What Happened
Silicon Valley Bank was the go-to bank for venture capital firms and their portfolio companies. By 2022 it held $212 billion in assets — but had made a catastrophic bet on long-duration bonds when interest rates were near zero. When the Fed hiked rates 525bps, those bonds lost enormous value. SVB was sitting on $15 billion in unrealized losses.
On March 8, 2023, SVB announced it was selling its bond portfolio at a $1.8 billion loss and raising equity. Instead of reassuring investors, the announcement triggered panic. Peter Thiel’s Founders Fund told its portfolio companies to withdraw deposits. Within 48 hours, $42 billion in deposits had been withdrawn or were in queue. SVB was closed on March 10.
The FDIC, Treasury, and Federal Reserve announced on March 12 that all depositors would be made whole — above the $250,000 FDIC limit. Signature Bank was closed the same day. First Republic followed in May. Credit Suisse required emergency merger with UBS. The crisis demonstrated that the regional bank model of borrowing short and investing long is fundamentally unstable in a rising rate environment.
The Mechanism
Duration mismatch and the social media bank runBanks borrow short (deposits, which can be withdrawn on demand) and invest long (mortgages, bonds). This is inherently unstable but manageable when rates are stable. SVB invested in long-duration Treasuries at near-zero yields. When rates rose 525bps, those bonds fell dramatically in value. The bank was technically insolvent on a mark-to-market basis. When depositors found out — via Twitter and group chats — the bank run took 48 hours instead of days or weeks. Social media fundamentally changed the speed of bank runs.
What the Consensus Believed
The prevailing view before the reckoning
SVB was a well-run bank serving the innovation economy. The deposit base was sticky (tech companies do not move banking relationships frequently). The duration risk was manageable. The bank had been growing rapidly and was a model for serving the startup ecosystem.
What the Record Shows
Duration mismatch is always a risk
The oldest risk in banking — borrowing short and lending long — destroyed SVB. This is not a new risk. It is the same risk that destroyed the S&Ls in the 1980s.
Social media changes crisis dynamics
SVB failed in 48 hours. Pre-social media bank runs took weeks. Group chats among startup founders coordinated the withdrawal in real time. Regulators are not equipped for this speed.
The FDIC limit creates perverse incentives
Above $250K, deposits should be at risk. Instead, the government made all depositors whole. This creates a guarantee that depositors at banks with concentrated, sophisticated depositor bases had not earned.
The Fed failed to supervise
The San Francisco Fed knew about SVB’s duration risk and did not force corrective action. Regulatory capture in the startup ecosystem was real.
↑ Cognitive pattern: Recency Bias — Assuming the recent era of low rates was the new normal
Key Voices
Called It Right
Jamie Dimon
JPMorgan
“SVB failure was entirely predictable. When you borrow short and lend long in a rising rate environment this is what happens.”
March 13, 2023 Right
Bank Analysts / Short Sellers
Various
“The regional bank model is broken. SVB is not the last. Signature, First Republic, and others have similar duration mismatches.”
March 10, 2023 Right
Wrong
SVB Management
Silicon Valley Bank
“SVB is a well-run bank serving the innovation economy. The deposit base is sticky and the duration risk is manageable.”
December 2022 Failed March 2023
First Republic Management
First Republic Bank
“First Republic will survive. The deposit outflows are slowing and the large bank consortium deposit is stabilizing.”
March 20, 2023 Failed May 2023