Apollo's Torsten Slok Warns Agentic AI Could Trigger Bank Run, Bolstering Bitcoin (BTC)

Apollo economist Torsten Slok warns agentic AI could drain cheap deposits and spark a bank run, boosting the case for Bitcoin (BTC) and stablecoins.

(04:25 PM UTC)
5 min read
AI SummaryAI
  • Apollo's Torsten Slok warns agentic AI could spark a bank run by sweeping deposits from 0.1% checking accounts.
  • Fintech accounts from Revolut, SoFi and Wealthfront pay 3.3% to 5% on deposits, per Slok's memo.
  • A $10,000 balance earns about $100 annually at 5% versus $10 in a typical checking account.
  • Meta's Muse agent launched September 8 with Plaid-linked accounts but cannot yet initiate transfers.
v3xn8bwc

The 0.1% Checking Problem

Slok's warning lands at the intersection of two of finance's fastest-moving stories: agentic artificial intelligence and the migration of household cash. Torsten Slok, chief economist at Apollo Global Management and one of the most widely read economists on Wall Street, argued in a Sunday memo titled “Is an Agentic bank run coming?” that AI assistants managing household finances will soon sweep idle cash out of checking accounts — which pay a national average of just 0.1% — into high-yield alternatives automatically, with no human in the loop. He points to fintech deposit products from Revolut, SoFi, Varo, LendingClub and Wealthfront, the category of high-yield cash accounts brokers like Robinhood helped popularize, which pay between 3.3% and 5% a year. On a $10,000 balance, that gap is roughly $100 of annual interest versus $10 in a standard checking account. Apollo Global, where Slok is a partner, manages about $1 trillion in assets, which gives the warning unusual reach. If households broadly deploy AI agents to optimize cash returns, Slok writes, banks could lose “a large share of the cheap deposits” they rely on to fund loans — a problem for the entire financial system. Agentic finance, as the concept is known, describes AI that acts rather than merely answers: agents monitor balances in real time, compare yields across institutions, move idle funds into higher-paying accounts and shift the money back before bills fall due. Market-size estimates vary widely: Mordor Intelligence values agentic AI in financial services at $7.78 billion in 2026, projecting $43.52 billion by 2031. The payment layer is already being built — Coinbase's x402 protocol, the most widely used agentic payment standard, lets an AI agent pay for online services in stablecoins within seconds, with no account, card or human approval, and has reportedly processed roughly 188 million to over 205 million cumulative transactions across around 69,000 active agents.

Meta Muse and the Yield Gap

The memo's yield-gap evidence is granular: drawing on Federal Deposit Insurance Corporation (FDIC) and Haver Analytics data, Slok lists 11 fintech and neobank deposit accounts paying 3.3% to 5%, led by Adelfi at 5% and SoFi at 4.5%. The trigger he singles out is Meta's personal agent Muse, launched on September 8 with account connections built through Plaid. Muse can currently display a user's balances, transactions and investment holdings, but cannot yet initiate transfers — a capability that would reshape the deposit landscape the day it ships. Muse arrived in the same week that Jakob Coxon, a researcher who worked at both OpenAI and Anthropic, left the field and accused the two companies of “gambling with our lives,” while Meta's leadership declined calls to slow the sector's overall pace. Banks keep paying near zero on checking balances because those deposits are their cheapest source of loan funding, and retail banking assumes most of the money stays put. Regulators have flagged synchronized machine behavior before: the Bank of England's April 2025 Financial Stability in Focus paper warned that advanced AI trading strategies could drive firms into increasingly interconnected positions that act alike during crises, amplifying shocks, and defined agentic AI as systems that pursue specific goals independently using tools while adapting to their environment. Timing sharpens the concern — the Federal Reserve lifted its target range to 3.75%-4% on September 16, its first hike in three years, and bank stocks slid afterward on investor worries that a compressed spread between short- and long-term rates would cut into lending profits. The herding idea dates to a 2023 speech in which SEC Chair Gary Gensler warned that a few dominant AI models could reach “similar decisions because they receive the same signal.” At the United Nations Security Council on September 23, Anthropic's Dario Amodei pledged to slow development by as much as necessary, and OpenAI's Sam Altman urged extreme caution. Readers tracking the market in real time can follow live spot and futures prices on Bitget.

Stablecoin Rails, Bitcoin (BTC) in Focus

For COINOTAG, both items trace a single arc: software that moves money is colliding with a banking model built on sticky, near-free deposits. The load-bearing primary document is Slok's memo itself, and it states plainly that household-scale yield optimization threatens the deposit base that funds credit. If deposits become contestable at scale, the case for bank-independent digital dollars strengthens — the circulating supply of stablecoins gives agents somewhere to sweep cash without a bank's permission. Standardization is moving fast: the x402 Foundation, joined by Cloudflare, Google, Visa, Mastercard, AWS, Circle and Stripe and now governed by the Linux Foundation, is building those rails on networks such as Polygon and the NEAR Protocol network. Nate Geraci, co-founder of the ETF Institute, has previously said AI and crypto are both coming for the traditional banking model. For Bitcoin (BTC), trading near $84,000 at the time of writing, the warning reinforces its hedge narrative against legacy-banking fragility.

COINOTAG News Desk

COINOTAG News Desk

COINOTAG's editorial and research desk.

How our News Desk works
AI-Assisted

AI-generated, AI-reviewed, under COINOTAG editorial oversight.