On the ninth of September, Alan Lane, who ran Silvergate Bank through both its improbable rise and its deliberate dismantling, offered the market a sentence that should have unsettled anyone still paying attention. The bank was solvent when it chose to die. Not temporarily liquid, not rescued through a shotgun merger, but solvent, capitalized, and still wound down voluntarily by a board resolution in March 2023. His account places the cause upstream of the balance sheet, with the Biden administration, the Department of Justice, the Federal Reserve, the FDIC and the OCC, whose overlapping scrutiny he describes as a pressure campaign rather than an investigation. I have read enough founder post-mortems, and written a few inside my own projects, to recognize the genre. The person closest to the failure always holds the most detailed map of the fire, and the least incentive to describe the accelerant. What makes Silvergate worth revisiting now, in the middle of a bull market that has quietly rebuilt its dollar plumbing on foundations nobody has audited, is not the assignment of blame. It is that both explanations — the state killed it, the market killed it — can be true at once, and only one of them is portable to the next institution.
Silvergate was not born a crypto bank. It opened in 1988 as a small community lender in La Jolla, California, and spent a quarter century doing the unremarkable work of a regional deposit institution. The pivot began around 2013 and accelerated after 2017, when it built the Silvergate Exchange Network — a permissioned, twenty-four-hour settlement rail that let exchanges and institutional clients move dollars between accounts instantly, including on weekends, when the Federal Reserve's own rails are closed. SEN was not a blockchain. It was a bank ledger with an API and a service-level agreement, and that was precisely why it worked. By 2021 Silvergate held roughly $14.3 billion in digital-asset-related deposits and had become the connective tissue between the crypto industry and the US dollar system. Coinbase, Gemini, Kraken, Circle and a long tail of market makers and OTC desks ran operating balances through it.
The composition of those deposits is the entire story. They were demand deposits — payable on demand, by definition — and they came overwhelmingly from a single, tightly correlated industry. When FTX failed in November 2022, Silvergate's depositors did not evaluate Silvergate. They evaluated each other. In the weeks that followed, something close to seventy percent of the bank's demand deposits left the building. On March 1, 2023, Silvergate delayed its annual report, a filing delay the market read as an inability to certify its own survival. On March 8 it announced an orderly liquidation. Two days later Silicon Valley Bank was seized. Two days after that, Signature Bank was closed by New York regulators. Three banks, one week, and the first of them had built an entire franchise on the assumption that crypto's dollar problem was an engineering problem, when in fact it was a concentration problem.
Lane's defense rests on a specific claim: that Silvergate managed its balance sheet against volatility deliberately, holding high-quality liquid assets to meet deposit swings. On its face this is exactly what a prudent treasurer would do, and I have made a similar argument in my own work — when I helped refactor the consensus layer of a privacy payments startup in Berlin, we spent three months proving a design could hold under load before we let it touch real money, cutting gas costs by forty percent and still refusing to ship until the failure modes were enumerated. That discipline is why I want to be precise about what Silvergate's liquidity actually protected against.
High-quality liquid assets protect against a withdrawal you can sell into. They do not protect against a withdrawal that coincides with a repricing of the assets you are selling. Through 2022 and into early 2023, the Fed's rate cycle had driven the mark-to-market value of long-duration, high-credit holdings well below par across the entire banking system. This is the same arithmetic that destroyed Silicon Valley Bank. A portfolio that is money-good at maturity is not money-good at the moment of a run. Selling to meet redemptions converts an unrealized loss into a realized one; realized losses eat capital; and capital is what the market actually prices when it decides whether to keep its money inside your building. Silvergate's own disclosures described selling liquid assets to repay borrowings, which is the balance-sheet equivalent of a patient selling organs to survive one more quarter.
So when Lane says the bank was solvent, I believe him in the accounting sense. Solvency is a balance-sheet fact measured at a point in time. Survival is a flow variable measured across a week. The two numbers can diverge catastrophically, and the gap between them is not a technical failure — it is a narrative failure, and narrative failure is not fixable with a liquidity buffer. Truth is not what is seen, but what is trusted, and a trust run does not care about your capital ratio.
The second thing Silvergate's depositors did, and this is where the case becomes a template rather than an anecdote, is act as creditors and customers simultaneously. Crypto firms in late 2022 were not merely holding cash at Silvergate; many were counterparties to FTX, Alameda, Three Arrows, BlockFi, Celsius and Genesis. When one node in that network failed, every other node learned at the same instant that the safest move was to pull all liquid balances from any institution whose solvency was correlated with the rest of the network. This is a run driven by covariance, not by fundamentals.
Traditional bank-run models assume a mostly independent depositor base: the bakery, the dentist, the pension fund. Their withdrawals are loosely correlated because their reasons are loosely correlated. Crypto's deposit base at a bridge bank is the inverse — a single shock triggers everyone at once, because everyone's balance sheet is exposed to the same shock. Seventy percent of demand deposits left not because Silvergate was worse than its peers, but because Silvergate's depositors were indistinguishable from one another.
