Pinal County officials seek more guidance on large-scale data centers ahead of La Osa decision

Pinal County officials will soon decide whether to approve the county’s first large-scale data center, known as La Osa. Before doing so, the Board of Supervisors held a special session Wednesday to gather more expert opinions.

The University of Arizona’s chief AI and data science officer, David Ebert, gave the first presentation. He explained that data centers have been around since the computer and that Google opened its first hyperscale data center in the early 2000s.

He says for years, data center operators made enough efficiency gains so that energy demand stayed relatively stagnant at these facilities despite rising internet demands. But, with the integration of artificial intelligence into basic Google searches, as well as countless other web applications and jobs, that’s no longer the case.

“There’s a big difference with the use of computers for AI, in that the systems that are being deployed have very dense circuits, require a lot of computation, and that generates a lot of heat and a lot of electrical use,” said Ebert.

To add to this challenge, Ebert says it's hard to distinguish which data centers are operating sustainably and which are not.

He says that’s because when people discuss specifics of these sites, they tend to choose only one metric to focus on — such as water or energy use — and won’t explain how those are linked or how they may vary depending on the activities taking place within the facility.

“We need to get away from just thinking of consumption, to, ‘what are they doing?’” said Ebert. “A lot of data centers can use exactly the same amount of heat and cooling, whether there is no computation going on or whether they’re doing productive computation.”

He used the example that UA will power down its cloud servers when students are away for the summer, allowing the school to save money and energy.

County Supervisors Rich Vitiello and Jeff Serdy asked similar questions about locating data centers. Vitiello pointed to the fact that Arizona is a drought-prone state, while Serdy said it wasn’t entirely clear why a data center must be local, if the facilities seem to be delivering information to everyone from any location.

To Vitiello’s point, Ebert responded that many data centers are finding ways to be water-neutral, meaning they don’t consume more water than they give back through other programs like recycling or watershed restoration.

There’s also new evidence that data centers don’t need to be kept as cool as initially thought — cooling is where most data center water use tends to originate — and that dry cooling may be just as effective in a climate like Arizona’s.

“If you’re running a data center at 95 degrees for a large portion of the year in Arizona, given how dry it is, you can actually use air being blown from outside to pull the heat away from the computer systems,” said Ebert.

As far as location, Ebert says it depends on what kinds of tasks the data center will be completing for the user.

“If you Google something, you don’t need a data center next door to answer that and get the answer 10 times as fast,” he said.

“But when you’re using them in terms of deploying systems that are doing forecasting for weather, or real-time flood prediction, or doing air traffic control, or controlling emergency response systems, that multi-second to 10-second delay to get something computed remotely can mean the difference between life and death.”

The energy picture

The two-hour special meeting allowed the five county supervisors and members of the county’s Planning and Zoning Commission to air their concerns.

Two members of the commission, Robert Klob and Gary Pranzo, said they don’t know enough about the energy picture to comfortably recommend large-scale data center development.

Pranzo said the county might need to grow its power by about a third, to account for data centers and the potential loss of hydroelectric power.

“A couple months back, we had a visit with APS, and during that presentation, we learned that the county needed to grow in power by 13%,” said Pranzo. “Last month we had a visit with (Electrical District) 3, and we learned that when our hydroelectric power fails, we’re going to lose about 20 percent.”

Because multiple utilities are serving Pinal County, Klob and Pranzo agreed it's hard to know how much energy supply is available; which uses it should be going toward; and how supplies could be impacted in an emergency situation.

“Are we robbing resources from residential? Who’s going to take precedent if we have power failure or not enough power? Will the data center, because they generate cash flow, or will the residential?” Pranzo asked. “It feels like we’re heading in that direction that we want to build data centers, but we don’t have the resources to support them.”

Scott Scharli, director of strategic energy management, handles all of the utility’s large-load customers. These customers require 20 megawatts of power or more and tend to be data centers.

SRP is predicting data centers will continue to make up much of the growing demand on the utility’s commercial side and in the large-load category for years to come. For now, though, Scharli says the 20 data centers located within SRP’s service territory contributed to only a fraction — 7% — of peak energy demand recently. The rest came from residential customers.

But, about a year and a half ago, he says the utility realized it had a problem on its hands when considering new data center loads.

“We had unprecedented interest in load growth requests, and there’s a lot of misinformation to this as to, ‘well, why the heck are all the data centers coming to Arizona?’” said Scharli.

Scharli said there are three reasons so many data centers are being built here. First, there are little to no natural disasters. Second, TSMC and Intel are based here, which means fast and easy access to the latest computer chips. Third, Arizona is known for having reliable power.

The flood of demand made it hard for SRP to know which data centers were serious about building and how much new generation and transmission would be needed. So, the utility developed a new system.

Instead of having data center developers apply one by one, it had them apply in clusters. That way SRP could collect all the fees necessary to study what it would mean if several large loads connected at once.

The first study was done in April 2025 and included 24 interested data centers, representing 7.2 gigawatts of energy demand. The current total capacity of SRP’s system is 9 gigawatts. The study determined that the total transmission upgrades would cost $12.8 billion.

In order to continue the process, applicants had to agree to a 30% deposit on those transmission costs. In the end, only three customers held, bringing anticipated load growth to under one gigawatt and expected transmission costs to $1.8 billion.

Scharli says data centers will also need to commit to a 15-year cost-sharing contract for their energy generation needs, because that’s typically how long it takes to bring all the necessary equipment online.

“We are very up-front in our study results to say, ‘hey, it may take us five years to build the transmission lines, but you need to know that it could be 10 years before we get the generation for you,’” said Scharli.

He says that while some data centers claim they’re bringing their own power, that power is usually meant only as a temporary bridge while they await a connection to the grid.

“Because connection to the grid equals reliability and redundancy,” said Scharli.