EP182 · Economy · first published 2023-03-16
The SVB Banking Crisis Step by Step | Jyri Engeström | Negotiator 182
Yes VC's Jyri Engeström narrates the collapse of Silicon Valley Bank day by day from the inside — the episode was recorded on 16 March 2023, a week after it happened. This was not an ordinary bank run: the balance sheet held not junk but government paper bought when rates were low, and the money left in seconds because more than 90 per cent of customers held deposits above the $250,000 insurance limit. Engeström also says he advised his own portfolio to move its money out, and calls that a mistake in hindsight. The sharpest observation in the episode is that the trigger was not Peter Thiel but Monday's payroll: in the United States it is safer to lay people off than to pay them late. It closes on the reversal in rates, the unsolved problem of deposit insurance, and the AI year that GPT-4 had just begun. Published 16 March 2023.
The SVB Banking Crisis Step by Step | Jyri Engeström
Summary: In episode 182 Sami Miettinen interviews Jyri Engeström of Yes VC about the collapse of Silicon Valley Bank. The episode was recorded on 16 March 2023, eight days into the crisis, and it is a first-hand account of what happened in Silicon Valley rather than a retrospective analysis. Published 16 March 2023.
Why this was not an ordinary bank run
Miettinen draws the distinction up front, and it is worth reading carefully, because it inverts the usual story.
An ordinary bank run goes like this: the bank has made risky investments on the asset side, losses appear, depositors see the hole and leave.
SVB went the other way round. There was no junk on the balance sheet. There were near-riskless government securities and property-backed loans bought when rates were low, which now had to be sold at a loss for liquidity. Wednesday’s update disclosed the losses and a small share issue to cover them, and that was the spark.
What made it exceptionally fast was the customer base. More than 90 per cent of customers held deposits above the $250,000 insurance limit, because they were essentially all venture-backed startups.
This is probably the fastest bank run in world history. It happened terribly fast.
Engeström’s own part, which he tells himself
This is the most honest passage in the episode, and it deserves highlighting because the guest does not protect himself.
Asked whether he was in “Peter Thiel’s early-warning ring”, Engeström first corrects the fact: in his experience there was no actual ring, and Peter Thiel was not personally involved. It was a few finance people at Founders Fund — the organisation Thiel originally founded but which, in Engeström’s words, seems to do quite well without him these days.
The message was that Founders Fund portfolio companies had been given guidance to reduce their deposits at SVB.
And then he says what he did himself:
In hindsight I made a mistake in advising our portfolio to pull out of that bank. It was in a sense obvious that if everyone does this, the bank falls.
He judges that those who started it did not really understand what they were setting off. He himself, as late as Wednesday, advised people to move only a couple of months of payroll to a reserve account, because he thought it impossible the situation would escalate to the bank actually failing — and he was wrong.
Why the venture fund itself was exposed
Engeström gives a thirty-second lesson here on how a fund works, and it explains why the crisis landed at the worst possible moment.
A fund does not normally hold cash. Capital is called from its investors only when an investment is made, and otherwise the account should be near zero. Yes VC had called several million for three imminent investments. One closed; two slipped, as they do in last-minute negotiations.
Suddenly our LP money is at risk, which should never normally happen.
The structural fix he describes discussing with his own investors: borrow, make the investment with the borrowed money, and call capital only once the investment is done, so the cash sits on the account only briefly. As a by-product it improves the fund’s IRR. He adds drily that a fund is not a cash management shop trying to earn yield on cash: that is not our job.
Miettinen notes that private equity did comparable IRR optimisation with short loans during the zero-rate era, and that the practice disappeared once rates rose.
What actually triggered the panic: payday
This is the episode’s most important correction to the public account, and Engeström offers it explicitly as a correction.
There was a lot of talk that this was Peter Thiel’s fault, that he started the bank run. I would just like to correct that a little — what mattered more was that this Monday was payday in America.
The bank failed on Friday — the day before payroll. Tens of thousands of employees were due to be paid on Monday, and companies faced millions in outgoings.
