Neuvottelija.com

Episode 182 · 2023-03-16 · 64:23 · Original in Finnish

The SVB Banking Crisis Step by Step | Jyri Engeström | Neuvottelija 182

Originally published as “SVB Pankkikriisi vaihe vaiheelta | Jyri Engeström | Neuvottelija 182”

Jyri Engeström of Yes VC narrates the collapse of Silicon Valley Bank from the inside, eight days after it began. This was not an ordinary bank run: the balance sheet held near-riskless government paper bought when rates were low, not junk, and the money left in seconds because more than ninety per cent of customers held deposits above the $250,000 insurance limit. Engeström says he advised his own portfolio to pull out and calls that a mistake, and corrects the public account of the trigger — it was not Peter Thiel but Monday's payroll, since in the United States it is safer to lay people off than to pay wages late. The episode covers what the bank meant to Silicon Valley as the only lender that understood startups, why the deposit base and the securities portfolio turned against each other at once, the unsolved problem of deposit insurance in both the US and Europe, and Finland's own concentration, where an informal poll found 68 per cent of startups banking with Nordea. It closes on the reversal in rate expectations and on the AI year GPT-4 had just begun.

Guest: Jyri Engeström · Host: Sami Miettinen

Watch and listen

Watch on YouTube JSON Markdown

Key moments

  1. 00:34 — Intro: Jyri Engeström of Yes VC, back for a second time
  2. 00:58 — Engeström's rollercoaster week as Silicon Valley Bank collapsed, 8-15 March 2023
  3. 04:16 — Why SVB was not an ordinary bank run, how bank risk forms, and Signature Bank falls
  4. 06:15 — How the week of horrors continued through three capital calls
  5. 11:51 — The $250,000 deposit insurance ceiling, and the trouble with removing it
  6. 13:00 — What if the crisis had happened in Finland? Nordea versus SVB
  7. 15:36 — Unpredictable depositors as systemic risk — even the Californian wine nearly went
  8. 17:19 — Dinosaur banks JPM, Citibank and others circling SVB's carcass
  9. 22:24 — Is SVB a good bank again after the bailout? New leadership, VC support, and a possible sale
  10. 26:31 — Rapid rate rises as root cause, and broken risk-taking incentives
  11. 29:09 — The structure, risks and benefits of venture debt
  12. 31:24 — Payday in the middle of the crisis — layoffs would have begun without a solution
  13. 34:57 — The political blame game: Trump's Dodd-Frank rollback versus Tucker Carlson's wokenomics
  14. 36:54 — Management share sales and criminal liability
  15. 40:28 — A systemic crisis of the monetary system — crypto or central bank digital currency as alternatives to deposits
  16. 43:20 — People want safe money: must everyone become a bond expert? The system is vast — over EUR 20 trillion in Europe
  17. 44:34 — Falling rate expectations as a side effect, and back to the starting point
  18. 46:34 — The AI revolution set off by GPT-4, and Yes VC's investment in Adept
  19. 55:39 — Engeström's and Miettinen's wild visions of neural-link-driven AI
  20. 57:30 — Is GPT creative, or an algorithm extrapolating web content?
  21. 58:56 — Will people hand over their lives, or at least copies of them, to AI?
  22. 01:01:43 — "Kahneman's juice": removing cognitive friction. Continuous flow, or AI dystopia?
  23. 01:03:25 — Outro

Summary

Jyri Engeström of Yes VC came on eight days into the collapse of Silicon Valley Bank to describe it from the inside. The episode is a snapshot rather than a retrospective, and several judgements in it are forecasts made before the outcome was known.

This was not an ordinary bank run. Normally a bank makes risky investments, losses appear, and depositors leave. SVB’s balance sheet held near-riskless government securities and property-backed loans bought when rates were low, which had to be sold at a loss for liquidity. Wednesday’s disclosure of those losses and a small share issue was the spark. What made it exceptionally fast was the customer base: more than ninety per cent held deposits above the $250,000 insurance limit, because they were essentially all venture-backed startups. Engeström calls it probably the fastest bank run in history.

He does not protect himself in the telling. Asked about “Peter Thiel’s early-warning ring”, he says there was no ring and Thiel was not personally involved — it was finance people at Founders Fund. And he volunteers his own part: he advised his portfolio to pull out, and in hindsight calls that a mistake, since it was obvious that if everyone did it the bank would fall. As late as Wednesday he thought escalation to actual failure was impossible, and was wrong.

The trigger was payday. This is his explicit correction to the public account. The bank failed on Friday, the day before Monday’s US payroll. In America, failing to pay wages on time exposes an employer to being sued, so it is safer to make people redundant than to pay late. Companies therefore began drawing up redundancy plans and hunting emergency loans at the same time. Founders asked Engeström whether he had personal money available for their payroll. Miettinen’s conclusion: this was not a bailout of venture capital’s privileged, it was also wages and the survival of the companies.

What the bank meant. SVB had run for forty years and was practically the only bank that understood how to lend to startups, banking about half of all venture-backed companies and essentially every venture firm, and leading in venture debt. The alternatives do not integrate — Engeström still could not open a JPM account, and the big banks will not take venture investors at all. His systems run on Carta, which integrates with SVB, First Republic and Mercury. More than 600 venture firms signed a petition within a day asking that SVB continue.

Why it happened. Deposits had grown from about 50 billion to over 190 billion in two years and had to be placed somewhere, so they went into low-yield paper. When rates rose, venture investment fell, startups stopped receiving new money and began eating their deposits, and the deposit base turned down exactly as the portfolio went under water. “They had not prepared at all for that combined effect.” His real question is structural: how should a system account for the cyclicality of the startup sector, which will recur.

Deposit insurance is left unsolved. You should be able to trust the system’s liquidity while only being able to rely on $250,000 of it; someone must carry the excess. The American workaround is a sweep account distributing balances across banks nightly. Finland’s limit is €100,000, which covers no company’s payroll, and Engeström’s informal poll of Finnish startups found 68 per cent banking with Nordea and only 16 per cent using more than one bank — more concentrated than Silicon Valley was. An unlimited euro-area guarantee would put trillions of risk on taxpayers, for which, as Miettinen puts it, there is no political capacity.

The two disagree, civilly and without resolution, on what happens to SVB next: Miettinen expects it to be zombified into a piggy bank, Engeström reports from his own sources that it will continue and be sold. The episode closes on how a single bank failure reversed market rate expectations within hours, and on the AI year GPT-4 had just begun.

Watch

The recording lives on the Neuvottelija channel: SVB Pankkikriisi vaihe vaiheelta | Jyri Engeström | Neuvottelija 182. A Finnish edition of this episode is published at www.neuvottelija.fi.

Go deeper

Guides connected to this conversation, with frameworks and further reading.

People and topics

Guests: Jyri Engeström

Topics: Investing & Markets Finnish Economy & Policy SaaS & Software

AI and agent resources


Source and content status

Provenance: Finnish source: Owner page assembled from YouTube metadata, the neuvottelija.fi episode record and a MacWhisper transcription of the Finnish audio. The chapter marks are the publisher's own list as released with the video, each verified to fall within ten seconds of a real transcript cue. No English caption track was produced, so no transcript is published here. The episode was recorded on 16 March 2023, eight days into the crisis, so several judgements in it are forecasts made before the outcome was known.. English subtitles: not available on this page; this is an episode summary, not a curated transcript. QA coverage 0% (transcript timecoded). Original episode: neuvottelija.fi. Imported 2026-09-11 · last reviewed 2026-09-11. Passages the source audio left genuinely ambiguous are marked [unclear] rather than guessed.