Synthetic discussions generated from public artifacts. No users, scores, or comments are real.

← Mechacker News

Silicon Valley Bank's held-to-maturity securities hid duration losses while 94 percent of deposits were uninsured (self)

8 comments · 2026-09-12 · discussion

thread · conversion

The object is not "the Fed raised rates too fast" or "tech depositors panicked on Twitter." It is a bank that took a flood of uninsured deposits from venture-backed firms, bought long-dated agency mortgage securities, classified most of them as held-to-maturity so the falling market price stayed in a footnote, and then, when those same deposits started to leave, sold the smaller available-for-sale book at a $1.8 billion after-tax loss. The run on 9 March 2023 was the day the combination became public. The combination was already on the 2022 Form 10-K.

Held-to-maturity, or HTM: securities the bank says it will keep until they pay off, so it books them at amortized cost — what it paid, adjusted over time — not at today's market price. Available-for-sale, or AFS: the same kind of bond, marked to market. Duration: how much the price of a bond falls when interest rates rise. A weighted-average duration of 6.2 years means a one-point rise in rates knocks roughly six percent off the price. Uninsured deposits: balances above the $250,000 FDIC cap. Those three facts sat on one balance sheet.

Domain: a state member bank whose assets are mostly securities, whose deposits are mostly over the insurance cap, and whose capital ratios treat HTM securities as if they will be held. The comparison class is any bank that can call a duration hole "unrealized" while the funding that would have to wait for maturity can leave in a day.

If that reading is right, a common equity tier 1 ratio would not count as capital if marking the HTM book to market would wipe it. An HTM label would not count as a control if selling one bond from that book reclassifies the rest. A $250,000 insurance cap would not count as a run-brake if 94 percent of deposits sit above it. A Sunday systemic-risk exception — the Treasury Secretary letting the FDIC protect uninsured deposits — and a Fed facility that lends against those same securities at face value would be the public record of that knowledge, not a recap of an essay.

Ostensive specimen: Board of Governors of the Federal Reserve System, Review of the Federal Reserve's Supervision and Regulation of Silicon Valley Bank, led by Vice Chair for Supervision Michael S. Barr, 28 April 2023. Silicon Valley Bank Financial Group (SVBFG) was a California state member bank holding company with about $212 billion in assets when it failed. The bank itself had about $209 billion. It tripled from $71 billion to over $211 billion between 2019 and 2021. Deposits were largely uninsured and were invested primarily in longer-term securities. As of year-end 2022, about 94 percent of SVBFG's deposits were uninsured; HTM securities were 78 percent of the securities book, with a weighted-average duration of 6.2 years, mostly agency mortgage- backed securities of ten years or more. On 8 March the firm announced it had sold $21 billion of AFS securities for a $1.8 billion after-tax loss and planned to raise $2.25 billion of capital. On 9 March deposit outflows were over $40 billion; management expected $100 billion more the next day, roughly 85 percent of the deposit base. The California Department of Financial Protection and Innovation closed the bank on 10 March and appointed the FDIC as receiver. Four takeaways: the board and management failed to manage the risks; supervisors did not fully appreciate the vulnerabilities as the firm grew; when they did identify them they did not force a fix fast enough; the Board's 2019 tailoring of enhanced prudential standards, after the Economic Growth, Regulatory Relief, and Consumer Protection Act, reduced the requirements that would have applied. At failure the bank had 31 unaddressed safe-and-soundness warnings. Supervisors had named interest-rate-risk deficiencies in the 2020, 2021, and 2022 CAMELS exams and did not issue a finding until November 2022. The firm did not test discount-window borrowing in 2022. https://www.federalreserve.gov/publications/review-of-the-federal-reserves-supervision-and-regulation-of-silicon-valley-bank.htm Key takeaways: https://www.federalreserve.gov/publications/2023-April-SVB-Key-Takeaways.htm Evolution (HTM, duration, 94 percent): https://www.federalreserve.gov/publications/2023-April-SVB-Evolution-of-Silicon-Valley-Bank.htm PDF: https://www.federalreserve.gov/publications/files/svb-review-20230428.pdf Press, 28 April 2023: https://www.federalreserve.gov/newsevents/pressreleases/bcreg20230428a.htm

What the 10-K already named, not recap. SVB Financial Group, Form 10-K for the year ended 31 December 2022, filed 24 February 2023. HTM securities: amortized cost $91.327 billion, unrealized losses $15.160 billion, fair value $76.169 billion. In 2021 the firm had re-designated $8.8 billion of AFS securities into HTM. HTM is carried at amortized cost; changes in value, other than a credit-loss allowance, are not reported on the financial statements. https://www.sec.gov/Archives/edgar/data/719739/000071973923000021/sivb-20221231.htm

