LaCross AI Institute Series Explores Opportunities and Challenges in Data Center Development

Since artificial intelligence stepped into the mainstream with OpenAI’s launch of ChatGPT in 2022, growth in the industry has exploded. Construction of data centers, the engines that power the digital world ramped up. Today there are more than 11,000 around the world—with more in the works.

On 17 February, the LaCross Institute for Ethical AI in Business at the UVA Darden School of Business convened industry experts and faculty, students, staff and alumni for a deep dive into data centers. Against the backdrop of historic sector growth, participants discussed the current state of data center development.

The event was part of the institute’s Value Chain of Ethical AI series. The series explores the value chain of artificial intelligence which offers a framework for making sense of the massive ecosystem where many rapidly evolving opportunities, challenges, and ethical considerations vie for attention, says institute director Marc Ruggiano.

The value chain examines five areas: infrastructure, measurement and data, algorithms for models and training, new business applications and managing AI and monitoring its outcomes. These are described in more detail in a forthcoming LaCross AI Institute white paper – Exploring the Value Chain of Ethical AI.

“This approach makes it easier to think of ethical AI decisions in a broader context. It opens the aperture for potential solutions to specific problems that can sometimes seem intractable when they are approached more narrowly,” Ruggiano said.

Panels made up of industry experts and faculty noted the AI sector, though now in its fourth decade, is still in its nascency, making this a new era ripe for innovation as data centers evolve.

The building boom

Dan Ephraim, vice president of sales and leasing at PointOne Data Centers, who sponsored the event, joined Ed Socia, regional director for North America at DatacenterHawk, to discuss the nuts and bolts of digital infrastructure: not only the data centers but the land they sit on and the resources needed to run them effectively: chips, storage, fiber, optics, networking—as well as the drywall, generators, switch gear, concrete, and mechanical-electrical components inherent in a physical plant.

“We are the plumbers of the internet,” said Ephraim, whose company builds data center campuses in Virginia and North Carolina. “Physical assets are so critical—you just can’t ignore it. The industry needs people to sell it, they need people to develop it, and to operate and manage it. And it’s only going to grow.”

“We want inquisitive people—who love to learn, who love to build, who love to think—to be in this space,” Socia added.

Data centers, measured in power use rather than square feet, range from double-digit megawatts to gigawatt facility. New builds are increasingly large out of the gate, contributing to what Socia calls the No. 1 bottleneck in developing digital infrastructure: the ability to find power.

“We used to say bring the power to the data center. Now we say bring the data center to the power,” he said.

The Value Chain of Ethical AI series explores the value chain of artificial intelligence which offers a framework for making sense of the massive ecosystem.

The power challenge

The amount needed to power the U.S. at peak load is 530 gigawatts, said energy finance consultant Patrick Worrall (MBA ’02).

“When you start talking about even tens of gigawatts for a single data center, that’s a massive amount of power,” Worrall said. “This type of industrial growth is like nothing that we’ve ever seen. When you talk about adding that kind of capacity on an already constrained system, it’s a lot.”

Development has only intensified an existing need to upgrade the grid, panelists agreed.

“Utilities can no longer assume steady state. It’s a step function increase in the amount of demand,” said Andy Stewart, CEO of Iconic Digital. Additionally, the 10- to 20-year planning cycle of a utility doesn’t work with data centers, typically planned out for 10 weeks to two years.

Hungry for power sources, companies are combining a variety—from wind to solar to natural gas—depending on their use cases, Worrall said. “It’s not just a matter of which one; it’s a matter of all of them.”

Addressing community concerns about water supply bottlenecks and wastewater treatment, panelists noted water-saving technologies continue to advance. Closed-loop cooling systems and new infrastructure, such as retention ponds and pump station upgrades, show promise.

Financing AI infrastructure

Roughly $7 trillion in global data center expenditures are projected by 2030, 40% of it in the U.S., said Joe Harar (UVA ’05), CFO of EdgeConneX, a major developer. The seven largest companies in the world make up about 90% of the market—EdgeConneX’s customers among them. Many are investment grade and AAA rated.

“It doesn’t insulate us by any means, but they’re making long-term investments. Customers are signing 10- to 15-year leases,” Harar said.

Traditional real estate and infrastructure finance options can’t cover large projects. With the rising cost of capital and longer contracts for servicing debt, developers are increasingly turning to private equity. The data center industry is now the largest investment category for PE firms.

Developers are high-intensity, capital-intensive businesses that need to think in terms of how their customers are parlaying their leases over five to 15 years, rather than a quarter at a time, said Irtiaz Ahmad, president of Sobina Advisors. “Being outside the public light has been a bit of a blessing to developers.”

The technology stack

Technology is the heartbeat of a data center, and it gets more powerful practically by the day, with GPU racks scaling rapidly in power demand and density.

While the size and density of a project depends on the use case, space considerations are a balancing act for all of them, said Emerson Whitney (MBA ’24), who works in the supercomputing group in Microsoft Azure.

“For a big training supercomputer, we need all of it: the GPUs, the storage, the traditional CPU, and the networking, all together.”

A shortage of readily available organic data has also led in recent years to a surprising boost in demand for CPUs for synthetic data generation, he added.

Chip innovation illustrates the need for speed across the sector. By the time a multisite supercomputer and infrastructure are ready to go, the chip they are designed around is likely not the latest-generation silicon. However, Whitney said, this is generally not an issue for companies about to press start.

The career landscape

As the industry scrambles to keep up with the rapidly evolving technology that drives it, a talent shortage ranks as an acute challenge for the sector, along with power constraints and supply chain bottlenecks.

In a session devoted to careers, recruiters discussed opportunities and answered questions from Darden students. Among key takeaways: Successful candidates thrive in matrix environments and think strategically. They have had a relevant internship or other practical experience.

They’ve also done their homework on the industry, the company and the job they’re applying for, and they bring thoughtful questions to the interview, said Abbey Kang (EMBA ’21), a senior manager at ServiceNow.

Development, supply chain, power generation, community engagement, finance, real estate, and technology all need people with curiosity, initiative, and a desire to learn, said Trey Dean, global vice president of Vantage Data Centers.

Jobs abound in construction, real estate, site development, project management, and facility engineering, as well as G&A areas like finance and insurance. Many of the jobs are well-paying roles in traditional companies that build the physical plant with drywall, cement, fiber, optics, lighting, and electrical and mechanical systems.

“The supply chain is very constrained right now. That’s our biggest risk,” Dean said. “Some of the things that you don’t instantly associate with data centers are the most challenging areas to find people. It’s very, very hard to find people that are able to do those jobs.”

For candidates eyeing careers in larger tech companies, Kang cautioned that they tend to hire from within first.

“Be open to just getting into the company, and you’ll have more options to move into what you want to do,” she advised.

Once you’re there, put your hand up, added Kendall Jennings Hayden (MBA ’12), a briefing lead for the CEO at Accenture.

“You have to be naturally curious in a technology firm, where people are changing jobs every year. You don’t have to be perfect. But to have that curiosity about AI and the technology and trying it out yourself is really important.