And then the mechanism that makes this worse than it sounds:
In the US, if you do not pay wages as agreed, you create a risk that employees can sue you. So it is safer to lay people off than to leave wages unpaid.
From that followed what Engeström calls a perverse incentive: companies simultaneously started drawing up redundancy plans and hunting for emergency loans. By Friday evening it did not look at all likely that anything beyond the insured $250,000 would be available by Monday — nowhere near enough for payroll.
He describes being asked by founders whether he had personal money available for their payroll. Investors ended up committing to lend their own money to companies without being able to be sure of getting it back.
Miettinen’s conclusion is the one to remember when the episode is later described as an investor bailout:
This was not a bailout for venture capital’s privileged — this was also personal wages and the future of the companies.
What the bank meant to Silicon Valley
This explains why the reaction was so strong, and it is not nostalgia.
Silicon Valley Bank had operated for 40 years and was, in Engeström’s account, practically the only bank that understood how to lend to startups. About half of all venture-backed startups banked there, and essentially every venture firm. Many founders and investors also held personal deposits at the same bank — at worst all three concentrated in one place.
It was also a leading venture debt lender. Engeström says hundreds of startups and dozens of funds owe their existence to it.
The contrast with the alternatives is concrete and slightly comic. The big banks do not integrate:
JPM, for instance — I still have not been able to open an account. And they will not even take venture investors.
Yes VC’s systems run on top of Carta, which has ready integrations to banks like SVB; he names First Republic and Mercury as the others that integrate at software level. He estimates the same fund administration takes one and a half people in the US where in Europe it could mean ten times the headcount, because less is automated. (Vauban, which Carta acquired, builds the European equivalent.)
This also explains the outcome: more than 600 venture firms signed a petition within a day supporting SVB’s continuation and saying they would keep their deposits there.
The reversal nobody expected
The timeline is the most dramatic part of the episode and is worth giving as it stands.
- Wednesday, everything looked normal.
- Thursday, the guidance spread and money began to move. Most transfers initiated after Thursday got stuck and never went through.
- Friday, the bank failed.
- Over the weekend, crisis meetings. Engeström’s ski trip to Switzerland was cancelled.
- Sunday evening a rumour, Monday confirmation: deposits guaranteed without limit.
And then everything flipped:
This was the only bank that was now actually safe.
Yes VC cancelled its own transfers. Engeström’s description is precise: let’s hold your horses and see where we are, so we don’t jump out of the frying pan into the fire.
The bank got new leadership — a former Fannie Mae chief executive, who introduced himself to venture investors on a Zoom call. Engeström mentions in passing that he was not even sure of the new entity’s name, which conveys how fresh all of it was.
He is appreciative of the speed of the authorities: they simply turned up within hours and took charge, even if the communication could have been better.
Miettinen’s forecast for the bank is more sober: he expects it to be zombified — the deposit insurance fund has no interest in running a competitor, so it becomes a piggy bank from which money is returned. Engeström reports the opposite from his own sources: SVB will probably continue and be sold, and there are several candidate buyers — he mentions General Catalyst’s effort to assemble a consortium, and General Atlantic, which had nearly invested in SVB on the Wednesday of the crisis week. The disagreement is left standing; neither of them knew.
Signature Bank and First Republic
Miettinen raises the other failed bank, Signature Bank, which had positioned itself around lawyers and had some crypto exposure. Engeström’s answer is revealing: “Signature Bank is completely new to me.” The US has an enormous number of small specialised banks — “probably fewer after this.”
Through that week nobody knew whether First Republic would be next; Engeström says it had to take on around 70 billion in additional financing.
Why it happened: rate rises and a two-sided squeeze
Engeström’s own analysis of the root cause has two parts and is the most analytical passage here.
First the balance sheet grew too fast. SVB’s deposits had risen in a couple of years from about 50 billion to over 190 billion — Miettinen gives the scale by comparing Nordea’s balance sheet of over 600 billion. The money had to go somewhere, and it went into low-risk paper at low yields.
Then the flow reversed. As rates rose, venture investment fell sharply. Startups stopped receiving new money into their accounts and started eating their existing deposits as growth slowed. The deposit base turned down at exactly the moment the securities portfolio was under water from rate rises.
