Marcellus Shale Proximity Drives Billions in AI Investment

The data center build-out in the US mid-Atlantic region is attracting tens of billions of dollars in investment, driven by the vast reserves of natural gas in the Marcellus and Utica Shales. But that enthusiasm is being tempered by persistent challenges in permitting and building the necessary network of gas pipelines and power plants.

Pennsylvania’s potential to become a global artificial intelligence hub was touted at an event this week in Pittsburgh attended by officials ranging from US Energy Secretary Chris Wright to Exxon Mobil CEO Darren Woods — and included an appearance by President Donald Trump.

Specifically, the Pennsylvania Energy and Innovation Summit highlighted the role the state’s 21 billion cubic feet per day of gas production could play in making the US a global leader in AI development. The state currently ranks 14th in the number of existing data centers, according to Datacenters.com.

“Today, you will hear companies commit more than $90 billion to make sure the road to AI and energy dominance comes right through Pennsylvania,” US Sen. Dave McCormick (R-Penn.) told the summit, which he helped organize. “The AI revolution has the potential and promise to transform our nation’s economic outlook.”

Trump later said Pennsylvania had received firm or preliminary private-sector agreements to spend $56 billion on new energy projects and $36 billion on data centers in the state, “and a lot more than that are going to be announced in the coming weeks.” It’s unclear how much of that represents new investment, and he provided few details of those agreements. However, he did highlight a pledge by global investment firm Blackstone to invest more than $25 billion.

“What makes us so excited about this area is that you can colocate the data centers next to the source of power,” Blackstone President Jon Gray told the event. “The special sauce here is putting these together, because building transmission lines, pipelines — that’s really difficult.”

‘Turn the Wrenches’

AI data centers require large amounts of round-the-clock electricity, meaning they will need a rapid build-out of power generators. Gas is expected to be the predominant fuel for those plants in the short run, spurring AI developers to look for sites in close proximity to large gas reserves and robust pipeline networks — both of which Pennsylvania offers.

But to continue attracting the vast sums of capital required to grow the AI sector, hyperscalers must have confidence that additional pipelines, power lines and generators can be quickly approved and built to avoid over-stressing the grid.

“We are sitting on one of the biggest gas fields in the world, but its potential has been throttled back” by anti-fossil fuel policies, Toby Rice, CEO of Pittsburgh-based gas producer EQT, told the forum. “We need to unleash the potential of these resources. … We are going to need to create both federal and state policies that will accelerate all of this, and we’re going to need to turn the wrenches to make it happen.”

Brendan Bechtel, CEO of engineering giant Bechtel, suggested the US would fall behind China and other nations if it doesn’t speed up energy projects that currently take years to permit and construct.

“We need comprehensive permitting reform, and the moment is right now,” he said. “It’s when [Congress] comes back from the August recess and before the end of the year. If it doesn’t happen then, it’s not going to happen for another generation.”

Bechtel said congressional permitting reform, which has failed to advance for years, “is the single biggest thing that could help this AI and energy infrastructure build-out. There is a lot of bipartisan support, we hear it in private every day.”

He said the most important aspect of such reform must involve limiting judicial interference. He lamented that lawsuits targeting large federally approved pipelines take an average of four years to resolve.

“About 80% of those litigated end up with a permit exactly as approved,” Bechtel said. “That is four years lost.”

Beyond the Keystone State

While Pennsylvania’s wealth of gas reserves may propel the next wave of AI development, the data center build-out is well under way in the Washington, DC, and the suburbs of northern Virginia, currently the largest AI hub in North America.

And growth in that region is set to explode, driving the installation of up to 15 gigawatts of gas-fired generation capacity across the PJM Interconnection by 2030, according to a recent report from Aurora Energy Research.

That could pose some major challenges to the power grid under current market conditions and give potential AI developers pause. “Adding 10-15 GW of firm generation capacity needed to supply these data centers and keep the lights on in Virginia will not be easy,” said Zachary Edelen, PJM research lead at Aurora. “It can take three to four years for the transmission organization just to green-light a new generator, and market prices are currently too low for developers to build the kind of capacity required.”

Dominion Energy, the main utility serving northern Virginia, has said it had 40 GW of data center capacity in various stages of contracting as of December, up nearly 90% from six months earlier.

Because the region must pipe the gas in from Pennsylvania, the whole mid-Atlantic pipeline grid could become even more strained, analysts say, creating further pressure on regulatory changes.

“I don’t think permitting reform is about taking anybody out of the process,” GE Verona CEO Scott Strazik told the Pittsburgh conference, adding that public input on energy projects should be preserved. “But it’s like a game; there needs to be a shot clock.”

Don’t Forget Nuclear

While gas is expected to take the early lead in powering the new generation of data centers, the nuclear industry is looking to position itself as a carbon-free alternative.

“Even if you build renewables, if you build gas, if you build coal, you will not build enough if you don’t also use nuclear energy,” said Bruce Flatt, CEO of Brookfield Asset Management, which owns 51% of Pennsylvania-based nuclear energy developer Westinghouse.

“Today, the discussion with most companies is, ‘Can you get us power, and when can you get us power and how can we get it delivered?’” Flatt told the forum, with most being agnostic on the fuel source.

He said Brookfield is talking with AI developers about building small on-site nuclear plants that would be “sole-sourced” into data centers, which would help take pressure off an increasingly strained grid.

“The number one thing that will slow down this build-out of energy is if we start blacking out people’s power,” Flatt said. “We need to ensure we don’t do that, and that’s why baseload nuclear is super, super important.”