Banks do not blow up from lack of cleverness. They fail because cleverness scales faster than caution. If the best-capitalized bank chief in America says competitors are doing dumb things again, the paradox is simple: high IQ, low margin of safety. In markets, the line between innovation and fragility is thin, and it narrows under pressure. The rub is not whether a few lenders push too far. It is that competition itself can force even prudent firms toward the cliff’s edge. We do not need a new villain. We need to relearn how systems break.
Jamie Dimon’s remark lands because it echoes a pattern. When spreads compress and growth is scarce, lenders search for yield by easing terms, stretching tenor, or ignoring correlation. In game-theory terms, this is a prisoner’s dilemma with balance sheets. If one bank loosens, rivals must follow or lose share. Everyone optimizes for the next quarter. The system optimizes for disaster. Judgment gets outsourced to models. The models work until the assumptions fail, which is when they matter most.
History makes the point. The savings and loan crisis married rate mismatch to weak supervision and produced hundreds of failures. In 2008, the dumbness was standardized: warehouse lines, off-balance-sheet conduits, triple-A labels on mezzanine risk. None of it looked reckless in real time because the average deal was fine. The flaw was correlation under stress. When housing turned, the entire capital structure was the same bet in disguise. That is how smart people build fragile systems: piece by piece, by winning local games and losing the big one.
Today’s version lives in private credit and nonbank lending. Banks tell themselves they are safer because they have more capital, tighter stress tests, and less complex derivatives exposure. That is true at the individual level and incomplete at the system level. Risk has migrated, not evaporated. Direct lending, covenant-lite structures, payment-in-kind toggles, and generous EBITDA adjustments are now features, not bugs. When yield is the selling point, underwriting is a concession. The leverage hides in NAV facilities, in warehouse lines behind private funds, and in liquidity promises that rest on calm markets.
Regulation helps and hurts. Tighter bank rules reduce direct exposure but push activity into shadows where disclosure is weaker and incentives are misaligned. The Financial Times has warned about aggressive lending standards and the need for oversight that looks across charters, not only within them. The mechanics are engineering 101: you do not make a bridge safer by moving trucks to a road you do not monitor. You lower measured stress on the bridge and raise unmeasured stress on the network. When one span fails, traffic reroutes through the same weak joints. Markets mirror that math.
The rate regime shifted, and the easy money phase ended. Funding costs are higher and deposit betas are stickier. That compresses net interest margins and sharpens internal pressure to keep return on equity intact. The path of least resistance is to reach for duration, water down covenants, or quiet volatility with accounting. You can load earnings today by sacrificing resilience tomorrow. Dimon has long criticized short-term scorekeeping because it manufactures this choice: meet the next print or hold the line. The language of performance seduces firms into fragile posture.
We should stop pretending this is about rogue desks. This is structure and incentives. Quarterly earnings are a speed limit sign that traders read as a target. Leaders who know better get cornered by peer comparison. Meanwhile, investors chase what is working with textbook recency bias. Retail flows tilt toward whatever promises steady yield with low mark-to-market pain. The result is a shared narrative that mistakes smoothness for safety. In probability terms, the variance is hidden, not gone; the tails are fatter, the center looks calm. That is how tails ambush.
Liquidity is not a thing you own. It is a behavior others perform for you under benign conditions. We relearned this in the 2019 repo spike, the 2020 Treasury dysfunction, and the 2022 UK gilt crisis. Seemingly safe collateral becomes unstable if a few large players need cash at once. In credit, the mismatch is sharper. Funds promise periodic redemptions against assets that are hard to sell without price concessions. Gating policies are a tourniquet, not a cure. Banks, for their part, backstop these vehicles with drawer lines that count on normal exit windows. Normal is not a strategy.
In stress, correlation goes to one because funding is the common factor. Securities that share nothing but the need for dollar liquidity start to trade together. Spread products behave like equities. Basis trades that looked like free money develop gap risk. The mechanical analogy is resonance: a bridge tolerates heavy trucks, then falls to a light wind at the right frequency. The cause is not the last gust. It is the design that amplifies it. Tail events are design failures revealed by time.
Robust capital and liquidity rules are necessary. They are not sufficient. The industry’s new comfort blanket is the belief that capital ratios and stress tests immunize large banks. They improve survival odds but do not cure collective action problems. If bank-affiliated entities extend leverage to private credit funds that then lend to the same borrowers banks used to serve, the system has not de-risked. It has become harder to map. The Financial Post has noted how competitive behavior pushes institutions to defend share in ways that elevate system-wide exposures. This is coordination risk disguised as market share.
There is also model monoculture. When rating agencies, consultants, and managers converge on the same risk metrics, they hardwire common failure modes. Averages look strong until they all break together. Investors forget that credit cycles do not die of old age. They die when refinancing windows close. The roll-down math that flatters returns in quiet periods becomes an anvil when spreads gap and exit doors narrow. The lesson from Long-Term Capital Management was not to avoid leverage forever. It was to avoid leverage that depends on the kindness of others to exit.
Retail sentiment spikes after high-profile warnings because people still believe timing is a skill. The better framing is Bayesian. A credible insider increases the probability that underwriting standards are eroding at the margin. That does not mean collapse is imminent. It means the payoff distribution is shifting: more small gains for longer, then a sudden drawdown when funding, collateral, or covenants get tested. Watch the terms, not the tickers. When you see more PIK features, EBITDA add-backs, and looser reporting, you are watching the fuse, not the fire.
Game theory explains why this persists. In repeated games with short reporting intervals, defection dominates cooperation unless you can enforce patience. Boards and investors rarely reward patience until after a blowup. They discount cash flows but not risk culture. So firms adopt what looks like a rational strategy: arbitrage the measurement window, assume liquidity, and trust the exit. That is not evil. It is fragile. The remedy is dull and unpopular: redundancy, simpler structures, and the acceptance of underperformance in benign phases.
The system will not save itself by naming villains or hoping regulators police every shadow. The only reliable defense is margin of safety institutionalized as culture. That means writing covenants that bite, funding with terms that match assets, and stress testing for the scenario where counterparties do not show up. In nature, forests that ban all small fires guarantee a big one. Controlled burns are costly on purpose. Banking has an analogue: pass on hot deals, raise terms when others drop them, hold capital that looks lazy in good times. It feels wrong until it is the only thing that works.
Dimon’s warning should not be read as theater. It is a reminder that the game resets the same way each cycle. Competition squeezes standards. Accounting smooths the bumps. Unseen linkages grow. Then a catalyst arrives and correlation completes the work. The dumb thing is not that some rivals cut corners. The dumb thing is believing their choices will not become your problem when the system is asked to prove its resilience.
