Forty Hours: Silicon Valley Bank, March 2023

Silicon Valley Bank was not a typical commercial lender. Founded in 1983 in Santa Clara, California, it had grown into the primary banking relationship for nearly half of all US venture-backed technology and life-sciences companies, along with a substantial share of their investors. By 31 December 2022, it held $175.4 billion in deposits, a figure recorded in its final annual filing with the Federal Deposit Insurance Corporation. Its customer base was structurally unusual: deposits were overwhelmingly from corporate accounts, most of them well above the FDIC's $250,000 insurance ceiling, which meant that in any crisis, most depositors had strong incentive to move fast.

The bank's vulnerability had been building since 2021. During the pandemic-era liquidity surge, SVB had deployed a large portion of its deposit base into long-duration US Treasury bonds and agency mortgage-backed securities — assets that carried minimal credit risk but were acutely sensitive to interest rates. When the Federal Reserve began its rate-hiking cycle in March 2022, the market value of those holdings fell sharply. Under accounting rules that classified most of the portfolio as "held-to-maturity," the losses were not recognized on the income statement — until the bank was forced to sell.

The sequence that destroyed the bank began on the afternoon of Wednesday, 8 March 2023. SVB Financial Group, the bank's parent, announced that it had sold approximately $21 billion of its available-for-sale securities portfolio at a loss of roughly $1.8 billion after tax, and that it intended to raise $2.25 billion in new equity to shore up its balance sheet. The announcement was intended to reassure; it did the opposite.

By the evening of 8 March, messages were circulating in the group chats and Signal threads that connect founders, general partners, and CFOs across Sand Hill Road and beyond. Several prominent venture firms communicated to their portfolio companies that they should evaluate their exposure. The advice varied in its urgency, but the direction was consistent: understand what you have at SVB and consider your options. Some accounts have described individual general partners advising withdrawal; the documented record is that deposits began leaving at scale within hours of the announcement. According to the California Department of Financial Protection and Innovation, depositors attempted to withdraw $42 billion on 9 March alone — roughly a quarter of the bank's total deposit base in a single day.

By the morning of Friday, 10 March, SVB had a negative cash balance. California regulators closed the bank that afternoon and appointed the Federal Deposit Insurance Corporation as receiver — making it, at that moment, the second-largest bank failure in American history by assets, surpassed only by Washington Mutual's 2008 collapse. The entire sequence from the capital raise announcement to federal takeover consumed approximately forty hours.

The FDIC initially created the Deposit Insurance National Bank of Santa Clara to hold insured deposits. The status of uninsured deposits — the vast majority, by dollar value — was immediately uncertain, and that uncertainty produced a second wave of anxiety across the startup ecosystem over the weekend of 11–12 March. Payroll cycles were days away for hundreds of companies. Treasury officials and regulators spent the weekend assessing systemic risk. On Sunday, 12 March, the Treasury Department, the Federal Reserve, and the FDIC issued a joint statement invoking the systemic risk exception, guaranteeing all deposits at SVB regardless of size. A separate announcement covered Signature Bank, a New York-based institution also closed that weekend.

The SVB collapse was not a credit crisis in the traditional sense — the bank's loan book was not rotting. It was a duration mismatch and a concentration failure, accelerated by the structure of its depositor base and the speed of modern communication. The bank run that ended SVB did not require depositors to queue at a branch; it happened in group chats, on Slack channels, and through wire transfer portals that processed instructions instantly.

The concentration was the more durable finding. SVB had become, over four decades, the de facto operating bank for the venture-backed technology sector. Founders opened accounts there because their investors banked there, and because SVB offered services — venture debt, warrant coverage, relationships with term-sheet signers — that conventional banks did not. That network effect, which had made SVB so valuable to founders, had also concentrated an extraordinary proportion of startup operating capital in a single institution. When the run began, the interconnectedness that had built the bank became the mechanism of its destruction.

The episode also foregrounded the role of venture capital firms in accelerating the crisis. Whether the advice circulating on 8 and 9 March constituted rational fiduciary behavior toward portfolio companies or a collectively self-fulfilling prophecy — or both simultaneously — was a question that regulators and commentators raised in the weeks that followed. No firm was found to have acted unlawfully. The speed and network-mediated character of the run nonetheless prompted the FDIC to commission a review of deposit insurance limits and the particular risks posed by highly networked, institutionally concentrated depositor bases.

SVB's UK subsidiary was sold to HSBC for £1 on 13 March 2023, following an intervention brokered by the Bank of England and HM Treasury, protecting deposits in the UK entity without calling on public funds.

The Federal Reserve's own post-mortem, published in April 2023, identified supervisory failures alongside the bank's internal risk management. Examiners had flagged interest-rate risk concerns in prior years; escalation had been insufficient. The report noted that SVB had grown so quickly — its assets roughly tripled between 2019 and 2021 — that supervisory resources had not kept pace.

For the startup ecosystem, the episode illuminated a structural assumption that had gone unexamined: that the banking infrastructure supporting venture-backed companies was as robust as the capital behind them. The forty hours of March 2023 demonstrated that it was not, and that the same network density that had made Sand Hill Road a functional community could, under the right conditions, become a contagion mechanism. The FDIC's guarantee prevented a broader cascade, but the failure itself was already complete.