Why Silicon Valley Bank Failed: The 36-Hour Collapse Explained

Silicon Valley Bank failed because it invested a sudden flood of deposits in long-dated, fixed-rate bonds at the bottom of the rate cycle, and those bonds lost about $15 billion of value when the Federal Reserve raised rates in 2022. That loss was roughly equal to the bank's entire equity. When SVB announced a loss-making bond sale on 8 March 2023, its uninsured, tightly networked depositors pulled about $42 billion in a single day, and California regulators closed the bank on the morning of 10 March.

SVB's loans were mostly sound. What failed was a promise that money placed with the bank could be had back, in full, whenever its owners asked. Forty years of careful lending to start-ups were undone in about 36 hours, and the failure became the second-largest in US history at the time.

What was Silicon Valley Bank?

SVB was founded in 1983 by Bill Biggerstaff and Robert Medearis, with Roger Smith as its first chief executive, to serve technology companies that big banks ignored. Its insight was about relationships, not technology. A start-up that had just raised a venture round had cash but no earnings and no collateral. A bank that knew the venture investors could take the company's deposits on day one, lend against the next funding round, and take small equity warrants as a sweetener.

The venture firms steered their portfolio companies to the bank, and the bank also handled the funds' own banking. SVB said it served nearly half of US venture-backed technology and life-science companies (a company claim, as reported). Its credit judgment was good: losses on venture debt and fund lending stayed low through 2022. In the terms of the book The Money Equation, SVB was strong on the third factor, distribution. It had the best network in its niche. Its company profile traces that history in full.

How did a deposit boom become a problem?

Between the end of 2019 and the end of 2021, venture money poured into start-ups and from there into their SVB accounts. Deposits roughly tripled.

Start-ups do not borrow much, so SVB lent out only about a third of its deposits. Most US banks lend 70–90%. The rest had to go somewhere, and in 2021 SVB bought about $80 billion or more of mortgage-backed securities and Treasuries yielding around 1.5–2%. It classified most of them as held to maturity (HTM). Under that accounting label, falls in market value do not reduce reported capital unless the bonds are sold.

The trade looked sensible in 2021. Short-term rates were near zero, and long bonds paid more. But a bond portfolio with a long duration loses a lot of value when rates rise. Between March 2022 and July 2023 the Fed raised its policy rate from 0–0.25% to 5.25–5.50%. By the end of 2022, SVB's HTM bonds were worth about $15 billion less than it had paid, close to its total shareholders' equity of about $16 billion. On paper, SVB was well capitalized. Economically, its capital was nearly gone if it ever had to sell.

Why were SVB's deposits so dangerous?

A bank can live with paper losses on bonds if its depositors stay, because it can hold the bonds until they mature. SVB's depositors were the least likely of any large US bank to stay.

About 94% of SVB's deposits were above the $250,000 FDIC insurance limit at the end of 2022 (as reported). Uninsured depositors have every reason to leave at the first sign of trouble: if a bank's assets are worth less than its deposits, the first to withdraw are paid in full and the last may not be. The depositors were also concentrated in one industry and connected through a few dozen venture firms and their group chats. The network that made SVB dominant was the same network that could empty it.

The deposits were already draining in 2022. Venture funding fell, start-ups burned cash, and clients demanded interest on what remained. SVB faced three pressures at once: shrinking deposits, rising deposit costs, and a bond book yielding far less than new bonds. The book's study of balance sheets and asset-liability management explains why this mismatch is the oldest trap in banking.

What went wrong inside the bank?

The Federal Reserve's review of 28 April 2023 blamed management, the board, supervisors and the 2019 rules that had lightened requirements for banks of SVB's size. Several decisions stand out.

• Hedges removed. SVB let most of its interest-rate hedges on its available-for-sale bonds run off or terminated them in 2022 to lock in gains (as reported).
• Models changed instead of risk. The bank's own models showed rising rates would hurt the economic value of equity. In 2022 it reportedly changed its assumptions about how long deposits would stay, rather than reduce the risk (Fed review).
• No chief risk officer. The post was vacant for about eight months in 2022, from roughly April until a successor was named in January 2023 (as reported).
• Pay tied to earnings. Executive pay rewarded earnings and return on equity, not risk measures, which favored the yield-chasing bond purchases of 2021 (Fed review).
• Open findings. By the time of failure the bank had 31 open supervisory findings, according to the Fed review, yet supervisors had not forced it to fix interest-rate risk quickly.

