Silicon Valley Bank was in "sound financial condition" on Wednesday. By Friday it was insolvent — the second-largest bank failure in U.S. history. $42 billion withdrawn in 10 hours. $1 million every second. The stress tests were running. The auditors were signed off. The cascade was invisible to all of them.
On the morning of Wednesday, March 8, 2023, Silicon Valley Bank held $209 billion in assets, ranked as the 16th largest bank in the United States, and served nearly half of all U.S. venture capital-backed healthcare and technology companies.[5] One month earlier, Forbes had added SVB to its "Financial All-Stars" list. The California Department of Financial Protection and Innovation considered the bank to be in sound financial condition.[1]
By Friday morning, it was insolvent. The California regulator seized it and placed it into FDIC receivership.[7] It was the second-largest bank failure in U.S. history, the fastest ever, and the beginning of a cascade that would destroy three of the four largest banks to ever fail in the United States — all within two months.[10]
SVB was audited by KPMG. It filed stress tests. It maintained regulatory capital. It was rated, reviewed, and supervised. Every traditional risk tool was running.
The core failure was not exotic. SVB took pandemic-era deposit inflows ($62 billion to $198 billion in two years), invested them in long-duration government bonds, and watched as rising interest rates destroyed the market value of those bonds. By the end of 2022, the bank had $117 billion in a bond portfolio that was underwater. Its chief risk officer had stepped down in April 2022 and was not replaced until January 2023 — a nine-month gap coinciding with the steepest rate increases in decades.[1]
But the textbook failure is not the story. The story is what happened when the failure became visible — the speed, the propagation, and the cross-dimensional cascade that no traditional risk framework was built to model. SVB repeatedly failed its own internal liquidity stress tests, then switched to less conservative assumptions to mask the results.[1] The Fed's own post-mortem admits that supervisors did not fully appreciate the extent of the vulnerabilities.[1] KPMG signed off on all three banks that failed.[5] The tools were running. The cascade was invisible to them anyway.
Pandemic-era tech funding floods SVB. Deposits grow from $62 billion to $198 billion in two years.[5] The bank invests in long-duration government bonds, reaching for yield in a low-rate environment. The $117 billion bond portfolio is the ticking clock no one hears.[9]
D5 Origin PlantedSVB's CRO steps down. A replacement is not named until January 2023.[1] This nine-month vacuum coincides with the steepest interest rate increases in four decades. The position responsible for catching the cascade is empty while the cascade builds.
D5 Risk GapFederal Reserve supervisors begin warning SVB management about interest rate exposure.[1] The bank repeatedly fails its own internal liquidity stress tests. Management responds by switching to less conservative assumptions, masking the risk rather than addressing it.[1]
D4 Regulatory FailureTo cover withdrawals, SVB sells its entire available-for-sale portfolio at a $1.8 billion loss and announces a $2.25 billion stock offering. The announcement — made shortly after Silvergate Bank began liquidating[12] — terrifies the market. Moody's downgrades SVB that evening.[9] Peter Thiel's Founders Fund tells portfolio companies to pull their money.[3]
⚡ Cascade Trigger$1 million exits every second.[6] SVB's stock plummets 60%. CEO Greg Becker holds a video call begging VCs to "stay calm." They do not.[3] By close of business, SVB has a negative cash balance of $958 million.[11] Another $100 billion in withdrawals is queued for Friday morning.[4] The bank cannot process the Fed's overnight cash letter.
🏦 D1 Bank RunCalifornia regulators close SVB and place it in FDIC receivership before markets open. Trading in SVB shares is halted. Employees receive their annual bonuses hours before the government takes control. CEO Greg Becker had sold $3.6 million in company stock two weeks earlier.[5]
D3 CollapseSignature Bank collapses — $20 billion withdrawn in two hours.[7] Treasury, Fed, and FDIC issue joint statement guaranteeing all deposits at both banks, far exceeding the $250,000 FDIC limit.[1] The Bank Term Funding Program is created. A new precedent: all deposits, regardless of size, are now implicitly guaranteed.
