AI Boom Drives U.S. Bank Commercial Lending to Fastest Growth in Nearly a Decade as Bankers Simultaneously Tighten Risk Controls
nashnova research
Commercial loans at America's 25 largest banks grew nearly 10% year-on-year — the fastest pace in over a decade — yet bank stocks are up just 5% this year versus the S&P 500's 13%, signaling deep market doubt over whether the AI-driven credit expansion can last.
How fast is lending actually growing?
Fed data show that C&I loans — commercial and industrial lending — at the top 25 US banks rose nearly 10% year-on-year in Q3, the fastest clip in over a decade once Covid-era distortions are excluded.
Commercial real-estate construction loans at large banks also grew quarter-on-quarter for the first time since early 2024. This means → capital is flowing not just into software and compute, but into the physical buildings that house AI.
In plain terms = AI construction is revving the bank lending engine to a speed not seen in ten years.
Lending is surging — so why are bank stocks lagging?
The KBW Nasdaq Bank Index is up roughly 5% year-to-date. The S&P 500 is up nearly 13%. Banks are clearly trailing.
This reflects one core worry: how long can the AI buildout last? If capital spending slows, today's new loans could become tomorrow's bad debts.
In plain terms = banks are earning the interest, but investors doubt the math works out in the end — so they refuse to pay a higher multiple.
What are big banks and regional banks each betting on?
Large banks — JPMorgan, Wells Fargo and peers — primarily serve hyperscalers (the giant cloud operators like AWS and Azure) and major developers, providing loans or bond-underwriting support for data-center construction.
Regional banks lend more to the companies that physically build data centers and smaller firms riding AI capex. Fifth Third Bancorp CEO Tim Spence put it plainly: "We're not lending to the people funding data centers — we're lending to the people building them."
This means → risk is layered: big banks bet that tech giants won't default; regionals bet that construction won't stall. The two layers are linked, yet each has its own blind spots.
What exactly are bankers worried about?
JPMorgan CFO Jeremy Barnum warned that AI is spawning innovation across sectors and capital is being mobilized through many mechanisms: "We need to track very, very closely to make sure nothing is being missed."
Wells Fargo CFO Mike Santomassimo stressed the bank is focused on "who the tenant is, where the collateral sits, and all the layered risks that could have an impact," adding that Wells has "very strict limits on what we will and won't do — and we stick to them."
The Wall Street Journal reported months ago that large banks had already begun capping loan sizes to specific clients such as Oracle. This reflects concern not just about aggregate exposure, but about single-name concentration and the tangled web linking multiple AI financing arrangements.
What do regional banks' guardrails look like?
Citizens Financial Group acquired DH Capital — a boutique advisory firm focused on data-center infrastructure — about four years ago, then committed its own balance sheet to AI-linked deals once the boom took off. But it simultaneously erected multiple guardrails.
The core logic: lend only to data-center projects that have signed hyperscaler leases, secured power-supply commitments, and obtained regulatory approvals — and hedge part of the exposure.
CEO Bruce Van Saun was the most candid: "Even though data centers look like a can't-lose proposition right now, you always think: could the bubble burst one day? We have to consider those scenarios."
Can this credit expansion land softly?
Record-decade lending growth and bank stocks trailing the market coexist. This means → the market has no consensus on whether AI-driven credit expansion can achieve a soft landing.
Bankers collectively tightening controls and publicly emphasizing limits — rather than growth — is both an internal risk signal and an open response to market skepticism.
In plain terms = banks are stepping on the accelerator with one foot and buckling the seatbelt with the other. That "charge-but-fear" posture is itself the truest snapshot of where the AI credit cycle stands today.
市场有风险，内容仅供研究参考，不构成投资建议。