GOVERNANCE WATCH
By InsidEntity Editorial Desk · Sep 28, 2026 · 13 min read
Sep 28, 2026 · @Insidentity
Supervisors knew, or should have known, about Silicon Valley Bank’s weaknesses by March 2022, according to the initial findings of an independent review commissioned by the Federal Reserve, announced by Vice Chair for Supervision Michelle Bowman on 18 September. Several of the same weaknesses were in the public record too. A $15.1bn gap between the carrying value and the fair value of the bank’s held-to-maturity securities was disclosed in financial statements covered by the clean audit opinion KPMG, SVB’s auditor since 1994, signed on 24 February 2023. The bank failed fourteen days later. A clean opinion covers the accounts as they stood; it is not a forecast of what depositors will do next.
InsidEntity record: Silicon Valley Bank (SVB Financial Group, formerly NASDAQ: SIVB)
| Instrument | Score |
|---|---|
| Company Risk Rating (CRR) | 2.75 |
| Director Independence | 5.00 |
| Director Capacity | 3.00 |
| Auditor Independence | 0.00 |
| Shareholder Influence | 3.00 |
| Financial Stability Rating (FSR) | 4 (audited years FY2020–FY2022) |
The CRR is calculated on equal 25% pillar weights, the methodology’s default until directors complete the independence questionnaire; Director Independence carries its starting value of 5.00 until then. On those weights, pillar scores of 5.00, 3.00, 0.00 and 3.00 average to 2.75. The published scale runs from 5 (Benchmark) to 1 (Risk); at 2.75, SVB sits below the 3.00 Good mark.
The review, and a different question
Speaking at Mansion House in London on 18 September, Fed Vice Chair for Supervision Michelle Bowman set out the initial findings of the independent review she commissioned from the Starling Advisory Group. SVB, it found, failed from a confluence of weaknesses: unrealised losses on its securities that exceeded its capital, a run-prone deposit base that was 94% uninsured and concentrated in venture-backed technology companies, and no operational readiness to borrow from the discount window. Supervisory staff knew, or should have known, about them as early as March 2022 and did not act promptly or decisively, which the review attributes in significant part to a culture of risk aversion, compounded by unclear decision rights. It also addresses two explanations that have featured prominently in accounts of the failure: the delays were not caused by the 2018 regulatory tailoring law, and social media did not trigger the run, with no evidence that it accelerated it. Both conclusions depart from the Fed’s own April 2023 review, which listed tailoring among its key takeaways and said the run appeared to have been fuelled by social media. The Fed has published Bowman’s speech but not, as far as we can find, Starling’s report itself, which she described as the first in a series, so its evidence cannot yet be checked.
That is a finding about supervisors, who see confidential information. The question for an investor, a depositor or a director is narrower and, in some ways, harder: what could someone outside the bank see before the run? On the public filings, the answer is more than the speed of the collapse suggests. It sits in three layers: the financial statements, the audit, and the governance record.
Layer one: the notes
At 31 December 2022, SVB classified $91.3bn of securities as held-to-maturity, about 43% of a $211.8bn balance sheet. Held-to-maturity (HTM) accounting lets a bank carry bonds at amortised cost, broadly what it paid for them, rather than at fair value, provided it intends and is able to hold them until they mature. Price movements stay out of earnings and out of equity. The fair value is still disclosed. SVB’s audited balance sheet showed it in parentheses beside the carrying amount, and the securities note gave the detail.
| At 31 December 2022 | $bn |
|---|---|
| HTM securities at amortised cost | 91.3 |
| HTM securities at fair value | 76.2 |
| Unrecognised loss | 15.1 |
| Total equity (book) | 16.3 |
| Equity if the HTM book were marked to market, before any tax benefit | ~1.2 |
Source: SVB Financial Group Form 10-K, FY2022.
Applying a hypothetical 21% tax rate would cut the illustrative hit to about $12bn, still most of the cushion; the actual tax effect would have depended on realisation and the group’s tax position. A year earlier, the unrealised loss on the HTM book had been below 1% of its cost. By the end of 2022 it was about 16.6%. Nor was the year-end figure the first sign: the court in the case discussed below found, from SVB’s own quarterly report, that by 30 September 2022 the unrealised losses on the HTM book already exceeded the bank’s entire equity capital.
So the losses were not hidden. They were disclosed, as the rules require, and not recognised, as the rules allow. Both numbers were true under US GAAP, and both sat in the same annual report. A reader who put them side by side could see how much of SVB’s reported equity rested on the assumption that those bonds would be held to maturity. Whether that mattered turned on the other side of the balance sheet: how likely the bank was to be forced to sell.
