Pittsburgh Is Poised to Lead the Next Generation of AI

Next week, an Energy and Innovation Summit convened by Sen. Dave McCormick and hosted by Carnegie Mellon University will bring industry leaders, academics and the president of the United States to the Oakland campus.

Aside from its high-profile draw, the July 15 summit is another in a series of ongoing conversations, such as Energy Week, about AI and energy on CMU’s campus. Still, the president and lawmakers aren’t the only ones looking to Allegheny County for AI leadership.

In April, the Department of Energy included the National Energy Technology Laboratory in South Park in a request for information as a prospective site for a national AI research hub.

The Allegheny Conference on Community Development, a nonprofit focused on economic development, spearheaded the region’s proposal, which was supported by 24 other local organizations.

“It’s a really important initiative not only for the specific project at [the National Energy Technology Lab], but because it identifies this region as one that is and should be in the lead on AI data center energy development across the globe,” said Matt Smith, the Conference’s chief growth officer.

If selected, the site will receive investment from the Department of Energy to develop and test equipment necessary for data centers, according to Smith.

Data centers are collections of connected computers that process and store digital information. They’ve existed as long as the internet itself, but the recent AI boom has significantly increased the demand for data centers, which are essential for AI training and operation.

Data centers require vast amounts of energy to power them, but Smith said the Pittsburgh area’s numerous energy options and innovations are a primary selling point.

“There’s significant access to really vast amounts of energy here — natural gas first and foremost — but we also have significant expertise … in advanced nuclear,” Smith said. “You look at companies like Westinghouse, Holtec [International], they’re doing significant things in the region around nuclear.”

The request for information response period ended in May, and Smith expects to hear back on the proposal in the coming weeks.

Pittsburgh leads the way

Although national recognition could further boost the AI sector, the Pittsburgh region is already a global leader in energy and AI development.

In January, sustainable data center developer TECfusions purchased the former Alcoa research and development site in Upper Burrell to create a data center campus that, in the next six years, will be able to handle electrical loads of 3 gigawatts (GW) — the same amount it would take to power 300 million LED light bulbs at once.

At the beginning of June, Ardent Data Centers — a subsidiary of German-based Northern Data Group — reopened a data center in McKees Rocks that it purchased last year. The site currently operates with a 2.4-gigawatt capacity, but at the event, officials announced plans to scale it up to 12 gigawatts by 2027.

In a speech at the opening on Wednesday, June 4, Wayne Ginders, Northern Data’s vice president of global operations, cited the region’s existing energy and labor infrastructure as key reasons for the company’s commitment to the region.

“Northern Data has the luxury of being able to build data centers all over the world, and we’re in McKees Rocks for a reason,” Ginders said. “We want this facility to serve as an example to the rest of the industry that Western Pennsylvania is not only open for business when it comes to AI infrastructure but has everything you need to develop and run the next generation of data centers.”

Another lure for data center developers is the region’s workforce, which boasts an abundance of skilled laborers for buildouts, as well as robotics and autonomy professionals for when sites are operational, the Allegheny Conference’s Smith said.

As evident by the projects that are already emerging, he continued, the types of sites available for such projects are inherently significant.

“It’s not just open land that we have here — we have sites that were former industrial facilities … that are not only great sites, [but] they’re already connected to the power sources, whether it’s natural gas or the grid,” Smith said.

In the short term, the biggest challenge developers face is energy availability. At a Pennsylvania Public Utility Commission meeting on Thursday, April 24, a Duquesne Light Co. representative said that the region’s peak load in 2024 was 2,700 megawatts or 2.7 gigawatts.

But local projects are emerging to fill the gap.

In April, the Homer City Redevelopment Group in Indiana Township announced that Kiewit Power Constructors Co. would turn the former Homer City Generating Station into the nation’s largest natural gas-powered data center campus capable of producing 4.5 gigawatts that can be transmitted through the local grid.

Smith called attention to manufacturers like Mitsubishi Electric Power Products Inc. and Hitachi Investments, which are both working on parts to upgrade the grid for more energy transfers.

“When you are able to put all these pieces together in one puzzle, … we’re able to present this value proposition that not only is this region capable of doing these projects, we have the skills, the resources, to solve the problems that developers are looking to solve when they think about doing projects like this,” he said.

Carnegie Mellon University paved the way

While much of the AI-driven growth in the Pittsburgh region is a byproduct of its readily available resources, it also has historic precedent connected to Carnegie Mellon University.

Reid Simmons, a research professor and director of the University’s Artificial Intelligence major, said that the field was created in the 1950s. At the time, two of its four founding fathers, Herbert Simon and Allen Newell, were teaching and conducting research at CMU.

Although AI classes wouldn’t formally be included in university course lists for at least 20 more years, the topic began to emerge in Simon and Newell’s other courses — especially in computing, which was also a relatively new field, Simmons said.

“What the founders of AI, and particularly Newell and Simon, realized is that you could use computers to manipulate symbols,” he said. It wasn’t just numbers, but you could talk about symbols — like words or variables — and by manipulating symbols, you could do logical reasoning or planning or searching.

“So, the early aspects of AI were very symbolic in nature — reasoning at a much higher level than the numerical calculations that most people were doing.”

The pair created a “thinking machine” that was programmed to solve mathematical proofs that had previously “taken 10 years of longhand work,” according to the university.

Simmons said that courses and degrees in machine learning emerged in the 1970s and 1980s but were initially available only to graduate students. Throughout the 1990s, the programs began trickling down to the undergraduate level. Simmons started working at CMU in 1988.

Throughout the remainder of that century and into the 2010s, the program remained relatively unchanged, occasionally experiencing spikes in interest with the rise of neural networks and the Dot-com bubble.

It wasn’t until May 2018 that CMU launched an undergraduate program in AI.

“I was just finishing up my stint at [the National Science Foundation] and they offered, when I came back, to finish establishing the program and direct the program,” Simmons said. “I jumped at the opportunity. It seemed like a great thing.

“Who would have known that four or five years later, AI was going to go through the roof?”

Since then, the program has averaged about 40 graduates per year.

To Simmons, the serendipity of Simon and Newell working at CMU in the 1950s served as the first wingbeat in a butterfly effect.

“They attracted students and colleagues with like mind, and those students and colleagues attracted people of like mind,” he said. “It was just fed on itself.”

Today, prospective students are drawn to AI programs at CMU not only because the institution has a rich history in the field, but also because Pittsburgh’s establishment as a tech hub enables students to secure internships and jobs more easily.

“And to be frank, when people come here, they like Pittsburgh and they’re happy to stay,” Simmons says

“They don’t have to go to the Bay Area, they don’t have to go to Boston in order to reap the same benefits. They can come here and stay here, and they’ve got those opportunities.”

Top Photo: Sen. Dave McCormick tours the National Energy Technology Laboratory in South Park. (Courtesy of NETL)