D4 Systemic ResponseDespite a $30 billion emergency lifeline from 11 major banks, First Republic loses over $100 billion in deposits and is seized by the FDIC.[8] JPMorgan acquires the remains. Three of the four largest bank failures in U.S. history have now occurred in a 49-day window.[10] Total FDIC cost: $33+ billion.[7]
D6 ContagionThe U.S. Senate Permanent Subcommittee on Investigations releases a 292-page report.[5] KPMG audited SVB, Signature, and First Republic at the time each failed. The report calls out inadequate auditing practices. The same auditor signed off on every bank that died.
D4 Audit FailureThe cascade originates in D5 (Quality/Risk Management) — a fundamental asset-liability mismatch that was visible on the balance sheet, flagged by Fed supervisors, and masked by management through less conservative stress test assumptions. The failure propagated across all six dimensions within 48 hours, making it one of the most rapid cross-dimensional cascades in the case library.
| Dimension | Score | What Happened |
|---|---|---|
| Quality / Risk Management (D5) Origin — 78 | 78 |
$117B bond portfolio underwater. 88% of deposits uninsured. CRO position vacant for 9 months during steepest rate hikes in decades. Bank repeatedly failed its own internal liquidity stress tests — then switched to less conservative assumptions to mask the results. KPMG signed off.[5]
Duration Mismatch |
| Customer (D1) L1 — 88 | 88 |
$42 billion withdrawn in 10 hours — the fastest bank run in U.S. history. Another $100B queued for the next morning. Peter Thiel's Founders Fund pulled all deposits and urged portfolio companies to do the same. Nearly half of all U.S. venture-backed tech and healthcare companies banked with SVB. Circle's $3.3B in deposits at risk broke the USDC stablecoin peg.[12]
Twitter-Fueled Run |
| Revenue / Financial (D3) L1 — 85 | 85 |
$1.8B realized loss on bond sale. Stock fell 60% in hours. SVB Financial filed Chapter 11 bankruptcy one week later. FDIC estimated $20 billion cost to the Deposit Insurance Fund. CEO sold $3.6M in stock two weeks before collapse. DOJ and SEC launched investigations.[5]
Total Wipeout |
| Regulatory (D4) L1 — 85 | 85 |
Unprecedented federal intervention. Treasury, Fed, and FDIC jointly guaranteed all deposits — far beyond the $250K FDIC limit. Created the Bank Term Funding Program. Fed's own post-mortem admitted supervisory failures. 2025 Senate investigation blamed KPMG. 2018 deregulation (which SVB's CEO lobbied for) had weakened stress test requirements.[1]
Systemic Override |
| Operational (D6) L2 — 75 | 75 |
Contagion to Signature and First Republic. Signature Bank lost $20B in deposits in two hours. First Republic lost $100B in one month despite a $30B emergency lifeline. Three of the four largest bank failures in U.S. history in 49 days. Moody's changed its outlook on the entire U.S. banking system from stable to negative.[7]
Systemic Contagion |
| Employee (D2) L2 — 65 | 65 |
Employees received annual bonuses hours before seizure. CEO investigated for pre-collapse stock sales. HSBC acquired SVB UK for £1.[5] Industry-wide talent disruption as startups scrambled to make payroll — companies like Etsy, Roku, and Roblox faced immediate operational uncertainty.[8]
Workforce Shock |
Methodology (expected risk management capability): 90
Performance (actual risk detection): 15
DRIFT = 90 − 15 = 75 (Extreme gap — risk tools running but completely failed to prevent or predict)
Chirp (weighted signal across 6D): 63.4
|DRIFT| (methodology − performance): |90 − 15| = 75
Confidence (source quality × data completeness): 0.75
FETCH = 63.4 × 75 × 0.75 = 3,566 → EXECUTE — CRITICAL (threshold: 1,000)
SVB is not a story of absent risk management. It is a story of present, functioning, audited risk management that could not see the cascade. Every traditional tool was running: internal stress tests, external audits, regulatory examinations, credit ratings. And every one of them failed to prevent or predict the 48-hour collapse.
The internal liquidity stress tests (ILSTs) flagged problems repeatedly — then management switched to less conservative assumptions to mask them.[1] This is not a failure of the tool; it is a failure the tool enabled. When a stress test can be neutralized by changing its inputs, the test is a compliance artifact, not a risk control.