I learned a version of this in 2022, auditing twelve failed lending contracts while the market unwound, several of which I had written about favorably the cycle before. The common thread was not fraud, and mostly not incompetence. It was that every design assumed the rest of the system would keep behaving. They were all long the same invisible asset — the assumption of continuity — and when that asset went to zero, they went to zero together. Silvergate was that assumption, incorporated in La Jolla. Which is why I find the retrospective debate slightly beside the point. Whether the regulators squeezed or the depositors fled, the outcome was over-determined. A bank with a heavily correlated deposit base and a fractional reserve has one tail risk, and it is not a tail at all. It is a scheduled event waiting for a trigger.
There is an uncomfortable symmetry between Silvergate and the cross-chain bridges this industry has spent five years learning to distrust. Bridges have been drained of more than $2.5 billion cumulatively, and the industry's response was not to stop using them but to build second-order insurance, monitoring and layered multisig around them. We accepted that the connecting tissue is the fragile part, and we kept connecting. Silvergate was the fiat-side bridge: a single institution every major exchange and market maker needed in order to convert between the crypto economy and the dollar economy. It was not decentralized, not trust-minimized, not even particularly innovative — but it was necessary, and necessity is what made it a single point of failure.
The distinction that matters is between dependency and substitutability. A system can tolerate dependence on a fragile component if that component can be replaced within the tolerance window. Bridges can be replaced slowly and expensively. A bank can be replaced almost never, because opening a compliant dollar account for a crypto-native firm requires a willing counterparty on the other side of the table, and after 2023 the number of willing counterparties in the United States collapsed. Silvergate and Signature Bank together carried more than $300 billion in deposits at peak; when both vanished within four days of each other, the substitution that was supposed to happen did not happen. It migrated instead — to stablecoin rails, to offshore banking relationships in Switzerland, Singapore and Hong Kong, and to a small set of surviving US institutions that took on the business while quietly capping their exposure. That migration is the real legacy, and it is more consequential than the liquidation itself. The dollar rail for crypto did not disappear. It changed hands, jurisdictions and legal character, and every one of those changes was driven by risk-management instinct rather than design intent.
The most visible transmission channel in March 2023 was not a token price. It was a stablecoin. Circle held roughly $3.3 billion of the USDC reserve at Silicon Valley Bank and a smaller sum, approximately $33 million, stranded at Silvergate. When SVB was seized on March 10, USDC traded down to about $0.87 within hours. A dollar-denominated instrument that exists precisely to be worth exactly one dollar spent a weekend trading at eighty-seven cents — not because the reserves were insufficient in aggregate, but because the market could not verify in real time whether forty billion dollars of backing existed behind a promise that lived in a bank account nobody could see.
That episode is the cleanest demonstration of a principle I keep returning to in my work on decentralized identity. Verification latency is itself a risk. It is not enough for a guarantee to be true; it has to be independently checkable at the moment the market needs to check it. Circle's reserves were, in the end, sufficient. The panic lasted a weekend. But the panic was rational — under the information available, the market was correct to price uncertainty. This is the argument for tokenized treasury rails and on-chain reserve attestation, and it is also the argument against the current fashion of treating them as a solved problem. A tokenized money market fund is only as good as the chain it settles on and the custodian holding the underlying. We have moved the single point of failure, not eliminated it. The same instinct that made a bridge bank necessary in 2019 is now making a tokenized reserve rail necessary in 2026, and the industry has learned remarkably little about making the replacement institutionally diverse rather than merely technically novel.
Here is the detail that reframes the whole episode. On March 12, 2023, the Federal Reserve announced the Bank Term Funding Program, a facility allowing banks to borrow against the par value of their government securities rather than their depressed market value. It was, in effect, a purpose-built antidote to the exact mechanism that had just killed three institutions. Silvergate announced its liquidation on March 8. The bank chose to die four days before the government built the tool that would have let it survive. I raise this not to argue that Silvergate would have been saved — a facility does not restore a deposit base, and depositors were gone by then — but to locate the actual failure point in time. The decision was made under a set of rules that were about to change. Management did not know that, could not know that, and acted rationally on the information it had. That is what makes it a structural story rather than a moral one.
And why voluntary liquidation at all, rather than a sale? Because a sale requires a buyer willing to absorb a crypto-concentrated deposit book in a market where no such buyer existed and no regulator was going to bless one. The other two banks were seized; Silvergate was not, because it moved first. An orderly wind-down preserves some残余 franchise value, avoids receivership, and spares the board the spectacle of a takeover. It is the rational exit for an institution whose franchise is its counterparty relationships and whose counterparties have already left.
