Sources: SVB bought long-term assets that paid more interest based on a misplaced profit-driven strategy, falling out of compliance with its internal risk model
and over the concerns of some staffers — SVB executives simply changed the model's assumptions, according to the former employees and securities filings.” https://www.washingtonpost.com/ ... Matt O'Brien / @obsoletedogma : It's like thinking you can solve climate change by getting rid of thermometers. Or, more accurately, that you can make more money until the crisis comes by getting rid of the thermometers, and then saying “who could have known there'd be a crisis?” https://twitter.com/...
Context & Ripple Effects
The collapse narrative now has a mechanism. Earlier reporting traced the damage through unrealized losses that snowballed from $0 in June 2021 to $16B by September 2022 on SVB's long-duration bond book; this Washington Post piece explains how the bank got there — executives chasing higher-yielding long-term assets until the portfolio broke the bank's own internal risk model, then editing the model's assumptions over the objections of staff rather than shrinking the book.
That detail retroactively sharpens the earlier arc: Greenoaks' November warning to founders and the quiet withdrawals that followed now look like the informed reading of a risk framework management had already hollowed out. The story moves from 'SVB misread rates' to 'SVB changed the ruler when it didn't like the measurement' — which is why it lands as a governance failure, not just a duration bet gone wrong.
First-order effects
- SVB executives and the staffers who objected become the central actors in any post-mortem, with securities filings cited as evidence that the model's assumptions were changed to accommodate — or mask — non-compliance.
- With the bank closed and deposits frozen, the distressed unwinding continues on the asset side too: SVB agreed to sell its VC arm SVB Capital, managing roughly $9.8B, to a Brookfield- and Sequoia Heritage-backed entity for $340M in cash.
Second-order effects
- Sophisticated depositors like Greenoaks demonstrated that reading a bank's disclosed risk posture — and acting before the crowd — is now a core VC operating skill, accelerating the herd-style runs the Bloomberg coverage identified as a collapse factor.
- Rival banks and their regulators face pressure to treat internal risk-model assumptions as audited disclosures rather than internal preferences, since SVB shows a model can be quietly re-parameterized until it approves whatever the balance sheet already did.
Third-order effects
- If the pattern holds, banking oversight shifts from checking whether a bank complies with its risk model to checking whether the model itself is honest — an audit of assumption-setting, not just limit-breaches.
- The deeper structural lesson is that duration-mismatch risk doesn't disappear when the measurement does; the $0-to-$16B loss curve shows the exposure compounding silently for over a year while the internal framework was adjusted to tolerate it.
The trend: Bank risk management is moving from model compliance to model integrity, as SVB shows the binding constraint on risk-taking is only as real as the willingness to leave the model's assumptions untouched.