Silicon Valley Bank 2023: The Fastest Bank Run in History

Situation

It is March 9, 2023. Silicon Valley Bank — the 16th largest US bank by assets, with $209 billion in total assets and $175 billion in deposits — opened for business this morning. By 9:00 AM Pacific, it is already too late.

SVB's depositor base was unlike any other bank: 97% of deposits were uninsured (above the FDIC's $250,000 limit), concentrated among technology startups, venture capital firms, and the founders of Silicon Valley. When interest rates were near zero and venture funding was abundant (2020–2021), deposits flooded in. SVB deployed those deposits into long-dated US Treasury bonds and agency mortgage-backed securities — buying safety, but creating a massive duration mismatch.

Then the Fed raised rates at the fastest pace since the 1980s. SVB's $117 billion bond portfolio lost approximately $17 billion in market value — losses that were masked by HTM accounting classification. By January 2023, SVB had unrealized losses equal to 100% of its total equity.

Yesterday — March 8 — SVB management announced three things simultaneously:

1. It had sold its entire $21 billion AFS portfolio, realizing a $1.8 billion after-tax loss
2. It was seeking to raise $2.25 billion in equity (Goldman Sachs was the underwriter)
3. Moody's had downgraded SVB two notches

Within hours, Founders Fund, Coatue, Union Square Ventures, and other major VC firms were telling portfolio companies to withdraw. By end of day March 8, $42 billion in withdrawal requests had been submitted — more than $1 million per second at peak.

The decision moment

You are the Chief Risk Officer at a regional bank with a similar deposit profile (tech-heavy, largely uninsured). It is March 9, 2023 — the morning of SVB's collapse. You have been watching SVB closely for two months. Before you speak to your CEO about your own institution's position, you must understand what went wrong.

Three analytical questions:

1. The HTM accounting shield. SVB classified $91.3 billion of its bond portfolio as Held-to-Maturity — a designation that removes unrealized losses from AOCI (Accumulated Other Comprehensive Income) and thus from regulatory capital calculations. This legally reduced SVB's visible capital impairment for nearly a year. The AFS portfolio ($26 billion) did flow through AOCI. When SVB sold the AFS portfolio and realized the $1.8B loss, it simultaneously signaled that the HTM portfolio had similar (but hidden) losses. How does HTM classification change the information available to investors, depositors, and regulators — and what are the appropriate limits on its use?

2. The depositor network effect. SVB's customers were not independent retail depositors who would each individually decide whether to withdraw based on their own analysis. They were a tightly networked community of VCs and founders who shared information in real time via text, Slack, and Twitter. When Founders Fund advised withdrawal, that advice spread to hundreds of portfolio companies within hours. Traditional bank run models assume independent action. How does networked behavior change the dynamics — and what does that imply for deposit insurance policy, early warning systems, and bank run prevention?

3. The regulatory failure. The Federal Reserve's San Francisco regional office had flagged SVB's interest rate risk in its supervisory review. A Matter Requiring Attention (MRA) was issued in 2022. SVB also had no Chief Risk Officer for nine months of 2022 during a critical period of rate increases. The Fed upgraded its concern level but did not force remediation before the collapse. What is the appropriate standard of care for regulators who identify but do not resolve systemic risk in individual institutions?

Key financial datapoints

The collapse mechanics

SVB's failure was a textbook asset-liability mismatch amplified by depositor concentration:

Step 1 — Deposit surge (2020–2021): Low rates + abundant VC funding → tech startups deposited excess cash at SVB. Deposits grew from $62B (2019) to $189B (peak, 2021). SVB needed to deploy capital and bought long-duration Treasuries and MBS at near-zero yields.

Step 2 — Rate shock (2022): Fed raised rates from 0.25% to 4.75% in 13 months. SVB's long-dated bonds fell in market value. The HTM portfolio ($91B) could not be sold without realizing the loss — and selling it would also reclassify remaining HTM bonds into AFS, creating further required loss recognition.

Step 3 — Deposit outflow (2022–2023): As tech funding dried up, startups burned cash faster and maintained smaller balances. SVB deposits fell from $189B to $173B over 2022. This forced SVB to sell assets to fund withdrawals — but its only liquid assets were in the AFS portfolio.

Step 4 — The AFS sale announcement (March 8): Selling the AFS portfolio and announcing a $2.25B equity raise signaled to sophisticated depositors that SVB was in capital distress. Networked VC firms immediately advised withdrawal.