They had not prepared at all for that combined effect.
From which follows his actual question, structural rather than accusatory: how can the system account for the cyclicality of the startup sector, which will in all likelihood recur.
The blame phase
The episode catches the political reaction in real time, and both sides get the same treatment.
First wave: this was Trump’s fault, because Dodd-Frank was rolled back for smaller banks.
Second wave: Tucker Carlson argued the cause was a bank full of “woke types”. Miettinen dispatches this: the board apparently had two black members and one woman, “so it’s a woke bank then.”
Engeström’s reading is that Republicans have an interest in framing it as a Democratic failure and the other side as a problem of Californian money — and that venture investors are also pointing fingers at each other over who started the run. Lawsuits over management share sales are already filed; neither man rules out that someone ends up liable.
Venture debt explained
A useful detour for anyone thinking about growth financing.
Venture debt is lending against a won customer base and its predictable cash flow — in effect an intangible-world solution to an old banking problem in which the collateral is usually property or another tangible asset. Miettinen’s reason why the product is not wildly risky for the bank: borrowers are typically SaaS companies whose won customers are unlikely to churn and instead continue year after year.
Engeström balances the picture. Not everyone gets it, terms can be hard, and the bank can call the loan if it judges growth has stopped — which can push a company into insolvency. The appeal comes from founders already having been diluted and not wanting to take more VC money.
The unsolved problem of deposit insurance
The closing stretch is broader and applies to European readers too.
Miettinen states the dilemma sharply: you ought to be able to trust the system’s liquidity, and at the same time you can only rely on the bank for $250,000. Someone has to carry the excess risk — there is no free risk.
The practical American answer is a sweep account: when the balance passes the limit, the excess is automatically distributed to other banks at the end of the day. Several accounts at the same bank do not help — the protection is per bank.
Finland’s scale is smaller and the limit lower: €100,000, which does not cover any company’s monthly payroll. During the crisis Engeström ran an informal poll in the Finnish startup community’s WhatsApp group: 68 per cent used Nordea and only about 16 per cent had more than one bank. His point is that concentration in Finland is even higher than it was in Silicon Valley.
Miettinen puts the risk in proportion, though: Nordea’s balance sheet is roughly three times SVB’s, and the deposit base is of an entirely different character — small, mostly under €100,000, not a hyperconnected group that can move everything in minutes.
And the euro area’s scale makes an unlimited guarantee politically impossible: bank balance sheets total roughly €40 trillion, of which about €20 trillion is the same kind of deposit base. Unlimited cover would mean taxpayers carrying trillions of risk.
There simply isn’t the political capacity for that.
Engeström’s human objection: does everyone now have to become a bond expert? Crypto surfaces in the conversation with its decentralised-money argument, as does central bank digital currency — of which Miettinen observes drily that customer enthusiasm is running at about minus three per cent.
The turn in rates
A short but important macro observation. Rates had been raised decisively against inflation and government paper was being sold off the balance sheet. The failure of a single Silicon Valley bank reversed market rate expectations within hours: the consensus became that central banks would retreat and move to rescue the system, which in turn lifts asset prices through a lower discount rate.
What stays with you
Three things.
The speed was a property of the customers, not the bank. The same balance sheet would have lasted longer at any other bank. It fell in seconds because the depositors were a hyperconnected, always-online group used to deciding in minutes — and because almost all of them were above the insurance limit.
The trigger was payday, not a famous investor. That it is safer in America to make people redundant than to pay them late turned a liquidity problem into an immediate jobs crisis and made backing out impossible.
Deposit insurance was left unsolved. An unlimited guarantee creates moral hazard and is politically impossible in the euro area; a limited one forces every company to build its own diversification machinery. The episode does not resolve this, and that is honest — neither speaker claims to know the answer.
A note on when this was recorded
The episode was recorded eight days into the crisis, and it shows. Some of the judgements in it are forecasts written down before the facts were in — particularly whether SVB would continue and whether First Republic would fail. Its value lies precisely in being a snapshot from the inside rather than a tidied-up account written afterwards.