Too big to fail is a statement about topology, not size. An institution earns the label when its liabilities are woven so deeply into other balance sheets — as counterparty, funding source, or market-maker — that its failure would transmit losses and uncertainty everywhere at once. The authorities’ problem is that everyone can see this, including the institution itself and everyone who lends to it.
The status functions as an invisible subsidy. Creditors of a protected institution lend more cheaply because they price in the rescue, and the cheap funding lets the institution grow larger and more entangled — the protection amplifies the condition that justified it. This is moral hazard operating at the level of market structure rather than individual behaviour.
LTCM in 1998 proved the label extends beyond banks: a hedge fund’s derivative book made it systemically unignorable. In 2008, the doctrine was tested destructively — Lehman’s failure demonstrated what the system does when a load-bearing node is removed, and the bailouts that followed hardened the implicit guarantee the crisis had been meant to discipline. Credit Suisse in 2023 showed the endgame under the post-2008 regime: a globally systemic bank was resolved over a weekend into a merger, because even with resolution plans on paper, no authority would run the experiment of an uncontrolled failure.
The post-crisis machinery — capital surcharges, living wills, bail-in debt — aims to make the label survivable rather than abolish it. The atlas records the running score: so far, when the moment arrives, the merger or the rescue has always looked cheaper than the test.