Layer two: the audit
KPMG’s report on SVB Financial Group’s 2022 accounts is dated 24 February 2023. It carried no going-concern warning. The critical audit matter it identified concerned the allowance for credit losses on loans and unfunded credit commitments, rather than the interest-rate exposure that later became central to the failure. And it closed, as US audit reports must, with a line on tenure. KPMG had served as the company’s auditor since 1994.
That line is what the InsidEntity score reads. Auditor Independence is a published tenure clock on the current auditor: the score steps down from 5.00 for an auditor in its first one to three years to 0 once the relationship passes ten years. With KPMG in place since 1994, SVB sits well inside the terminal band. On this pillar a zero is an assessed result, not missing data.
Tenure is a governance signal, not a finding that the audit failed. KPMG’s US chief executive said after the collapse that the firm stood behind its reports and had followed professional standards, and the opinion addressed the financial statements for the year to 31 December 2022. It was not a forecast of future events, although auditing standards required the auditor to consider known going-concern conditions up to the report date. Nor was the audit committee short of expertise: it was chaired by Mary Miller, a former US Treasury Under Secretary for Domestic Finance, and four of its five members were designated financial experts. The question the score poses is narrower: what does an auditor relationship of almost three decades signal to a board and its shareholders, in a year when the gap between book and fair value on the securities book widened from under 1% to about 16.6%?
Layer three: the governance record
SVB’s 2023 proxy described a board of twelve directors, eleven of them independent. A separate, independent chair, Beverly Kay Matthews, who spent 36 years at Ernst & Young. A standalone Risk Committee of seven, chaired by Kate Mitchell, co-founder of Scale Venture Partners. And a reporting line written into the proxy: the Chief Risk Officer reports directly to the Risk Committee, and only administratively to the chief executive.
For eight months of 2022 there was no Chief Risk Officer to report. Laura Izurieta ceased to serve in the role on 29 April 2022 under a separation agreement; the proxy says the company had opened discussions with her about a transition early that year. She stayed on in a non-executive transition role until October. Her successor, Kim Olson, previously Chief Risk Officer for the Americas at Sumitomo Mitsui Banking Corporation and earlier a supervisor at the Federal Reserve Bank of New York, took up the role on 27 December 2022; SVB announced the appointment on 4 January 2023.
Source: SVB Financial Group preliminary proxy statement, filed 3 March 2023; SVB announcement, 4 January 2023.
Over the same year the Risk Committee met 18 times, against seven in 2021. The board as a whole met ten times, and its committees 76 more. Meeting counts measure attention. They do not show what the attention produced, and from the outside, before the run, there was no way to tell. A court has since made findings about what the committee did; they are set out below.
Source: SVB Financial Group preliminary proxy statement, filed 3 March 2023.
One caveat on timing matters for a piece about what the record showed. Olson’s appointment was public from January, but the full account of the vacancy, with the dates and the meeting count, came in the preliminary proxy statement filed on 3 March 2023: five days before the capital-raise announcement, a week before the bank failed. A screen reading the record that week would have seen it. Only just.
A CRO vacancy is not a CRR pillar, and it should not be scored as one. It is evidence in the governance record, and it sits beside the pillars rather than inside them.
Why the notes mattered: the deposit base
An unrecognised loss on bonds held to maturity stays on paper for as long as the bank is not forced to sell. SVB’s funding made a forced sale more likely than at most banks. SVB’s own annual report put estimated uninsured deposits in its US offices at $151.5bn at the end of 2022, and noted a further $13.9bn of foreign deposits outside any deposit-insurance regime. The FDIC, citing the bank’s year-end 2022 call reports, puts the uninsured share of domestic deposits at 94%, the figure the Starling review also uses. Those depositors were concentrated among venture-backed technology companies, and SVB’s own proxy summary of 2022 referred to declining deposits and elevated client cash burn.
The sequence in March 2023 was short:
- Wednesday 8 March. SVB announces it has sold about $21bn of available-for-sale securities at an after-tax loss of about $1.8bn, and plans to raise $2.25bn of new capital.
- Thursday 9 March. Depositors withdraw $42bn in a single day, which the FDIC puts at nearly 30% of deposits.
- Friday 10 March. With roughly $100bn more in withdrawal requests queued for the morning, California regulators close the bank and appoint the FDIC as receiver.
The $42bn is one day’s outflow, not a two-day total. The run did not create SVB’s weaknesses; it exposed them. The loss on the held-to-maturity book stayed unrecognised only for as long as the bank could hold the bonds, and a run of that size took the choice away.
Source: FDIC.
What has come out since
Three later sources add to the picture. They carry different weight.
The Fed’s first review, April 2023. The Fed’s internal review found that during 2022, over roughly the same period in which SVB’s interest rate risk models were loosened, the bank removed interest rate hedges that would have protected it against rising rates.
