美国六大银行的 6 位 CEO 在 2025 年各自都获得超过 $40 million，合计收入接近 $260 million；相比之下，2024 年为 $213 million。文章指出，这不包括特殊期权授予，例如 Jamie Dimon 在 2021 年约 $50 million 的留任奖金。Davies 表示，JPMorgan 在科技上的支出高于任何其他银行，且在六大行中，其薪资与福利支出占营收比重最低；Goldman Sachs 排名其后，并在 20 年内降低了劳动收入占比。他还说，自约 2010 年以来，CEO 薪酬增长快于普通员工成本，从 Goldman Sachs 与 Morgan Stanley 的约 120 倍，到 Jane Fraser 最新 Citigroup 薪酬方案的超过 300 倍。
Davies 描述了一条可能的 AI 路径：银行需要更少员工、薪酬差距收敛、薪资水准下降，CEO 也包含在内；但他认为完全自主不太可能，因为策略选择仍需要人类。他提到委托人—代理人问题，并表示高阶主管主要以股票支付以与股东利益一致，同时仍作为所有者与企业之间的中介。他举例包括 ArabesqueAI 的 AutoCIO 平台，在其中人类仍要选择目标，例如收益对成长、在地对全球焦点、以及气候限制；他也说 AI 无法对潜在黑天鹅提出假设，且在人员训练、文化、层级，以及与监管者、政治人物与客户的关系上，仍需要人类判断。
Paul J. Davies wrote on February 23, 2026 that pay for top bosses at the biggest US banks has reached new records, above the pre-crisis peak of 2007, while discussion about AI job automation has raised questions about the future role of executives. He says the US labor share of gross domestic product fell to less than 54% in the third quarter of 2025, the smallest recorded by the Bureau of Labor Statistics in data going back to 1947, versus the low 60s about 20 years ago.
The six CEOs of America’s big banks each received more than $40 million for 2025, with combined income near $260 million, compared with $213 million for 2024; the article notes this excludes special options grants such as the roughly $50 million retention bonus for Jamie Dimon in 2021. Davies says JPMorgan spends more on technology than any other bank and has the lowest salaries-and-benefits bill as a share of revenue among the big six, while Goldman Sachs is next and has reduced labor’s income share over 20 years. He also says CEO pay has grown faster than average employee costs since about 2010, ranging from roughly 120 times at Goldman Sachs and Morgan Stanley to more than 300 times for Jane Fraser’s latest Citigroup package.
Davies describes a potential AI path in which banks need fewer workers, pay differences compress, and pay levels decline, including for CEOs, but he argues full autonomy is unlikely because strategic choices still require humans. He cites the principal-agent problem and says executives are paid mostly in stock to align with shareholders, while still acting as intermediaries between owners and businesses. He gives examples including ArabesqueAI’s AutoCIO platform, where humans still choose goals such as income versus growth, local versus global focus, and climate constraints; he also says AI cannot hypothesize about potential black swans and that human judgment remains necessary for training, culture, hierarchy, and relationships with regulators, politicians, and customers.