- Key Insight: PNC is building its own AI models and data centers so it can use the technology without relying on tech companies.
- Forward Look: As AI becomes increasingly essential to banking, the Pittsburgh-based bank hopes to avoid the costs of working with third-party vendors.
- Expert Quote: "We are building our own AI factory. … That's a big deal." — PNC CEO Bill Demchak
Earlier this week, the Pittsburgh-based bank's CEO, Bill Demchak, said PNC is committed not only to using artificial intelligence, but increasingly to producing it in-house.
"We are building our own AI factory," Demchak said at the Morgan Stanley U.S. Financials Conference in New York. "We will have our own GPU compute. We will not be as reliant on burning external tokens [as] what we will do internally for our own large language models. That's a big deal."
In practical terms, that means the $603 billion-asset bank is acquiring its own data centers, buying Nvidia chips for its own use and building its own large — and small — language models for solving specific, banking-related problems, like blocking fraud and running call centers.
The goal, Demchak said, is not to compete with tech companies, but to reduce PNC's dependence on them. If AI providers were to increase their prices, PNC does not want to be at their mercy.
"As we roll forward, any impact that AI can have on the productivity of a bank, that productivity can be taken away by the cost of tokens," Demchak said, referring to the units of data used to calculate AI usage fees.
PNC is not alone in taking this approach. Many large banks, including JPMorganChase and Bank of America, own their own data centers. And BofA CEO Brian Moynihan has
But few banks have charted a course toward AI self-sufficiency as explicitly as PNC has. In an emailed statement, a PNC spokesperson confirmed that the bank is positioning itself to "avoid vendor lock-in."
"We are focused on reducing long‑term reliance on external, usage‑based pricing models," the spokesperson said. "Over time, that includes building more internal capabilities and selectively developing models or model components to support specific business use cases."
That doesn't mean PNC has stopped working with outside tech vendors. So far, the bank is still using a combination of internal and external sources for its AI.
"We continue to evaluate and use a mix of third‑party models and internally developed capabilities, depending on the use case," the spokesperson said.
But the bank is also finding ways to avoid paying for outside technology. In some cases, that's by tapping into open-source large language models. In others, it's by building PNC's own "small language models" that may not be on AI's cutting edge, but are perfectly adequate for particular banking tasks.
"What we have learned is you need to make sure that you've got the right tool for the right problem," Ned Carroll, PNC's head of data and automation, told American Banker. "Sometimes that tool is a smaller language model that maybe is [on-premises]. Sometimes the right tool is accessing a large language model" through a third party.
Demchak's comments at the Morgan Stanley conference pointed toward the need for this kind of flexibility.
"You get lots of people now saying, 'Hang on a second, these tokens are really expensive' — which they are," Demchak said. The most efficient AI tool, he said, "oftentimes isn't the $35 token, it's the $1.50 token, or may well be our token inside of our own data centers."