The FDIC’s suit. In January 2025 the FDIC sued six former officers and eleven former directors (FDIC v. Becker, N.D. Cal.). Its sharpest governance allegation concerns a $294m dividend from the bank to its parent in December 2022, which it says was approved while the bank was in financial distress. The FDIC notes that the two boards were identical, so the directors who approved the payment sat on both sides of it. In October 2025 the court refused to dismiss the case. That is a ruling on the pleadings, not a finding of fact; the allegations are untested at trial.
The SVB Financial Trust ruling, August 2026. In a separate case, Judge Beth Labson Freeman ruled after a twelve-day bench trial that the trust which succeeded SVB’s holding company cannot recover a $1.71bn deposit claim from the FDIC. Her 206 pages of findings, filed on 28 August, go further than any earlier public account. She found that the chief financial officer, the global treasurer and the rest of the officer-only asset-liability committee breached their fiduciary duties to the bank. They concentrated its portfolio in long-dated fixed-rate securities, designated most of them held-to-maturity out of concern for the metrics reported to investors rather than the bank’s economic health, and terminated hedges while keeping the securities those hedges had protected. That conduct, she found, caused more than $4.5bn in losses. She held that the $294m December 2022 dividend, whose main purpose was to give the holding company cash for a share buyback, was a voidable transfer. And she found that members of the board’s Finance, Risk and Compensation Committees, the chief executive and the former Chief Risk Officer knew of and encouraged the breaches. Her opening assessment was that the holding company still had a board and management “assembled to run a much smaller regional bank.” None of the individuals was a party to that case, and the FDIC’s own suit against them continues.
The dividend is the kind of fact a governance record can hold: a dated, board-level decision with named approvers.
What the rating scores, and what it does not
The CRR does not see a securities portfolio, an empty risk seat or a dividend. It scores four things: the independence of the board, its capacity, the independence of the auditor, and the influence of shareholders. On current inputs, SVB’s 2.75 breaks down like this:
- Director Independence, 5.00. The starting value the methodology uses until directors complete the independence questionnaire. It is not an assessment of this board; SVB’s own proxy counted eleven of its twelve directors as independent under Nasdaq rules.
- Director Capacity, 3.00. A board-level measure of bandwidth and how thinly directors’ commitments are spread, not a judgement on any individual director.
- Auditor Independence, 0.00. KPMG since 1994, far past the ten-year point where the tenure clock reaches zero.
- Shareholder Influence, 3.00. Scored on the number of shareholders and how concentrated the holdings are.
The auditor pillar is doing real work here. Holding every other input constant, an auditor in its first three years would have put the same record at 4.00. The tenure clock alone accounts for 1.25 points, and it is the reason SVB sits below the Good mark rather than in the Excellent band.
Source: InsidEntity platform.
The Financial Stability Rating answers a different question. It is built only on full-year audited statements, and on the three audited years to December 2022 it scored SVB 4, Excellent, on the same five-point scale. That is the argument of this piece in a single number. The FSR reads the accounts as reported, and as reported SVB’s total equity stood at $16.3bn. The $15.1bn was disclosed beside it, not deducted from it.
No screen could have told anyone the date of the run, and this one does not pretend to. What the public record did show before the run was a $15.1bn unrecognised loss set against $16.3bn of total equity, an auditor relationship that had passed every point on the tenure clock, and, in a proxy filed a week before the failure, a risk committee that had met 18 times in a year when the risk chief’s seat sat empty for eight months of it.
Several of SVB’s vulnerabilities were already visible in its public record before the run. The filings did not tell you when the run would happen. They told you what was sitting on the balance sheet before it did.
Sources
- Bowman, Initial Findings from Independent Review of Silicon Valley Bank, 18 Sept 2026
- SVB Financial Group preliminary proxy statement, filed 3 March 2023
- SVB Financial Group Form 10-K, FY2022
- KPMG consent referencing its report dated 24 Feb 2023
- SVB: SVB hires Kim Olson as Chief Risk Officer, 4 Jan 2023
- Barr testimony to the Senate Banking Committee, 28 March 2023
- Board of Governors of the Federal Reserve System, Review of the Federal Reserve’s Supervision and Regulation of Silicon Valley Bank, 28 April 2023
- FDIC memorandum on authority to sue former SVB officers and directors
- FDIC v. Becker, N.D. Cal., complaint (January 2025) and order on the motions to dismiss (October 2025)
- FDIC: Successfully Managing Systemic Risk (nearly 30% of deposits left in hours)
- SVB Financial Trust v. FDIC, No. 5:24-cv-01321-BLF, Findings of Fact and Conclusions of Law, ECF No. 417, 28 Aug 2026
This is independent research, not financial advice.