高盛在6月评估投资级非金融企业如何讨论AI之后，本期转向评估美国与欧洲银行在财报电话会及其他披露中的表述。核心结论是AI对银行信贷构成利好，主要通过两条路径：一是银行可借顾问、股权与债券承销、银团及资产管理等收入变现AI资本开支周期，同时一般避免留存长久期基础设施风险；二是内部部署AI以支持自动化与成本控制。AI资本开支主题在美国银行中最为突出，效率提升则在美欧两地均有体现：美国银行提及AI资本开支的频率约为非金融企业的九倍。大部分银行尚未明确披露AI相关直接敞口，芝加哥联储数据显示大型美银对AI相关行业的直接敞口仅占总资产0.8%、约占一级资本9%，计入未提取承诺后升至25%；银行整体保持资本较轻的"发起-分销"模式，AI活动已贡献今年至今投行业务量与收入增长的约50%至55%。
Following our June review of how investment-grade non-financial firms discuss AI, we assess US and European bank commentary in earnings calls and other disclosures. Our key conclusion is that AI is likely to be credit-positive for banks through two channels. First, they can monetize the AI capex cycle through advisory, underwriting, syndication, asset-management and selective lending revenues. Second, they can deploy AI internally to support automation and cost discipline. The AI capex theme is most pronounced among US banks, but the efficiency angle is evident in both regions. Banks have consistently discussed AI more frequently than non-financial firms since 2025H2, and US banks are roughly nine times more likely than non-financial corporations to reference AI capex on earnings calls. Direct retained balance-sheet exposure to AI infrastructure remains limited so far, though broader AI-related balance-sheet usage may grow as the capex cycle broadens. We reiterate our overweight recomm