KPMG audited not just SVB but also Signature Bank and First Republic Bank. All three failed.[5] The 2025 Senate investigation called out inadequate auditing practices. When the same auditor signs off on every institution that collapses, the audit process is not catching cross-dimensional propagation — it is certifying that each dimension, in isolation, looks acceptable.
The Federal Reserve's own supervisors began warning SVB about interest rate risk in November 2021. The warnings went unheeded. The Fed's post-mortem attributed this to a combination of bank mismanagement and supervisory inadequacy.[1] But the deeper structural problem is that traditional risk frameworks evaluate dimensions independently. An asset-liability mismatch (D5) is scored separately from deposit concentration risk (D1), which is scored separately from regulatory capital adequacy (D4). No tool modeled what happens when all three fail simultaneously — and propagate at social-media speed.
Silicon Valley Bank's board of directors and management failed to manage their risks. Supervisors did not fully appreciate the extent of the vulnerabilities.
— Federal Reserve Review of the Supervision and Regulation of SVB, April 28, 2023
The Self-Referential Cascade — UC-038 analyzed why traditional frameworks (FMEA, risk matrices, bowtie analysis) cannot model cross-dimensional propagation. SVB is the proof case: the cascade mapped in UC-038's framework comparison table is exactly what happened here. The tools scored each dimension independently. The cascade happened between them. Read UC-038 →
The previous largest bank run in modern U.S. history was Washington Mutual in 2008: $16.7 billion over 10 days.[11] SVB lost $42 billion in 10 hours — a 25× increase in velocity.[2] The next day, $100 billion more was queued.[4] As House Financial Services Chair Patrick McHenry put it, this was the first Twitter-fueled bank run.[3]
The velocity itself is a cascade dimension that traditional models cannot accommodate. Risk matrices assign static severity and likelihood scores. FMEA evaluates failure modes as independent events. Stress tests model scenarios over quarters or years. None of them include a variable for "how fast will panic propagate through a tight-knit community on social media and private Slack channels?"
SVB's depositor base was uniquely concentrated: venture capital firms and their portfolio companies, most of them within one degree of separation from each other. When Founders Fund pulled its money and urged others to follow, the message traveled through the same networks that funded the startups.[3] The bank run was not a random panic — it was a coordinated exit by a community that communicates in real time.
This velocity problem is not SVB-specific. It is structural. Any institution with concentrated, interconnected depositors or customers, operating in an era of instant communication and digital withdrawals, faces the same dynamic. The risk models built for a 10-day bank run cannot protect against a 10-hour one. The cascade doesn't wait for the quarterly review.
SVB repeatedly failed its own internal liquidity stress tests. Rather than address the shortfall, management switched to less conservative assumptions to produce passing results.[1] When a risk tool can be neutralized by adjusting its inputs, it is not a control — it is a compliance theater prop. The tool existed. The cascade was invisible to it because the tool was designed to be adjusted, not to be obeyed.
KPMG audited SVB, Signature Bank, and First Republic Bank. All three failed within two months.[5] The 2025 Senate investigation blamed inadequate auditing practices. When the same auditor certifies every institution that collapses, the question is not whether the audit was wrong — it is whether auditing, as currently practiced, can detect cross-dimensional cascade risk at all.
SVB's bank run was 25 times faster than Washington Mutual's in 2008.[11] Social media, digital banking, and interconnected depositor networks compressed what used to take 10 days into 10 hours.[3] Risk models calibrated to historical run speeds cannot protect against runs that propagate at network speed. The velocity itself is the unmodeled dimension.
SVB's asset-liability mismatch (D5) was visible on the balance sheet. Its deposit concentration (D1) was known. Its regulatory capital (D4) met requirements. Each dimension, evaluated in isolation, appeared manageable. The cascade happened in the space between dimensions — the propagation from D5 to D1 to D3 to D4 to D6 in 48 hours. Traditional tools scored each box. None of them modeled the chain.
SVB's stress tests were running. Its auditor signed off. Its regulators were watching. The 6D Foraging Methodology™ maps the space between dimensions — the propagation chains that traditional tools are structurally blind to.