker whose graphics processing units sit at the heart of nearly every AI training and inference workload reshaping a load curve. “The ability to power is an issue.”

At the same time, you’re ly well aware of the mood shifting outside the conference walls. Public opposition to data center development has hardened in communities across the country, and the acronym BANANA, for “build absolutely nothing anywhere near anyone,” surfaced more than once during the event as shorthand for the sentiment that utilities and developers are increasingly encountering.

Mary Sprayregen, global head of regulatory affairs and market development at Oracle Utilities, argued that the industry’s most significant challenge is public trust. She said utilities have a narrow window to demonstrate that adding new load can actually benefit ratepayers rather than burden them, citing Alliant Energy’s announcement of two data center developments in Cedar Rapids, Iowa, that the company has paired with a five-year residential rate freeze.

“We need to find more pathways to demonstrate that data center load can benefit everyday people,” she said.

Jake Tetlow, chief operating officer at Arizona Public Service (APS), agreed on the destination but identified a different obstacle. For APS, the binding constraint is physical infrastructure. The utility has already committed to nearly 5 GW of incremental growth, spanning a Taiwan Semiconductor Manufacturing Company expansion, residential growth, and data centers.

Tetlow said reliability has to come first, citing modeling showing that a sustained outage on a 110-degree day in Phoenix could result in thousands of casualties. He also emphasized that residential customers should not subsidize large load growth, and pointed to a proposed natural gas pipeline from Texas to Arizona as one of the resources APS is counting on.

Inside the Operational Reset at Salt River Project

If the keynote framed the stakes, Scott Scharli, director of strategic energy management at Salt River Project (SRP), showed what happens inside a utility trying to absorb the curve. SRP serves the Phoenix metropolitan area with a summer peak of 8.5 GW. Large business customers currently account for about 1.2 GW, or roughly 12 percent of the system. SRP projects that will climb to between 39 and 40 percent.

The pivot point came in 2023. SRP had historically handled one or two large load requests a year on an ad hoc basis. By 2023, the queue had ballooned to more than 20 active projects, engineers were routinely working 12-hour days, and the existing intake process was breaking down. SRP went back to customers in the queue, some of whom had investors and press releases out, and told them the process had to pause.

“The system is broken, and the results of the information that we gave you are not accurate anymore,” Scharli recalled, describing what SRP told customers at the time. “Sometimes the magic solution there is just the truth.”

The utility pulled staff from transmission, generation, customer service, and strategy into a six-month working group and established a governance structure that reported up to the CEO. SRP then rebuilt its large-load process around a multi-stage intake with significant financial gates. Customers pay a $40,000 nonrefundable application fee, then $250,000 for an engineering study, then post an irrevocable letter of credit covering 30 percent of project value, and ultimately up to 60 percent before construction. Transmission upgrades are not rate-based. Large-load customers pay full freight, a practice SRP has maintained for more than 40 years.

The most recent cluster study results, delivered to customers in October, are a stark measure of where speculation meets reality. SRP studied 24 projects and presented customers with roughly $13 billion in combined infrastructure and upgrade costs. Only a handful of customers stayed in and posted the 30 percent letter of credit to continue. Scharli estimated the original 24 were roughly 80 percent seasoned operators and 20 percent speculators or land developers. Cost per megawatt (MW) and time-to-power were the most common reasons for dropping out, with some customers receiving energization timelines in the 2034 range.

Scharli also flagged what he called the stickiest unresolved issue: load ramp. Data center sales models depend on guaranteed power, while utility planning works on forecasts and long generation lead times. “We don’t have a solution right now, but staying at the table is probably the best thing you can do,” he said.

Bring Your Friends

A separate session featuring Phil Jones, director of AI data centers and microgrids at Emerson, and Minesh Patel, vice president of commercial operations at Prevalon, the Mitsubishi Power energy storage joint venture, captured the supply-side scramble.

Jones used the phrase “bring your friends” to describe how OEMs, controls vendors, battery integrators, and EPCs now have to operate. The traditional engineer-procure-construct sequence has been scrambled, he said, because developers are buying generation equipment before designs are finalized.

“If I can get a spec that’s not on a bar napkin, I’m excited,” he joked.

The technical case for tighter collaboration is becoming clearer as project scale increases. Jones said the industry is now seeing 100 MW load swings in less than a second, multiple times a second, from AI workloads, and no rotating asset can absorb that mechanically without storage smoothing the curve.

Patel, drawing on Prevalon’s power-side experience, said the company would not have allowed anyone to direct-connect turbines to data centers at that load profile and expect it to work. The Emerson-Prevalon collaboration pairs Emerson’s plant controls with Prevalon’s battery energy storage to manage the transient response. Patel said Prevalon’s work in this space began roughly , when it was approached by a hyperscaler about supporting AI transient loads, though Emerson and Prevalon have a longer-standing relationship through ties to Mitsubishi Heavy Industries.

Jones also flagged a contractual disconnect that has yet to fully surface. Data center developers are signing five-nines reliability contracts, which allow roughly five minutes of downtime per year, and then asking power providers to deliver against that standard on assets designed to operate for 25 to 30 years.

“How are you putting your N plus one configuration?” Jones asked. “These are not two-year assets.”

Patel described running two to three coordination calls a day with peers to solve problems on these projects, a cadence he said became routine over six to nine months of collaboration. Asked whether the industry will eventually standardize, he said he had seen 15 to 20 of these opportunities in the past nine months, and no two were a.

“Nobody is bringing the exact same assets. Nobody is delivering the exact same compute requirements,” he pointed out.

The takeaway, he offered, is that the era of any one company solving these problems alone is over.

Grid-Enhancing Technologies as a Parallel Track

Another panel, moderated by Julia Selker of the WATT Coalition and featuring Portland General Electric and Georgia Power, highlighted what utilities can deploy on the existing grid while new transmission catches up. The U.S. Department of Energy (DOE) under the Biden Administration previously estimated that grid-enhancing technologies (GETs) and advanced conductors could unlock more than 100 GW of capacity on existing infrastructure.

Lee Recchia, grid operations senior manager at PGE, said the Oregon utility is operating 106 dynamic line rating sensors across 13 transmission lines and has 30 high-performance conductor projects underway. PGE has also committed its virtual power plant to provide 25 percent of its system load by 2030.

Kaleb Holbrook, manager of strategic projects at Georgia Power, said the utility has deployed more than 200 miles of advanced conductor with several hundred more planned over the next decade, and is evaluating digital dynamic line rating systems that could cover ten times the mileage of physical sensor deployments in the same timeframe.

Selker, who is WATT Coalition’s executive director, said the load-growth scenario has shifted executive-level thinking about capital spending. Utilities are facing “more [capex] than they can manage, potentially more than they can pass on to ratepayers,” she said.