SVB also benefited from regulatory tailoring. The 2018 Economic Growth, Regulatory Relief and Consumer Protection Act raised the threshold for enhanced prudential standards to $250 billion, and the Fed's 2019 rules placed SVB in a lighter category that let it exclude unrealized losses on available-for-sale bonds from regulatory capital.

What happened in the 36 hours?

With a Moody's downgrade reportedly in preparation, management decided to sell its available-for-sale portfolio and raise new equity to reposition the balance sheet.

The announcement was meant to repair the balance sheet. Instead it told depositors that the losses were real and that capital was short. Prominent venture funds told portfolio companies to move their money (as reported), and the messages spread over Twitter, Slack and WhatsApp. Northern Rock had lost about 8% of its retail deposits over three business days in 2007, and Washington Mutual about $16.7 billion over eight business days in 2008. SVB lost a quarter of its deposits in one day by smartphone. Supervisors began to speak of "digital runs."

SVB had also not tested its emergency borrowing lines at scale and was not ready to use the Fed's discount window on 9 and 10 March, according to the Fed review.

Could SVB have survived?

The book's stylized example in Chapter 1 shows why timing mattered. A $200 billion bank with $90 billion of securities, $182 billion of mostly uninsured deposits and $13 billion of equity suffers an unrealized loss of about 109% of its equity after a 300 basis-point rate rise. Its year-one net interest income falls only from about $5.6 billion to $4.5 billion. Earnings barely move while the bank becomes insolvent on paper.

The loss becomes real only when depositors leave. A 25% run, met by selling bonds, turns equity negative. The same run met by a central-bank loan against the bonds at face value realizes no loss at all. That is roughly what the Fed's Bank Term Funding Program offered from 12 March 2023, two days too late for SVB. The program closed to new loans on 11 March 2024.

What happened afterward?

The run spread to banks that looked similar. Signature Bank, with about $110 billion of assets and deposits about 90% uninsured, was closed on 12 March. First Republic lost more than $100 billion of deposits in the first quarter even after eleven large banks placed $30 billion with it, and on 1 May its roughly $229 billion of assets were sold to JPMorgan Chase. Two of the three largest US bank failures by assets happened in eight weeks. Our list of the biggest bank failures in US history puts them in context.

The FDIC's estimated cost of protecting uninsured depositors at SVB and Signature, recovered through a special assessment on large banks, was about $16.3 billion in November 2023 and about $16.7 billion as of 30 September 2025, successive estimates of the same loss. In January 2025 the FDIC sued 17 former SVB officers and directors for gross negligence and breach of fiduciary duty; the case was pending at last report. On 28 August 2026 a federal judge ruled that the bank's former parent, now SVB Financial Trust, was not entitled to about $1.7 billion it had held on deposit at the bank, as reported.

The deposits did not vanish; they moved to the largest banks, to money-market funds and to Treasury bills. Uninsured deposits were still 43.3% of US commercial-bank deposits at the end of 2025, so the conditions for another run have not gone away.

Which factor did SVB fail on?

In money-equation terms, SVB failed on the fourth factor: a balance sheet that could not survive the bad year. It was excellent at reaching customers and good at pricing credit. But it funded long fixed-rate assets with deposits that could leave in hours, and when the bad year arrived, the first factor, trust, collapsed in a day because depositors could see the weakness as quickly as anyone. The interest-rate cycle was the weather that exposed it, and regulatory tailoring was the weather that let it go unchecked.

The lesson for anyone reading a bank is to compare the balance sheet with the income statement, as our guide on how to analyze a bank explains, and to ask how stable the deposits are. Our explainer on net interest margin shows why a rising-rate year can look profitable on the income statement while the balance sheet bleeds.

The takeaway

Silicon Valley Bank did not fail because of bad loans or fraud. It failed because it put a deposit boom into long-dated bonds at near-zero yields, removed its hedges, and relied on depositors who were almost entirely uninsured and connected to one another. When rates rose, the losses were hidden by an accounting label until a single announcement made them visible, and the run did the rest. It is the clearest modern case of how banks make money in our guide to how financial companies make money, and how they lose it. The book tells the story in full in Chapter 1 and places it in the history of runs in Chapter 18, "Crises and Runs".