Market Autopsy: how Silicon Valley Bank died in 44 hours
A bank with $200bn in assets failed in less than two days. The cause was not fraud, and it was not a bad loan book. It was a duration mismatch, a badly sequenced announcement, and a group chat.
- What happened
- On March 8, 2023, Silicon Valley Bank sold about $21bn of securities at a $1.8bn loss and announced a capital raise. On March 9, customers tried to withdraw $42bn. On the morning of March 10, regulators closed it.
- Why
- The bank had funded long-dated bonds with deposits that could leave instantly. Rising interest rates made those bonds worth less; the announcement made depositors aware of it at the same moment.
- The numbers that matter
- $21bnSecurities sold$1.8bnLoss realised$42bnWithdrawals in one day44 hrsAnnouncement to closure
- Why it matters
- Every element of this failure was visible in public filings beforehand. The lesson is about how information becomes action — and how fast that can now happen.
- What happens nextas we saw it on 15 Aug 2026
- Nothing. This is history. The value is in the sequence.
This series reconstructs failures in the order they happened, using what was known at each step. The temptation with any collapse is to narrate it as though the ending were obvious. It rarely was — and pretending otherwise teaches nothing.
Silicon Valley Bank was, until the week it closed, an unremarkable institution in the way that matters here: it was not committing fraud, it had not made reckless loans, and its regulators were not raising public alarms. It failed anyway, in under two days.
The setup
SVB banked technology start-ups and the venture funds that financed them. During the funding boom of 2020 and 2021, those clients raised enormous sums and deposited them. The bank's deposit base grew far faster than its lending opportunities.
So the bank did what banks do with excess deposits: it bought securities. Specifically, long-dated government-backed bonds — safe from a credit perspective, in the sense that they would be repaid in full at maturity.
A long-dated bond is exposed to interest-rate risk regardless of credit quality. When market interest rates rise, the market value of an existing bond paying a lower fixed rate falls. Hold it to maturity and you are repaid in full. Sell it before maturity and you realize the loss.
SVB had funded assets that must be held for years with deposits that could be withdrawn in seconds. That mismatch is the entire story.
Then interest rates rose at the fastest pace in four decades. The bank's bond portfolio fell in market value. At the same time, its depositors — start-ups no longer raising fresh capital — began spending their balances. The bank needed cash and its most liquid assets were worth less than it paid.
The sequence
The mistake that turned a problem into a failure
The securities sale on its own was survivable. Realizing a $1.8bn loss to raise liquidity is painful, not fatal, for a bank of that size.
The failure was in the sequencing. The bank told the market about the loss and the need for new capital in the same announcement, without the capital raise already being secured. It handed depositors a complete argument — this bank has losses it cannot absorb and needs money it does not have — with nothing attached to resolve it.
SVB's depositor base was unusually concentrated, unusually large per account, and unusually well connected. A significant share of deposits exceeded the FDIC insurance limit, meaning those depositors had a rational, immediate incentive to move first. And they were in the same networks, funds and chat groups.
A bank run used to require queuing outside a branch. This one required a phone. The speed of the outflow — $42bn in a day — is the genuinely new element, and it is an interpretation of the mechanism rather than a claim that any single group caused it.
What was knowable beforehand
This is the part where hindsight has to be handled carefully.
- Knowable: the composition of the securities portfolio, the unrealized losses on it, the deposit concentration and the uninsured deposit share. All disclosed in public filings.
- Knowable: that interest rates had risen sharply, and what that mechanically does to long-dated bond prices.
- Not knowable: that the bank would announce the sale and the capital raise together, in that order, on that day.
- Not knowable: that the withdrawal request would reach $42bn in a single session — a speed with no precedent to reason from.
The most common lesson drawn from SVB — 'the risks were obvious' — is the least useful one. The risks were disclosed; the trigger was not. What is genuinely transferable is narrower and more practical: an institution funding illiquid or long-dated assets with instantly withdrawable money is fragile regardless of the credit quality of those assets, and concentration in the funding base is itself a risk. Clearly labeled as our view.
Why it still matters
Duration mismatch is not a banking curiosity. It is the same structure that appears whenever a long-term commitment is funded with short-term money — in a business, a fund, or a household with a floating-rate obligation against a fixed income.
The question SVB leaves behind is not "was this bank badly run." It is: what do you own that is fine as long as you are not forced to sell it, and what would force you to sell it?
Sources
- Silicon Valley Bank: A Timeline and Summary of EventsLowenstein Sandler LLP · Mar 2023 · Secondary source
- Silicon Valley Bank is shut down by regulators in biggest bank failure since the financial crisisCNBC · Mar 10, 2023 · Secondary source
Related analysis
$10,000 a year ago: bitcoin, gold, and the part return charts never show
One of these turned $10,000 into $13,115. The other turned it into $5,310. The interesting question is not which — it's what you would have had to endure to still be holding.
Why small caps hit a record on the day the S&P 500 fell
On Friday, August 14, 2026, the Russell 2000 closed at an all-time high while the three big indices finished lower. The explanation is not that investors turned bullish on small companies.
The economy is sending two signals at once. The Fed is listening to the louder one.
Payrolls fell. Retail sales fell. Inflation held at 3.4%. And three Federal Reserve officials voted to raise interest rates. Here is how those facts fit together.