Can Western Pennsylvania Become Silicon Valley East?
How do you go from 100% to dead? That is the tragic enigma which changed Jim Foote’s life. In 2004, Foote’s 15-year-old son, Trey, was diagnosed with bone cancer. After chemotherapy and surgery, doctors told him, “We never say it’s [cancer] 100% dead. It’s 100% dead. You’re good.” Three months later the cancer returned. In October 2006, Trey Foote died; he was 17.
Bereft, Foote was haunted by the riddle “how do you go from 100% to dead?” When he learned bone cancer treatment protocols had not budged in 25 years, Foote saw an industry in need of transformation. Today, he is the co-founder of First Ascent Biomedical and sits at the forefront of twin technological innovations in western Pennsylvania: artificial intelligence and cancer mapping.
The birthplace of the petroleum and steel industry, western Pennsylvania, is a century removed from its economic zenith. Deindustrialization has turned the region into a center for deaths of despair and Trumpist politics. But Joanna Doven thinks Pittsburgh can become Silicon Valley East for artificial intelligence, data centers, and advanced manufacturing. Doven, the executive director of Pittsburgh’s AI Strike Team, told me, “The Pittsburgh region has the full stack of the AI economy. The same reason for the steel revolution is the same reason for the AI revolution.” Blessed with abundant water and natural gas, western Pennsylvania has the shovel-ready sites and university brainpower for an industry about to explode. In the next five years, analysts estimate that $6.7 trillion will be spent to build AI data centers worldwide.
Powering the AI revolution are the soon-to-be ubiquitous data centers. Dr. Kefei Wang, a computer science professor at Gannon University, noted that, “A data center is a fancy term for a giant room with a bunch of computers.” The data center’s physical footprint – an average-sized factory – belies their power density. Elon Musk’s Memphis-based supercomputer “Colossus,” for example, houses 100,000 NVIDIA H100s, each of which devours up to 700 watts of energy per day. Wang told me that one data center “consumes the same energy as a mid-sized city.” Powering billions of semi-conductor transistors, a data center computes gob smacking amounts of data cheaply.
Foote, a former Fortune 200 cyber-security executive, said, “AI finally is delivering all of the promises that’s been made for decades.” For First Ascent, and so many other tech entities, low-cost, high-powered computing is the difference between bankruptcy and breakthrough. Foote explained his company’s cancer mapping: “Every patient has a unique biological fingerprint. For every patient, certain cancer drugs will work, and other drugs may be lethal. First Ascent’s cancer mapping takes a patient’s biological fingerprint and matches it to the proper standard of care.” To do so requires enormous amounts of computing power. AI data centers offer it. As Foote told me, “Ten years ago to process this data was hugely expensive. Now you can process it for an individual patient for pennies.” When doctors use First Ascent’s platform to guide treatment, refractory cancer patients outperformed standards of care by 83%.
As with any significant innovation, data centers carry heavy environmental costs. David Masur, the executive director of Penn Environment, said, “In Virginia [a data center mega-site] 25% of electricity is used for data centers. By 2030, 10% of electricity will be going to data centers.” And that thirst for energy, according to Masur, means “they [data centers] keep dirty energy online. Dirty fossil fuel plants that were slated to close now are ordered to re-open.” In other words, data centers mean more fossil fuels are emitted in an already warming planet. And this is only one piece of the environmental puzzle. The electricity pouring into data centers create terrific amounts of heat. Wang equated it to having “50,000 space heaters in a single factory.” To cool the computers requires millions of gallons of water. To put that on a human scale, Emma Bast, a staff attorney with Penn Future, told me, “Every ChatGPT use to summarize an email uses a bottles’ worth of water.” Multiply that by a billion AI searches and a data center’s immense water usage comes into relief.
In addition to the dire environmental costs is the AI “hype-cycle.” In a tech hype-cycle, proponents so inflate expectations that an inevitable “trough of disillusionment” ensues. Dr. Tamara Kneese has skipped the Christmas rush and headed straight to the disillusionment stage. The Program Director for Data & Society’s Climate, Technology, and Justice initiative, said, “AI can’t do all the things they say it can do. I have no interest in using AI in my writing. That is the problem with this hype-cycle, it is like the meta-verse. AI is a hype-bubble.” This hype-bubble fuels much of the breathless predictions of a world turned upside-down by AI. Wang sees the hype-bubble up to a point. He cautioned, “AI. What can it do today? Not a lot.” But he counseled, “There is a clear and bright future in the field, that’s why there is a huge investment.”
Across the U.S., the AI hype-cycle has promised communities that data centers will bring economic riches. Kneese warns that “clearly there are [temporary] construction jobs but under 50 [permanent] full-timers.” In other words, data centers can soak the water and suck the energy from a community – yet employ remarkably few locals.