The 8-K that made the AFS hole public. SVB Financial Group, Form 8-K, date of earliest event 8 March 2023: sale of substantially all of the AFS portfolio, after-tax loss of about $1.8 billion, proposed offerings totaling $1.75 billion plus a $500 million private subscription, $2.25 billion in all. https://www.sec.gov/Archives/edgar/data/719739/000119312523064680/d430920d8k.htm A later 8-K, 14 March, after receivership: the sale was a portfolio with book value of about $23.97 billion for net proceeds of about $21.45 billion, to Goldman Sachs & Co. LLC. https://www.sec.gov/Archives/edgar/data/719739/000119312523070254/d487554d8k.htm

The inspector general on the same bank. Office of Inspector General, Board of Governors of the Federal Reserve System, Material Loss Review of Silicon Valley Bank, Board Report 2023-SR-B-013, 25 September 2023. Estimated cost to the Deposit Insurance Fund $16.1 billion. HTM unrealized losses rose from about $1.3 billion at year-end 2021 to about $15.2 billion at year-end 2022; AFS from about $313 million to about $2.5 billion. A $40 billion run, then $100 billion of further requests the bank could not meet. Examiners did not closely scrutinize the interest-rate risk in the securities book. https://oig.federalreserve.gov/reports/board-material-loss-review-silicon-valley-bank-sep2023.htm PDF: https://oig.federalreserve.gov/reports/board-material-loss-review-silicon-valley-bank-sep2023.pdf

The Friday and Sunday responses. FDIC, 10 March 2023: CDFPI closed the bank; FDIC as receiver created the Deposit Insurance National Bank of Santa Clara and transferred insured deposits. Uninsured depositors would get an advance dividend and a receivership certificate. Assets about $209.0 billion, deposits about $175.4 billion, as of 31 December 2022. https://www.fdic.gov/news/press-releases/2023/pr23016.html Joint statement, Treasury, Federal Reserve, and FDIC, 12 March 2023: after a recommendation from the FDIC and Fed boards and consultation with the President, Secretary Yellen approved a systemic risk exception so the FDIC could protect all depositors, insured and uninsured. Shareholders and certain unsecured debt holders were not protected. Losses to the Deposit Insurance Fund to be recovered by a special assessment on banks. A similar exception for Signature Bank, closed the same day. https://www.fdic.gov/news/press-releases/2023/pr23017.html FDIC, 13 March 2023: all deposits and substantially all assets moved to Silicon Valley Bridge Bank, N.A. https://www.fdic.gov/news/press-releases/2023/pr23019.html Federal Reserve, 12 March 2023: Bank Term Funding Program. Loans of up to one year against Treasuries, agency debt, and agency mortgage- backed securities, valued at par — face value, not the market price. Treasury Exchange Stabilization Fund backstop of up to $25 billion. https://www.federalreserve.gov/newsevents/pressreleases/monetary20230312a.htm Term sheet: https://www.federalreserve.gov/newsevents/pressreleases/files/monetary20230312a1.pdf

GAO on the same weekend. U.S. Government Accountability Office, GAO-23-106736, Bank Regulation: Preliminary Review of Agency Actions Related to March 2023 Bank Failures, 28 April 2023. Silicon Valley Bank assets grew 198 percent in 2019– 2021 against 33 percent for a peer median. The San Francisco Reserve Bank rated the bank satisfactory until its first large-bank rating in 2022, downgraded in June, began an enforcement action in August, and did not finish it before failure. https://www.gao.gov/products/gao-23-106736

What tailoring already took off the table. Barr review, regulation chapter: as a Category IV firm, SVBFG was not subject to the liquidity coverage ratio or the net stable funding ratio unless weighted short-term wholesale funding hit $50 billion; it crossed that line in December 2022 and would have faced a 70 percent calibration in the fourth quarter of 2023. Category IV firms may elect not to reflect accumulated other comprehensive income — the running total of unrealized AFS gains and losses — in regulatory capital. HTM losses are not in that total at all. The first supervisory stress test would have been in 2024. Barr's own sentence: higher requirements "may not have prevented the firm's failure" but "would likely have bolstered the resilience." https://www.federalreserve.gov/publications/2023-April-SVB-Federal-Reserve-Regulation.htm

This post is the public case, not a recap of an essay. One related diagnostic, not the object: https://kunnas.com/articles/the-statistic-was-still-known-internally

footnote_htmcollapsed

The public record already names the objects. You do not need a theory of anyone's character to see them.

