Forty Eight Hours: How Silicon Valley Bank Died

The Fastest Failure

On Wednesday March 8 2023 Silicon Valley Bank was a $209 billion institution banker to something like half of the venture capital world with a stock price that had enriched shareholders over the long term.the series is about how that happened because the mechanics are an ongoing lesson in how banks really fail not a one-time fluke that you can file away and forget

A Balance Sheet Built for One Interest Rate

SVB's problem wasn't crypto. It wasn't fraud. It wasn't bad loans. It was duration. During the boom of 2020 and 2021 covered earlier in this series venture capital-backed startups raised historic amounts of cash and deposited them into SVB and the bank's deposits roughly tripled in two years. That money had to go somewhere so SVB bought Treasury and long-term mortgage-backed securities with very littleperformance because in 2021 almost everything yielded very little.On paper it looked conservative.Government bonds.Secure guarantee.The kind of portfolio that a regulator would accept without looking twice

Then came 2022 and the fastest rate hike cycle since Volcker. When interest rates rise existing bond prices fall and long-term bonds fall more sharply. That's duration risk the sensitivity of a bond's price to a change in interest rates and grows the further the cash flows move away from the bond. By the end of 2022 SVB was accumulating more than $15 billion in unrealized losses on securities it had classified as held to maturity an accounting cube that allows a bank to hold bonds at cost rather than market value on the theory that it will never need to sell them. The losses were disclosed in footnotes. They were roughly equal to the bank's total equity capital. Visible to anyone who read the documents. Almost no one did until it mattered

A Worked Example: What Duration Risk Does to a Balance Sheet

Duration risk sounds abstract until you look at the numbers so let's break them down using a hypothetical bank instead of SVB's actual figures so you can check each step yourself. Call it Riverbend Community Bank completely made up

Suppose Riverbend owns $1 billion or $1 billion of face value in Treasury bonds with an average duration of 6 years. The duration of 6 is a rule of thumb that says: for every 1 percentage point increase in interest rates the market value of that bond falls about 6 percent. It's an approximation not exact physics but it gets you close enough to reason

Now let's assume rates rise 3 percentage points over the next year in the spirit of what actually happened in 2022 although these are round numbers not the actual path. An increase of 3 points versus a duration of 6 implies a price drop of about 3 times 6 which equals 18 percent

Apply that 18 percent to the $1 billion book: 1,000 times 0.18 equals $180 million of paper loss

Now let's assume that Riverbend has $100 million in equity or just 10 percent of that bond book even before counting the rest of its balance sheet. An unrealized loss of $180 million is equal to 1.8 times the bank's total capital. According to held-to-maturity accounting none of that loss affects the income statement. The bank is still declared solvent on paper. But if depositors get nervous and the bank has tosell those bonds to fund withdrawals the paper loss becomes real and $100 million of capital cannot absorb a $180 million hole. This is the same arithmetic that was quietly inside the SVB's actual $15 billion figure just reduced to numbers small enough to check on a napkin

The Two Day Run

The trigger came on March 8 when SVB announced that it had sold $21 billion in securities at a $1.8 billion loss and would raise new capital to plug the hole. The announcement was intended to be reassuring. It landed in the opposite direction. Instead of closing the story it turned a footnote into a headline and venture capitalists a closely connected group chat of an industry if ever there was one passed on March 9advising his portfolio companies to withdraw their money immediately

Here's the detail that makes SVB historic. Approximately 94 percent of its deposits were above the FDIC insurance limit of $250,000 because its clients were businesses that had payrolls not households that had savings. Deposit insurance It exists precisely to make the operation of a bank irrational and it simply did not apply to people banking at the SVB. On Thursday March 9 customers attempted to withdraw $42 billion about a quarter of the bank in a single day mostly through apps and wires rather than waiting in line on a sidewalk. Regulators expected more than $100 billion more to come out by Friday morning. There was nothing left to run on. The FDIC statedbankrupt to the bank before lunch

The Weekend the System Blinked

The ruling immediately raised a system-wide question: If SVB's uninsured depositors suffer losses what company keeps the money in any mid-sized bank on Monday morning? Signature Bank was closed that Sunday. That afternoon regulators invoked the systemic risk exception and guaranteed all deposits at both banks insured or uninsured while the Federal Reserve created a new mechanism that allowed banks to borrow against their underwater bonds at their face value rather than their market value. Shareholders and bondholders of the failed banks were wiped out. Depositors were recovered

The stress didn't end there. First Republic Bank facing the same outflow of uninsured deposits limped along for weeks before being seized and sold to JPMorgan in May 2023 an even bigger failure than SVB. Three of the four largest bank failures in U.S. history occurred in eight weeks. And yet surprisingly the panic stopped there. The backup worked

Case Study: Continental Illinois and the Original Too Big to Fail

SVB was not the first time a bank died because its funding base had no reason to remain loyal. In 1984 Continental Illinois National Bank was the seventh largest bank in the United States a Chicago wholesale lender that was funded not by a broad base of local retail depositors but by large uninsured deposits from other banks foreign institutions and money market funds seeking a slightly better rate. That funding structure is the direct ancestor of the SVB problem decades beforeSmartphones existed to speed it up

