Will AI-driven cash transfers disrupt banks and brokerages? Morgan Stanley sees less risk than markets fear
Large amounts of low-interest cash have already moved into higher-yielding products, while remaining balances also serve payment and trading needs. Morgan Stanley believes AI is more likely to accelerate deposit repricing than substantially raise the ultimate cost; brokerages can try shifting to fees, but customer acceptance and competition determine how much revenue they can replace.
Original source：Morgan StanleyFirst published on Unhidea：Original source date：
Key takeaway
Morgan Stanley believes transactional cash needs and account-access restrictions limit AI’s near-term impact on low-cost funding; the longer-term effect depends on deposit-rate pass-through and brokerages’ transition to fees.
Will AI move all low-interest cash away?
Morgan Stanley believes the impact of agentic AI on banks’ deposit costs and brokerages’ cash sweep revenue will be smaller than markets fear. The concern follows this chain: AI automatically seeks higher-yielding alternatives, low-interest cash becomes more rate-sensitive, banks are forced to raise rates to retain customers, and brokerages lose cash sweep revenue.
But the migration of funds is already well advanced. The share of non-interest-bearing deposits at US banks fell from about 30% in 2021 to about 21%. It has been broadly stable since 2024 and has returned to its long-term average excluding years of near-zero interest rates. As a share of combined bank deposits and money market funds, money market funds account for about 28%, above the historical average of about 24%.
Remaining low-interest balances have service value: companies hold deposits in exchange for cash management and transaction services, while individuals keep a buffer to pay bills and avoid overdrafts. Morgan Stanley therefore judges that the risk of additional outflows is concentrated mainly in idle cash, rather than all low-cost funding.
Will deposit costs rise faster, or ultimately rise further?
Morgan Stanley distinguishes between the speed of repricing and its ultimate magnitude. Deposit beta measures how much of a rate increase is passed through to deposit rates. In the tightening cycle that began in 2022, the terminal beta for interest-bearing deposits was about 52%, above 39% in the 2004 cycle and 36% in the 2015 cycle. But that tightening cycle was also more rapid: rates rose by a cumulative 525 basis points over seven quarters.
Even after digital channels became more convenient, deposit rates in the 2022 cycle peaked four quarters after the final rate increase, the same lag as in the 2004 cycle. The bank believes AI could shorten this lag, causing funding costs to rise sooner, but would not significantly increase the terminal beta.
The historical basis for this assessment is that, from 1992 through the second quarter of 2026, the relationship between interest-bearing deposit rates and the federal funds rate still corresponded to a pass-through ratio of about 52%, with the latest actual reading slightly below the trend line. The condition that would truly change banks’ economics is AI persistently increasing the pass-through ratio, rather than merely speeding up adjustments.
Can fees make up for a decline in brokerages’ cash revenue?
When clients move cash into products such as money market funds on the same brokerage platform, cash sweep revenue falls, but client assets and relationships may be retained. Brokerages can therefore offer higher cash yields while trying to charge account, platform or cash management fees.
Under Morgan Stanley’s 2027 model, cash-related revenue covered for LPL Financial, Charles Schwab, Raymond James and Ameriprise amounts to 15–59 dollars per account per month, or about 5–13 basis points of client assets. The cash revenue included differs across companies; these figures estimate the fees needed to replace that revenue.
If recurring fees can replace interest-rate-sensitive revenue, earnings visibility and valuation multiples have room to improve. But competition and clients’ sensitivity to explicit fees will limit the extent of compensation, and pricing power differs across retail and adviser channels.
What limits the near-term impact, and what could change the long-term assessment?
Automatically finding rates and automatically transferring funds are separated by account opening, authorization, identity verification and the allocation of responsibility. Several major US banks have already restricted third-party AI access; Charles Schwab also blocks unauthorized agents and limits fund transfers through approved connections.
Brokerage clients also need cash for trading, withdrawals, fees and collateral, while T+1 settlement maintains the need for immediate liquidity. Longer-term pressure may come from tokenization and faster settlement: if funds can move across accounts in near real time, demand for cash buffers may fall, but this requires broad changes to financial infrastructure.
