Apollo raises specter of an AI agentic 'bank run' hitting financial industry

The Invesco KBW Bank ETF fell 2.3% last week and is down 5.7% month-to-date, heading for its weakest month since its nearly 10% decline in March 2025. Goldman Sachs and Bank of America have each dropped more than 8% this month, while Charles Schwab and Interactive Brokers are down 9.8% and 8.3%, respectively, as multiyear-high Treasury yields raise concerns over slower loan growth. Apollo's Torsten Slok also warned that AI agents could automatically shift household cash from traditional bank accounts yielding roughly 0.1% into fintech accounts offering 3.3%-5.0%, pressuring banks' low-cost deposit bases, though no current bank run is evident.

Analysis

The relevant earnings risk is not deposit outflows alone but a permanently higher deposit beta: automated cash sweeping would force banks to reprice balances faster in each rate cycle, reducing the value of noninterest-bearing deposits and lowering through-cycle NII/ROTE multiples. This is most acute at retail-oriented institutions and sweep-dependent brokers, where cash balances are embedded in client economics; SCHW faces greater margin sensitivity than GS or MS, whose earnings are less reliant on low-cost consumer transaction deposits. BAC is exposed to both higher funding costs and a potentially weaker credit/loan-demand backdrop, making its current pressure more fundamental than the advisory-centric banks.

The near-term AI narrative is likely premature. Meaningful autonomous movement of household cash requires trusted permissions, account aggregation reliability, FDIC-insurance optimization, tax/reporting clarity, and consumers accepting liquidity friction; that is a 6-18 month product-adoption issue rather than a days-to-weeks deposit-run catalyst. The more immediate 1-3 month driver remains the yield curve: a further rise in long rates without an equivalent increase in short rates can initially support asset yields, but credit normalization and deposit repricing eventually dominate.

SOFI is not a clean beneficiary simply because it offers attractive savings yields: winning rate-sensitive deposits can raise acquisition costs and compress funding-margin advantage unless cross-sell and loan yields scale faster. Conversely, agent-driven rate shopping could favor scaled digital platforms with low servicing costs, but only after evidence of sustained deposit growth without elevated promotional expense. The contrarian opportunity is that market pricing may be treating all financials as deposit-beta casualties despite materially different funding models and cash monetization exposure.

Key Decisions for Investors

• Initiate a 3-6 month pair: long JPM / short SCHW, sized beta-neutral. JPM's diversified operating-deposit base and fee businesses should better absorb funding-cost normalization, while SCHW has more direct exposure to client cash sorting and sweep-balance margin pressure. Reassess if SCHW reports sequential bank-deposit growth and stable net interest revenue despite lower sweep balances.
• Avoid adding to BAC on the AI-deposit thesis alone; use the next earnings update as the decision point. A sequential increase in interest-bearing deposit costs, lower NII guidance, or rising criticized loans would validate downside; stable deposit beta and unchanged NII outlook would make the current selloff a cover signal rather than a fresh short.
• Keep SOFI on watch rather than buying the fintech-deposit narrative. Upgrade only if quarterly deposit growth remains strong while deposit costs and sales-and-marketing expense stay contained; otherwise high-yield deposit gathering is revenue-light and margin-dilutive. A long SOFI / short SCHW expression is appropriate only after that funding-cost data confirms digital share gains.
• Do not short GS or MS as a primary expression of automated deposit migration. Their earnings sensitivity is more likely to be driven by capital-markets activity and asset prices; use any continued financial-sector de-rating to rotate short exposure toward SCHW and retail-bank proxies rather than the advisory-heavy brokers.