Conor Lamb, a former two-term western Pennsylvania U.S. Representative, is aware of this. As he told me, “The data centers, this is coming. America needs to be in a place where it is thriving. We should be building [them] in Pittsburgh. My only hope is that the public gets a good deal.” And Lamb’s “good deal,” economically and environmentally, is Doven and the AI Strike Team’s aim.
For Doven, Pittsburgh’s “full stack” of talent, energy, water, and industrial sites, means western Pennsylvania can avoid the AI hype-cycle. “The Strike Team’s true north is jobs,” said Doven. One data center will scarcely disrupt western Pennsylvania’s economic trough. Dr. Kyle Kopko, the executive director of the Center for Rural Pennsylvania explained to me: “A one-off is not enough. If you have 2, 3, 4, 5 in an area that could be enough to revolutionize it.”
But having the full stack means scores, if not hundreds, of data centers dotting the landscape. And Doven explained these “flywheel for re-investment” means “[once] the big tech flag is planted that can result in hyper-scalers and follow-on investment.” As an example, she pointed to New Albany, Ohio, where nearly 100 data centers brought $1 billion in investment. Though this means jobs, Doven noted that, “Jobs in AI are not just software engineers. 70% are physical labor: electricians, steam fitters, HVAC, and data labeling.” Data centers, in the AI Strike Team’s plan, will make western Pennsylvania into a locus of advanced manufacturing and assorted tech jobs that can make the region as it was in the past: an economic hub delivering middle-class pay to working-class Americans.
Environmentalists are rightly worried that data centers only bring noise, dirty air, and biblical levels of water use. Masur advises: “Data centers should be powered with sustainable energy.” Bast urges that “data centers should have a BYOP (bring your own energy) policy.” She rightly fears that data centers will force an overtaxed grid to tap into fracked gas, slag coal, and even diesel fuel. Steve Brame, the president and CEO of the Pennsylvania Rural Electric Association, warns that a surge in energy demand from data centers has resulted in “[energy] prices going up by 800% in some areas. $27 megawatts per day became $200 megawatts per day.” But Brame sees a solution: nuclear power.
Pennsylvania has a complicated history with atomic energy. In 1957, Beaver County became home to the nation’s first-ever commercial nuclear power facility. In 1979, Dauphin County’s Three Mile Island suffered a partial nuclear meltdown and only narrowly avoided a cataclysmic disaster. The latter understandably cooled public fervor for atomic energy. But Brame reminded that nuclear power is commonplace in the Commonwealth. “Nuclear already provides 60-65% of power generated by the Pennsylvania Rural Electric Association,” he said. The nuclear energy of the present is not that of the future. Doven shared that Westinghouse is developing small, modular nuclear reactors to power western Pennsylvania’s data centers. The size of a telephone booth, the reactors will be powered by spent nuclear fuel. Far from a vision of Dr. Strangelove, the European Commission has recognized their utility in achieving the EU Green Deal. More important, mishaps with small, modular reactors avoid the cataclysms associated with accidents at giant, old generation nuclear energy plants.
Some data center critics are open to such solutions. Kneese said, “A lot of people want to be cautiously in favor of small nuclear power facilities. I think that could be a possibility.” As unsettling as nuclear power may be for some Americans, Wang stated the hard facts: “We don’t have enough electricity for our computing needs. Even if we combine all the electricity in the world, we don’t have enough. Nuclear power is the future, it is the most efficient. The latest module is supposed to be a lot safer.”
China powers its AI data centers with coal. Western Pennsylvania will power theirs, for now, with cleaner but still dirty, fracked natural gas. But as Doven said, “Don’t make perfect the enemy of the good.” The AI Strike Team’s plan is to transition from fracked natural gas to a more sustainable nuclear source. In a world of hard choices where generative AI is recognized, even by critics, as the future, nuclear is a consensus solution. Beyond the energy source, the key is to avoid what Masur terms a “build at any cost” mentality or what Kneese calls “a free-for-all.” Threading the needle between proper regulation and needless hurdles is the political task of the moment. The former can lead to environmental risks that degrade a community’s resources without economic benefit. The latter, meanwhile, could mean, as Doven warned, “missing the way we [Pennsylvania] did with [the] semi-conductor [industry].”
In the Rust Belt, threading that needle means jobs and even life itself. First Ascent’s lab at Gannon University in Erie is purported to create 38 jobs paying between $60,000 and $150,000. Five years from now, those positions will number over 100. But Foote understands there is a larger meaning for the 150,000 in Greater Erie – an 150-mile region encompassing parts of Ohio, Pennsylvania, and western New York – who die every year of cancer. “Cancer pissed off the wrong guy, and I’m not going to go silently into the night. I lost the battle but I’m going to win the war.”
