Banking Dive published the interview on Tuesday. Gabrielle Saulsbery spoke to PNC chief executive Bill Demchak at Money20/20 in Las Vegas the week before. His argument is simple. Banks that do not own their compute , meaning their own data centres and graphics chips, will fall behind on AI, and that will drive mergers.
"We want our own compute and our own data," Demchak said. "I don't want to share it with anybody else." Many banks cannot follow, in his view, because they are "still running mainframes," while others have bet heavily on the cloud. "Some people just don't have the investment dollars," he added.
Then came the line that matters for anyone selling a bank. "I think this transformation is going to cause a lot of mergers over the next handful of years, particularly with small banks," Demchak said. "A target-rich environment would be my expectation."
I read that as a forecast from a buyer who is choosing not to buy right now. That combination is unusual. I would expect a buyer who sees cheap targets coming to want to be first in line.
Demchak explained the pause in the same interview. "If we did a large deal and spent two or three years vetting everything onto the same platform … at that point you'd have to take a timeout on AI," he said. "This game could be over in two years."
I went looking for the earnings call to see whether this was new. The idea is not new at all. On 15 October, according to the transcript I read, he told analysts he was "hesitant to slow that down which would perhaps necessarily be the case were we to do an acquisition of any size." He went further: "If that gets slowed down because we otherwise got into some big M&A deal, I think you could destroy massive value."
He left room for small purchases. "There's always small things," he said on the call, and "you can never say never." Yet his summary was blunt: the coming transformation "is strategically more important than bulk."
The cost line is the part I keep thinking about. Demchak told Banking Dive the AI programme will cost "hundreds of millions to save billions" a year. He gave no breakdown, and he said a roadmap will come before the three-year mark he set for "meaningful progress."
The call transcript shows where the money already goes. Demchak described "creeping depreciation expense on the back of all of our data center rebuilds," adding that PNC has put GPU chips in its data centres. The aim, he said, is infrastructure "without vendor support, without compute costs coming from third parties and with your own personnel instead of vendors."
He was frank about jobs too. Much of the change will mean "offshore contractors no longer being employed with us," he told Banking Dive, and coders across the technology sector will feel it.
PNC is not a neutral observer. It is still integrating its FirstBank acquisition, and its third-quarter results carried $27 million of integration costs, against $261 million of integration costs and significant items in the second quarter. Demchak knows what an integration costs in money and management time.
For a small bank, his forecast cuts two ways. If AI spending forces consolidation, sellers will be plentiful. If the largest regional buyers sit out to protect their AI programmes, there may be fewer bidders for them. I cannot tell which effect wins, and Demchak did not say who the buyers in his "target-rich environment" would be.
My view is that his two-year deadline is the claim to test. I still do not know what PNC will publish in its roadmap, and that document will show whether "hundreds of millions" is a budget or a talking point.