When rumors spread in 1984 that Continental had a pile of bad energy loans largely impaired loans for Oklahoma oil and gas that it had bought from a smaller failing bank wholesale depositors didn't stick around to find out if the rumors were true. Unsecured money unlike a retail savings account has no particular reason to be loyal and it was gone in a matter of days moving via telex and phone rather than an app but leaving anyway

The FDIC's response set the template that regulators used again in March 2023. Rather than let the bank openly fail and risk a chain reaction across all the banks and money funds that had lent it money the FDIC guaranteed all of Continental's deposits insured or uninsured and injected capital directly effectively taking a stake in a bank it considered too large and too interconnected to allow to collapse in the usual way. The phrase "too big to fail"go bankrupt" entered the financial vocabulary largely because of this bailout

Continental Illinois and SVB are almost forty years apart and have completely different assets bad energy loans against Treasuries and duration but the underlying vulnerability is the same form: a bank largely funded by depositors with no insurance or loyalty who can and will leave in a hurry the moment doubt appears. The lesson did not need to be reinvented in 2023. It simply went from a telex machine to a push notification

Where This Model Breaks

The SVB story is clean almost to the point of being a parable and clean stories deserve some suspicion. This is where it gets less tidy

First unrealized losses on held-to-maturity books are common and rarely fatal. Many regional and community banks endured similar losses on paper through 2022 and 2023 without anyone taking advantage because their deposit bases were retail insured and boring. Duration risk only becomes fatal when it collides with a deposit base that can and will move overnight. Eliminate the concentrated unsecured and tightly held venture capital client baseinterconnected SVB's stock portfolio was a drag on the next decade of profits not a forty-eight-hour collapse

Second one could argue that this was not so much a solvency problem as a liquidity panic that became reality. Held to maturity most of those bonds would have actually paid par eventually. The loss only became real because SVB was forced to sell in a falling market to fund withdrawals that it otherwise could not cover. A bank with a stronger funding base or one that had created a larger liquidity cushion or permanent line of credit before the run began could have overcome thesame duration mismatch without ever having seen a receivership. I find it really uncomfortable to sit with this because it means that the line between a solid bank having a bad quarter and a failed bank was to some extent a question of who panicked first and how quickly the money could move. That's not a totally satisfying story if you want clear cause and effect

Third the regulatory response itself changed the rules of the model going forward. The Fed's new lending facility allows banks to borrow against underwater bonds at their face value rather than their liquidation value which exists specifically to prevent the next SVB from having to realize a duration loss under pressure. If that facility or something like it remains in place and banks actually use it the exact mechanism that wiped out the SVB in forty-eight hours may not be available to wipe out the next Fed bank.same way. A model built on a very rapid episode always runs the risk of explaining the last war instead of the next one

How I Actually Read a Bank's Balance Sheet

My read is that most of what matters here was revealed all along and the discipline that would have caught it is almost boring: reading the securities footnotes. When I look at a bank now for a stock presentation or just out of habit I go through a short checklist that this episode taught me directly

I start with the split between held-to-maturity and available-for-sale and the unrealized loss disclosed on each of them. A held-to-maturity loss that is small relative to capital is a rounding error. A held-to-maturity loss that approaches or exceeds the bank's equity capital as the SVB did is close to the scariest figure a bank can publish and typically appears quietly in a footnote rather than the income statement exactlywhere I would have been lost if this episode hadn't happened first

Then I look at the deposit base itself. What fraction is insured versus uninsured? How concentrated is it by industry or by a small number of large depositors? A retail bank with thousands of small insured checking accounts is a structurally different animal from a bank whose ten largest depositors could in theory coordinate a withdrawal via a group chat. The $250,000 insurance cap is not a footnote detail: it decides who has a rational reason to apply and whono

Finally I try to ask what the actual trigger would be. Every fragile balance sheet needs a spark. For the SVB this was a capital increase announcement that was meant to appear reassuring and rather be interpreted as a confession. I don't think you can predict the exact spark in advance but you can identify which banks are in a room full of firewood. The way I would actually use this is not to short all the banks that have unrealized bond losses - many of them are actually fine - but as a screen: first find theduration mismatch then check to see if the deposit base gives you something to turn on. This is an educational lens for reading a presentation not a trading signal and I'm not telling anyone to buy or sell bank stocks based on three paragraphs from a news site

What It Should Teach You

Some lessons worth following beyond this episode. Banks die of liabilities not assets. SVB bonds were worth money if held to maturity it was the deposits that could not wait so you have to be careful with the concentration of financing before becoming obsessed with the quality of the loans. Accounting categories do not change the economy. The maturity treatment hid the losses from the income statement not from reality itself. And the tail risks are concentrated around those who grewmost rapidly during the previous regime. The SVB tripled in size during the free money era making it the most exposed institution when that era ended. The fastest growing in any boom is usually the first name worth checking when the regime changes

The Bottom Line

Silicon Valley Bank failed because it financed long-term low-yielding bonds with the banking industry's most volatile deposit base and because the 2022 rate hikes made that mismatch fatal. The run took two days the bailout took a weekend and the lessons - about duration uninsured deposits and panic over the speed of smartphones - were captured directly in the way regulators and analysts look at each bank since. Continental Illinois shows thesame financing vulnerability that manifested itself four decades earlier at the pace of a telex machine and the counterargument is real: many banks endure long-term losses without dying and the difference often comes down to who panics first. Of all the episodes in this series this is the one where literally reading the footnotes would have allowed you to see it coming