Silicon Valley Bank collapse

A deposit flood looking for a home

For years, Silicon Valley Bank (SVB) was the main bank for the US start-up and venture-capital world, and that single fact shapes everything that follows. Its deposit base grew fast: from about $61.76 billion at the end of 2019 to about $189.20 billion at the end of 2021, as pandemic-era venture funding poured into its clients' accounts. Total assets roughly tripled over the same stretch, from about $71 billion to more than $211 billion. A bank holding that much new cash has to put it somewhere, and SVB put most of it into what looked like the safest asset available: long-dated US Treasuries and agency mortgage-backed securities.

Safe from default, exposed to rates

Those bonds were, in one sense, exactly as safe as advertised. They were backed by the US government, and every coupon was always going to be paid in full and on time. But a bond's safety from default is not the same as its safety from a change in interest rates. Starting in March 2022, the US Federal Reserve raised its policy rate from a range of 0-0.25% to 4.50-4.75% by February 2023, the fastest tightening cycle in decades. As rates rose, the market value of SVB's existing bonds, bought when rates were near zero, fell hard. By 31 December 2022, the bank's held-to-maturity securities had an amortised cost of $91,321 million but a fair value of only $76,169 million, an unrealised loss of $15,152 million. A further unrealised loss of about $2.5 billion sat in the available-for-sale book. Together, the two losses totalled roughly $15 to 17 billion on paper, close to the bank's entire $16.0 billion of equity.

A hedge removed at the worst moment

SVB had used interest-rate swaps to offset part of the interest-rate risk on its available-for-sale bonds. In the first half of 2022, it unwound about $11 billion notional of these swaps, booking about $517 million of gains from closing them out, and left only about $563 million of swap protection in place by year-end 2022. That decision removed the one instrument that could have cushioned the price fall as rates kept climbing through the rest of 2022. A hedge that has been closed out is a hedge no longer available, and the timing here could hardly have been worse.

A loss hidden by accounting, until it was not

None of this loss had to be reported as a hit to income or capital, because "held to maturity" accounting carries bonds at amortised cost, not market value, for as long as the bank does not sell them. The loss was real, but invisible, right up until SVB needed cash. On 8 March 2023, SVB announced in a single release that it had sold about $21 billion of available-for-sale securities at a realised after-tax loss of about $1.8 billion, and that it was launching a $2.25 billion capital raise to rebuild its balance sheet. Management intended this as prudent housekeeping. Investors and depositors read it as a confession that the bank was in trouble.

The run

What followed was one of the fastest bank runs in history. On 9 March, depositors, many of them venture-capital-backed start-ups warned by their investors to move their cash, tried to withdraw about $42 billion in a single day, close to a quarter of the bank's deposit base. SVB ended the day with a negative cash balance of about $958 million. The planned capital raise collapsed, since no investor wants to put fresh capital into a bank haemorrhaging deposits. On 10 March 2023, the California Department of Financial Protection and Innovation closed Silicon Valley Bank and handed it to the FDIC as receiver. The failure is often described as the second-largest bank failure in US history, though some counts place it third if Continental Illinois (1984) is included; both framings appear in the record, and nothing in the case turns on which one is used. Two days later, on 12 March, the US Treasury, Federal Reserve and FDIC jointly invoked a "systemic risk exception," guaranteeing all deposits, insured and uninsured, to stop the run spreading to other banks.

The underlying story: the price of time

SVB's failure is not a story about a bad trade or a rogue trader. It is a story about the price of time. A fixed-rate bond bought when rates were near zero is worth less once rates rise, even though nothing about the bond's promise has changed. In duration terms, SVB's held-to-maturity book had a weighted-average duration of about 6.2 years, funded by deposits with an effective duration close to zero. That mismatch, between long-duration assets and short-duration, on-demand liabilities, sat quietly on the balance sheet for two years. It only became a crisis once the bank needed cash and the market forced the loss into the open. Most of SVB's depositors were start-ups holding balances well above the $250,000 insurance limit, and they were unusually concentrated: many banked with SVB because their venture-capital investors told them to, and many of those investors talked to each other. Once doubt appeared, the rational move for each depositor was to withdraw immediately, and because everyone reasoned the same way at once, the doubt became self-fulfilling within hours. That is the same logic behind a margin call in a derivatives market: a paper loss becomes a demand for cash, and if the cash cannot be found fast enough, a position that is solvent on paper becomes an actual default.

This case is not a derivatives trade gone wrong. It is the price of time, the curve built in Lecture 4, and what happens when a balance sheet ignores it. SVB was long a large book of fixed-rate, long-duration bonds and short (funded by) deposits that could leave on demand. That is a duration mismatch, the same mismatch a duration hedge in Lecture 4 is built to close.

Two separate risks matter here: default risk and interest-rate risk. The bonds were federal government or federally backed securities, so default risk was close to zero. But price is not the same as safety from default. When the risk-free curve moves up, the present value of a fixed set of future coupons falls, because those cash flows are now discounted at a higher rate. That fall is real even though nothing about the promise to pay has changed. This is interest-rate risk, and duration is its first-order measure.

The accounting label "held to maturity" let the paper loss stay off the income statement and out of regulatory capital ratios, because HTM securities are carried at amortised cost, not fair value, as long as the bank does not sell them. That is not a derivatives concept, but it explains why the loss was invisible until a forced sale, needed to raise cash for deposit outflows, converted an unrealised, non-cash loss into a realised, cash one. Once realised and disclosed, it triggered exactly the liquidity event that made further realisation certain.