PNC 称将自建 AI 算力基础设施、减少对外部科技供应商依赖，暂无量化业绩信息

While most banks talk about their adoption of artificial intelligence in terms of how much to spend on outside vendors, PNC Financial Services Group is planning a more independent path: doing much of it on its own.

The goal is to build an infrastructure that “allows you to do it yourself without vendor support, without compute costs coming from third parties,” Chief Executive Officer Bill Demchak told analysts on a conference call Thursday.

The approach differs from what many of PNC’s competitors have discussed publicly, with most of them focusing on matching tasks with the right AI models, so that they don’t overspend on tokens. Pittsburgh-based PNC has instead been investing directly in the underlying computing infrastructure, putting GPU chips into its data centers to increase their power.

“It’s so easy to get hooked on the bells and whistles and cool toys that come with AI, but to succeed long-term, I think you need to have the core infrastructure that allows you to do this at scale,” Demchak said. “We’ve been purposefully quiet on all of the above. We’ll show up when we have something positive to show you.”

Demchak said the next step is to capitalize on what’s been several years of AI investments to use the technology to lower costs and develop new products.

That transformation could play out in the next two to three years, Chief Financial Officer Rob Reilly said on the call.

Demchak said he would prioritize the AI work before he considers other transformative moves, such as making a big acquisition.

The proliferation of AI agents has sparked concerns about deposit outflows, as robots could automatically move money from one bank to another that offers higher yields, just like how they can help make decision in shopping. Demchak said that would be “incredibly dangerous” if allowed to happen without secure connections and authentication.

It’s “not going to happen on my watch,” the CEO said.