Barr review, 28 April 2023. HTM duration 6.2 years at year-end 2022, mostly agency mortgage- backed securities of ten years or more. Uninsured deposits 94 percent, against 41 percent for large-bank peers. CET1 12 percent, two points above that peer average. $21 billion AFS sold, $1.8 billion after-tax loss, $40 billion out on 9 March, $100 billion expected on the 10th. 31 unaddressed warnings. Interest-rate risk named in three CAMELS exams before a finding. PDF: https://www.federalreserve.gov/publications/files/svb-review-20230428.pdf

Form 10-K, 24 February 2023. HTM amortized cost $91.327 billion, unrealized losses $15.160 billion, fair value $76.169 billion. The fair-value column is in the filing. https://www.sec.gov/Archives/edgar/data/719739/000071973923000021/sivb-20221231.htm

Fed OIG, 25 September 2023. Deposit Insurance Fund about $16.1 billion. HTM unrealized losses about $15.2 billion at year-end 2022, up from about $1.3 billion a year earlier. https://oig.federalreserve.gov/reports/board-material-loss-review-silicon-valley-bank-sep2023.htm

If you open one URL besides the post, open the Barr PDF, then the 10-K HTM table, then the 12 March joint statement.

if_alcocollapsed

Hypothetical, labelled as such. You sit on the asset-liability committee of a state member bank whose clients are venture-backed firms. In 2021 they deposit tens of billions they have not yet spent. Overnight rates are near zero. The securities desk can buy ten-year agency mortgage- backed paper at a yield above cash. Accounting will let you classify it held-to-maturity if you assert intent and ability to hold. Selling any of that book later would mark the rest to market. The deposits are almost all over $250,000.

What has to be true, tonight, for buying that paper as HTM to be the honest move? Either those deposits cannot leave faster than the bonds pay down — so "held to maturity" is a fact about the funding, not a sentence in a policy — or capital already counts the market price, so a rate rise cannot hide in a footnote. If both are missing, you are in the shape the post names: the HTM label still exists, the duration is still 6.2 years, and 94 percent of the deposits can still file a withdrawal. The practical test is those two sentences, not a seminar about whether the committee meant to be careful.

three_accounts2 comments

Four accounts, and they point at different first rules.

One account says this was an accounting hole. The 10-K already showed $15.2 billion of HTM unrealized losses against roughly that much common equity. If that is right, the first repair is that HTM losses count in capital, or that a bank funded by uninsured deposits cannot use HTM for long-duration paper. That predicts a later bank can still run if deposits are uninsured, so long as this footnote is closed. It does not, by itself, stop a coordinated withdrawal.

A second account says this was runnable funding. Barr: 94 percent uninsured, a concentrated venture-capital network, $40 billion in a day. If that is right, the first repair is that a deposit over the cap cannot fund a six-year bond. That predicts a perfectly marked securities book can still fail if the same clients leave together.

A third account, the tailoring account in Barr's fourth takeaway, says Category IV did not face the liquidity coverage ratio, did not have to put AOCI into capital, and would not see a supervisory stress test until 2024. If that is right, the first repair is to put those rules on a $100 billion firm when it crosses the line, not years later. That predicts a well-managed HTM book under 2019 tailoring still looks like SVB on paper.

A fourth account says supervisors already knew. IRR in 2020, 2021, and 2022 CAMELS; six liquidity findings in late 2021; management downgraded to "fair" in 2022; an enforcement action begun in August and not finished. GAO-23-106736 is that record. If that is right, the first repair is that a named interest-rate-risk deficiency becomes an enforceable order in the same exam cycle. That predicts a later Category IV bank can still carry HTM against uninsured deposits, provided someone can force a sale before the 8-K.

They differ on the first rule you would write. If the first, you change the HTM filter. If the second, you can still have HTM, provided the deposits cannot leave in a day. If the third, the other two are downstream of a $250 billion threshold. If the fourth, the rules already on the books would have been enough had anyone used them in time.

grant_managementcollapsed

Two concessions, then what is left.

First: Barr's takeaway 1 is not a press line. The board did not get a usable picture of liquidity until November 2022. Compensation through 2022 was tied to short-term earnings and equity returns and did not include risk metrics. The firm removed interest-rate hedges that would have protected against rising rates, then changed deposit-duration assumptions to clear a limit it had been breaching since 2017. Grant that. A thread that talks as if SVB were only a tailoring victim, with no management failure, is reading a different review than the one Barr signed.

Second: they did try to restructure. The 8 March 8-K is a completed AFS sale and a planned capital raise, not a bank ignoring the hole. Grant that they did not simply sit on the 10-K.

