The 2023 Banking Crisis Wasn't Contagion. It Was Three Bad Banks and a 48-Hour Fire Drill.

On March 9, 2023, Silicon Valley Bank's customers asked for their money back. Not slowly, not in a line outside a branch — by app, by wire, all at once. About $42 billion in withdrawal requests landed in a single day, roughly a quarter of the bank's deposits, with another $100 billion queued for the next morning.² By March 10 the bank that had financed half of America's venture-backed startups no longer existed; the California regulator seized it and handed the keys to the FDIC.¹ The official narrative formed within hours: contagion. A run that could jump from bank to bank until the whole system buckled. That fear was real. It was also mostly wrong.

The story that hardened over the following weeks was that 2023 was a systemic crisis — banks toppling like dominoes, the financial system itself in question. Strip the panic away and you find something narrower and more damning: a handful of badly run, dangerously concentrated banks, the kind regulators had been writing memos about for years, failing in quick succession. The dominoes didn't fall because they were connected. They fell because each was already standing on a tilted floor.

What actually broke inside Silicon Valley Bank

SVB's failure was not exotic. It was the oldest bank trick gone wrong: borrow short, lend long. Flooded with startup cash during the 2021 boom, it parked the money in long-dated bonds. Then rates rose, those bonds lost market value, and the bank sat on a growing pile of unrealized losses — its 2022 filing already carried an accumulated other comprehensive loss of nearly $1.9 billion against $16.3 billion of equity, a quiet hole in the balance sheet.⁷ On its own, a paper loss is survivable; you hold the bonds to maturity and wait. The problem was the other side of the ledger. SVB's depositors were a tight cluster of venture firms and their portfolio companies — sophisticated, networked, and overwhelmingly above the insured limit. When one of them moved, they all heard about it within the hour.

That is the mechanism, and it is the whole story. A concentrated, uninsured deposit base is not a stable funding source — it is a single customer wearing a thousand name tags. The unrealized bond losses meant the bank could only meet a run by selling assets at a loss and crystallizing the hole. The depositors knew it. So the run wasn't irrational fear; it was a correct read of a fragile structure, executed at the speed of a group chat. The $42 billion didn't appear out of panic. It appeared because the floor was already tilted, and everyone standing on it could see the angle.

Signature wasn't a crypto casualty. It was a management one.

Two days after SVB, New York regulators closed Signature Bank — $110.4 billion in assets, $88.6 billion in deposits.⁴ The convenient label was 'the crypto bank,' as if the digital-asset sector had reached out and pulled it under. Read the regulators' own autopsy and that explanation collapses. The FDIC's April 2023 report on Signature is blunt: the proximate cause was illiquidity, triggered by panic spilling off Silvergate's self-liquidation and SVB's failure, but the root cause was poor management — a board chasing rapid, unrestrained growth without the risk controls to match, and a management team that did not always heed its examiners.⁵ Both banks had grown at a pace that should have set off alarms: SVB and Signature expanded their assets 198% and 134% respectively over 2019 to 2021, far outrunning their peers.⁶

Notice the pattern repeating. Same flaw, different label. Signature, too, ran an overwhelmingly uninsured deposit base — the same flammable funding structure as SVB — so when the spark arrived from next door, it caught. Contagion was the match. Management was the kindling. Blaming crypto for Signature is like blaming the breeze for the fire while ignoring that someone soaked the building in gasoline.

The 48-hour backstop that worked too well

By Sunday, March 12, the authorities had decided the risk of waiting was greater than the risk of acting. Treasury Secretary Yellen approved a systemic risk exception, authorizing the FDIC to guarantee every deposit at both banks — insured and uninsured alike. The same day, the Federal Reserve stood up a new emergency facility, the Bank Term Funding Program, backstopped by up to $25 billion from the Treasury's Exchange Stabilization Fund.⁶ The FDIC moved all of SVB's deposits into a bridge bank and drew a hard line on who would not be saved: shareholders and certain unsecured creditors were wiped out.³ It was fast, it was total, and it stopped the panic. As a piece of crisis containment, it worked.

That sentence is technically true and quietly incomplete. Losses to the Deposit Insurance Fund — estimated at roughly $20 billion for SVB and $2.4 billion for Signature¹⁴ — were to be clawed back through a special assessment on other banks, not the federal budget. But the BTFP carried a $25 billion contingent taxpayer exposure via the Treasury backstop, and the speed of the guarantee delivered a lesson that no press release could undo: when a big enough bank fails badly enough, every depositor gets made whole, insurance limit be damned. The intervention that calmed the markets also taught uninsured depositors and bank managers alike that the limit is a suggestion. That is moral hazard, manufactured in a weekend.

Doesn't the speed of the run prove it really was systemic?

The fair objection is that the run jumped from Silvergate to SVB to Signature in days, and a guarantee of only insured deposits might genuinely have triggered a wider panic — so calling this 'idiosyncratic' is too tidy. There's truth in it. Once SVB fell, every depositor in every regional bank with a similar profile started doing the same math, and that reflex is the very definition of contagion fear. The honest counter is that the contagion found the banks that deserved it. It did not jump to well-capitalized, diversified institutions with stable deposits; it traveled along the network of banks that shared SVB's exact structural flaw — runaway growth, concentrated uninsured funding, weak controls. And the academic post-mortem refuses to let the systemic framing inflate: SVB, Signature, and First Republic together produced three of the four largest bank failures in US history within two months, which sounds apocalyptic until you notice the same review correctly demotes SVB from 'second-largest' to third once First Republic surpassed it — a reminder that the headline framing was always a beat behind the facts.⁸ The panic was systemic in feeling. The failures were specific in cause.

The 2023 banking crisis is remembered as the weekend the system nearly broke. It is better understood as the weekend a few badly built banks finally collapsed, and the authorities chose a total rescue over a measured one because the measured one couldn't be tested in 48 hours. The backstop did its job: it stopped the bleeding. But it also blurred the only fact that mattered — that the rot was narrow, named, and known. The regulators had the memos. The bonds were already underwater. The depositors could read a balance sheet. What 2023 proved was not that contagion is inevitable. It was that when you save everyone equally, you erase the difference between the banks that ran a tight ship and the ones that didn't — and the next set of managers is watching.