Back to Atlas
Finance6 min read

The collapse of Silicon Valley Bank

Reconstructing the sequence of signals before the failure

Overview

The collapse of Silicon Valley Bank in March 2023 was the second-largest bank failure in American history. It unfolded over 44 hours — but the signals that preceded it emerged over nearly two years.

Understanding what happened requires reconstructing the sequence: the deposit surge during the tech boom, the interest rate hikes that eroded the bond portfolio, the sector contraction that triggered withdrawals, and the 48-hour cascade that ended in regulatory intervention.

Each signal was visible. The sequence was not — until it was too late.

Timeline Reconstruction

The collapse can be traced through these key moments:

2021

Deposit surge

SVB deposits nearly triple during tech boom, reaching $189 billion

Q1 2022

Rate hike cycle begins

Federal Reserve begins raising interest rates

Q3 2022

Unrealised losses grow

Bond portfolio loses value as rates rise; losses reach $15.9 billion

Late 2022

Tech sector contraction

VC-backed companies begin drawing down deposits

Feb 2023

Internal stress signals

Liquidity concerns emerge internally

Mar 8, 2023

Securities sale announced

SVB sells $21 billion bond portfolio at $1.8 billion loss

Mar 8, 2023

Capital raise announced

Bank announces plan to raise $2.25 billion

Mar 9, 2023

Depositor panic

$42 billion in withdrawal requests in single day

Mar 10, 2023

FDIC takeover

California regulators close SVB; FDIC appointed receiver

Signals Involved

A comprehensive reconstruction would require signals from:

Federal Reserve announcementsQuarterly financial filingsInterest rate movementsDeposit flow dataTech sector funding reportsSocial media sentimentVC fund communicationsAnalyst reportsRegulatory correspondence

Why Chronology Matters

In hindsight, each signal in the SVB collapse was visible to someone. Analysts tracked the unrealised losses. Regulators monitored the deposit concentration. VCs knew their portfolio companies were burning cash.

What was missing was not information — it was chronology. No single view showed how these signals connected over time, how the sequence was accelerating, or how close the system was to cascade failure.

Understanding systemic risk requires understanding sequence. The same signals that appear stable in isolation can appear alarming when viewed as a chronology.

If warning signals are emerging across your organisation, could the sequence be reconstructed in time?

Request a demo