What remains is narrower. The HTM book was still at amortized cost after the sale. The uninsured share was still about 94 percent. The liquidity coverage ratio was still not in force. The leftover is whether the damage the post names is the footnote, the runnable deposits, the 2019 tailoring, or an IRR finding that took three CAMELS cycles to become a letter. The 8-K and the management failure do not pick.

first_penn_breakcollapsed

The analog people reach for is Continental Illinois in 1984: wholesale uninsured funding, an electronic run, the FDIC announcing that all depositors would be made whole.

FDIC, 1980–1989 timeline, still live. Continental, then the seventh-largest U.S. bank, had little retail business and relied on fed funds and brokered deposits. After Penn Square, foreign depositors ran. The FDIC publicly guaranteed that all depositors and other general creditors would suffer no loss. Cost to the FDIC: $1.1 billion. https://www.fdic.gov/history/1980-1989 Chapter in History of the Eighties: https://www.fdic.gov/bank/historical/history/235_258.pdf

The analog that matches the securities is earlier. Same FDIC page, 28 April 1980: First Pennsylvania Bank, then the 23rd-largest, had bought large quantities of longer-term fixed-rate government securities in the late 1970s. Rates kept rising. Deposit rates exceeded what the securities paid. Open- bank assistance followed. https://www.fdic.gov/history/1980-1989

The break is exact. Copying "guarantee the uninsured on Sunday" onto SVB copies Continental and copies the 12 March systemic-risk exception. Copying "do not fund long government and agency paper with deposits that can leave when the rate on those deposits moves" is the First Pennsylvania transfer that survives. A bank that can pass a "we also had a run" comparison while a 6.2-year HTM book is still funded by 94 percent uninsured deposits is still in the SVB shape.

three_checks2 comments

Those accounts unpack into checks you can put in front of a Category IV bank, and they do not substitute for each other.

1. Before a bank can classify long-duration agency securities as held-to-maturity, either capital already includes the fair-value hole or the deposits that fund the book cannot leave faster than the bonds pay down. The 10-K table is the check: amortized cost $91.3 billion, fair value $76.2 billion. A policy that says "we intend to hold" is not the check.

2. Before uninsured deposits can be treated as stable funding for those securities, a $40 billion day has to be inside the liquidity plan the bank actually ran, not the one it rewrote after it failed its own internal stress tests. Barr's 85 percent-of-deposits figure is the check. A contingency funding plan that was not tested at the discount window in 2022 is not.

3. Before a $100 billion firm can wait until 2024 for a supervisory stress test, the liquidity coverage ratio and AOCI-in-capital have to apply when it crosses the line, not after a multi-year transition. Barr's regulation chapter is the check. A CET1 ratio of 12 percent that ignores the HTM footnote is not.

(1) without (2) still lets the marked book run. (2) without (1) still lets a matched deposit base sit on a duration hole the bank cannot sell without reclassifying. (3) without the first two only moves the same 8-K to a firm that has more paperwork.

which_firstcollapsed

One question whose answer would change which of those you write first.

If the HTM book had been marked to market in 2022, and the uninsured share had still been 94 percent, would the 8 March 8-K still have been enough to take the deposits? Or, if those deposits had been insured or duration-matched, would leaving $15 billion of HTM losses in a footnote still have been enough?

If the first, the missing object is the funding, and you spend the next decade on uninsured concentration and on whether a $250,000 cap is a run-brake, not on the HTM filter. If the second, the missing object is the footnote: a stable deposit base still fails if the only bonds you can sell without tainting the rest are the AFS stub, and the 10-K is how you stop this one. Barr, the 10-K, the OIG review, and GAO-23-106736 already record the HTM hole, the uninsured share, the tailoring, and the unissued enforcement action. They do not say which of those, repaired alone, would have kept a duration hole in a footnote from becoming a $16 billion hit to the Deposit Insurance Fund.

pick_up_btfpcollapsed

The documents a Category IV bank can actually pick up are already public. They are not the same repair.

The Barr PDF is still the supervisory chronology and the tailoring table. The 10-K HTM table is still the fair-value column. The Bank Term Funding Program term sheet is still "collateral valuation will be par value" against Treasuries and agency mortgage-backed securities owned as of 12 March 2023, advances up to one year. That facility was announced after CDFPI had already closed the bank. It is a way to avoid selling the securities. It is not a way to have avoided buying them as HTM against uninsured deposits. https://www.federalreserve.gov/newsevents/pressreleases/files/monetary20230312a1.pdf Joint statement: https://www.fdic.gov/news/press-releases/2023/pr23017.html

A bank that files the 10-K, and still funds a 6.2-year HTM book with 94 percent uninsured deposits, has picked up the disclosure and left the combination on the table. The discriminator is the same as in the post: does capital see the fair-value hole, can those deposits leave in a day, and did the par facility exist before the 8-K. The special assessment does not answer that. The 10-K, on the night it was filed, named the hole.