Step 5 — The bank run (March 9): $42B in withdrawal requests filed in one day. SVB could not fund them. FDIC seized the bank at 9:45 AM Pacific time on March 10, 2023 — before West Coast business hours even began.

Frameworks invoked

• Asset-Liability Management (ALM). Duration mismatch is the fundamental risk in banking: borrowing short (deposits) to lend long (bonds/loans). When duration mismatch is severe and the rate environment shifts sharply, the gap between asset value and liability value creates a solvency crisis. ALM frameworks should stress-test for parallel shifts, steepening, and flattening yield curve scenarios.
• HTM vs. AFS Accounting. Held-to-Maturity securities are carried at amortized cost — unrealized losses do not flow through income or equity until realized. AFS securities mark to market through AOCI (equity). HTM classification is legitimate for genuinely illiquid, core holdings; it is misused when it obscures the true economic net worth of an institution from regulators and counterparties.
• Depositor Concentration Risk. Traditional bank regulation focuses on loan concentration. SVB's risk was on the liability side: a homogeneous, networked depositor base that could coordinate withdrawal at network speed. Herfindahl-Hirschman Index analysis of depositor concentration would have flagged this.
• Information Contagion. The VC/startup community's real-time communication infrastructure (Twitter, group chats, Slack) allowed bank run coordination that would have taken days or weeks in prior eras. This is a structural change in bank run dynamics that existing deposit insurance models were not designed for.

Discussion questions

1. SVB classified $91.3 billion of bonds as Held-to-Maturity in 2021–2022, removing those unrealized losses from regulatory capital calculations. The losses were real — the accounting just deferred their recognition. Should regulators require banks above a certain size to include HTM unrealized losses in capital calculations? What are the tradeoffs?
2. Greg Becker, SVB's CEO, sold $3.6 million in SVB stock on February 27 — less than two weeks before the bank failed. The sale was made under a pre-scheduled 10b5-1 plan filed in January. How do pre-scheduled trading plans affect your analysis of executive stock sales as a signal of insider knowledge?
3. KPMG signed an unqualified audit opinion on SVB's financial statements on March 1, 2023 — nine days before the FDIC seizure. The audited financials did not flag going-concern risk. What should have been visible to auditors in January/February 2023 that would have warranted either a qualified opinion or a going-concern disclosure?
4. The US government guaranteed all SVB deposits — including $151 billion in uninsured deposits — on March 12, 2023. Critics argued this implicitly expanded deposit insurance without Congressional approval. Defenders argued it was necessary to prevent contagion to other regional banks. Was the decision correct? What precedent does it set?
5. SVB's depositor base was 97% uninsured. Traditional retail banking is dominated by insured depositors who have little incentive to run. If you were designing a regulatory framework for "VC-primary banks" — institutions whose depositors are sophisticated, networked, and mostly uninsured — what specific rules would you impose that differ from standard bank regulation?

The real outcome (revealed at session end)

March 10, 2023: FDIC seizes SVB. Approximately $175 billion in deposits frozen.

March 12, 2023: US government announces that all SVB deposits — including uninsured — will be made whole. Signature Bank (NY) also seized the same day. The systemic risk exception is invoked for the first time since 2008.

March 13, 2023: First Republic Bank shares fall 62%; regional bank contagion begins.

March 27, 2023: First Citizens BancShares acquires SVB's loans and deposits in an FDIC-assisted transaction.

May 1, 2023: First Republic Bank seized by FDIC and sold to JPMorgan Chase — the second-largest US bank failure ever.

The lesson: Asset-liability mismatch is the oldest risk in banking. What changed in 2023 was the speed of the run (networked depositors), the visibility of the risk (HTM accounting masked losses), and the concentration of the depositor base (97% uninsured). Traditional early warning systems — designed for retail bank runs that take days — could not detect a networked run that compressed the crisis into 12 hours.

Sources

• Federal Reserve Board, "Review of the Federal Reserve's Supervision and Regulation of Silicon Valley Bank" (April 2023).
• FDIC, "FDIC's Supervision of Signature Bank" (April 2023).
• SVB Financial Group Form 10-K (December 31, 2022), filed February 24, 2023.
• KPMG audit opinion on SVB Financial Group (March 1, 2023).
• US Senate Banking Committee hearings on SVB collapse (March 2023).
• Wall Street Journal, "How Silicon Valley Bank Fell in 48 Hours" (March 2023).