Which brings us to Operation Choke Point 2.0, the phrase Lane's account quietly feeds. The original Operation Choke Point, run out of the Department of Justice between 2013 and 2017, pressured banks to sever relationships with legal but politically disfavored industries. The 2.0 version, as the crypto industry describes it, is informal discouragement of bank-crypto relationships through supervisory pressure that never produces a public rule. That description is not obviously false. But the honest reading of Silvergate is narrower and more useful than the political one. Regulators did not need to declare Silvergate insolvent, because declaring is not how a run is triggered. A filing delay is enough. An unannounced examination is enough. A single leaked concern about deposit concentration, circulating in a market that had just watched FTX evaporate, is enough. The mechanics of a supervisory kill are not legal, they are informational: you degrade the confidence of counterparties until the balance sheet has to shrink faster than it can be managed down. Lane's claim that the bank was solvent and still chose liquidation is, in this framework, not a contradiction at all. It is a description of a specific and reproducible method.
I say that with a caveat I apply to every post-hoc narrative I read, including my own. Lane is a single source describing a decision he made, in a political environment where his framing serves one side of an argument. There are documents he does not cite and questions he does not answer — most notably why the annual report was delayed on March 1 if the bank's position was certifiable. A delayed 10-K is not a neutral event for a bank under scrutiny. To any counterparty it reads as a signal that the auditors have raised the going-concern question. If your liquidity strategy depends on counterparties believing in you, the worst possible thing you can do is tell them, in regulatory filing language, that you need more time. I spent the better part of a year at a Nordic fintech translating cryptographic guarantees into risk-management language for executives who had built careers pricing things they could see. The lesson was consistent: institutions do not buy cryptography, they buy narrative certainty. A bank that cannot certify itself in ninety days is not, in the market's eyes, a bank.
That lesson should be uncomfortable in 2026, because in a bull market nobody asks where the dollar rail lives. Deposits are flowing in, not out. Stablecoin supply is expanding, not contracting. Exchanges are profitable, tokenized treasury products are humming, and the industry has largely convinced itself that the 2023 banking crisis was a regulatory story with a happy ending — the happy ending being that crypto no longer needs banks. This is the most dangerous belief in the market right now. The industry did not solve the bridge-bank problem. It replaced a regulated, supervised, insured institution with arrangements that are less transparent, less insured, and concentrated in ways the market does not measure. A dollar of USDC is now a claim on a reserve portfolio custodied across a handful of institutions, some of which publish no supervisory ratings, some of which sit in jurisdictions with different insolvency regimes, and all of which remain exposed to the same rate cycle that broke Silvergate. An offshore account in Zurich is a fine solution for a fund. It is not a solution for an industry that wants to be a system.
The concentration risk did not disappear. It moved from a bank's deposit base into a stablecoin issuer's reserve composition, and from a single point of failure into a shorter list of points nobody is stress-testing publicly. Based on my audit experience with the protocol failures I catalogued in 2022, structural fragility does not announce itself. It looks like a healthy balance sheet right up until the moment it looks like nothing at all. The posture of this market is exactly the posture of those contracts in late 2021: everyone long continuity, nobody pricing the alternative. The regulators, meanwhile, have not become friendlier — they have become quieter. Quiet supervision is harder to detect, harder to litigate, and more effective at shaping institutional behavior. The mechanism Lane describes did not retire with the administration that administered it.
The counter-intuitive reading, and the one I think the industry has still not absorbed, is that Silvergate's collapse was not primarily a regulatory event. It was a business-model event, and the regulatory pressure was the accelerant rather than the fire. A bank whose deposits come from one industry is not a bank. It is a leveraged bet on that industry's liquidity, wearing a charter. That is true of a crypto bank, and it would be equally true of a lender whose deposits all came from commercial real estate developers in a single metro, or from one supply chain's finance arms. The 2023 crisis produced three such institutions in a week not because the government held a vendetta, but because the rate cycle and correlated deposit bases had already made them fragile. SVB's depositors were venture-backed startups with correlated burn rates. Signature's were crypto firms. Silvergate's were crypto firms. The common variable was covariance, and covariance is a business decision, not a policy outcome.
Where the industry's preferred narrative fails is in its prescription. If Silvergate was killed by overreach, the fix is political: elect better regulators, pass better legislation, assume the next cycle will be kinder. If Silvergate was killed by concentration, the fix is structural: diverse funding, verifiable reserves, redundant rails, and an honest accounting of which parts of the dollar system a crypto institution can actually replace. The second diagnosis is more expensive to act on, so the industry has chosen the first. I have watched the identical substitution happen in protocol governance. It is always easier to blame the hostile external actor than to redesign the mechanism that made you dependent on them in the first place.
The question for 2026 is not who killed Silvergate. It is where the next bridge bank is, and whether anyone is stress-testing it while deposits are still flowing in rather than out. Bull markets do not reward that question, which is exactly why it goes unasked until it is asked inside a liquidation announcement. Truth is not what is seen, but what is trusted — and in a market this confident, trust is the one asset nobody audits.
