Some thoughts on the failure of Silicon Valley Bank

The tech and finance worlds have been consumed for the last four days by the collapse and closure of Silicon Valley Bank (SVB), a bank serving… Silicon Valley. I’ve been following the developments closely and wanted to organize some my thoughts into a short commentary piece.

What is/was SVB?

SVB is pretty much a normal bank. The only real difference between them and any other bank is the nature of their customers—they mostly serve tech startups, the venture capital community, and their employees. So not only are they geographically concentrated with a high number of clients in California, but they also have a big industry concentration. So they are sensitive to shocks in the technology industry in a way that a regular bank that serves a bunch of different industries and types of people is not.

What happened to them?

SVB’s banking relationships were strongly concentrated in the tech and crypto industries, both of which experienced big bubbles leading up to 2021. As firms in these spaces raised tons of cash towards the tail end of the bubble, they turned around and deposited it at SVB (SVB claimed to serve 44% of U.S. venture-backed startups in one way or another). According to the WSJ, SVB’s deposits climbed 86% in 2021 to $189 billion and capped out at $198 billion in 2022.

When a bank has deposits (liabilities), it has to hold assets against those liabilities. A bank paying you 2% in your savings account can only afford that if it holds some asset that pays it more than 2%. Traditionally, the main asset for banks was making loans—they borrow from you at 2% through your savings account, for example, and turn around and lend the same amount to someone else for their mortgage at 5%. That 3% difference is the bank’s “net interest margin,” which is the main way that banks make money.

Silicon Valley Bank’s deposits grew so fast that they basically couldn’t find enough productive loans to make to keep up with their deposit growth. Instead, they took that money and bought various types of bonds, such as U.S. government debt and mortgage-backed securities. It’s normal (actually it’s required) for banks to hold at least some of their assets in these types of high-quality securities, but what was unusual about SVB was that they held nearly 50% of their assets in this form. That’s a really high number—for most banks this number will be less than 20%.

Bond prices are sensitive to changes in interest rates. When interest rates go up, bond prices go down—it’s a mathematical relationship. The Fed started raising interest rates in March of 2022. Based on how much interest rates have gone up since then, I would expect a typical 10-year Treasury bond that was issued in February 2022 to be down 15-20% in value today. This is a gross oversimplification of the math, but to just explain the concept, if SVB bought $100 billion of 10-year U.S. Treasuries in February of 2022 when their deposit base hit its all-time peak, I would expect those Treasuries to be down $15-20 billion in value today.

Since banks are required to publish information about their holdings every quarter, analysts who were paying attention to SVB knew full well that they had significant losses on the asset side of their balance sheet due to rising rates. (When a bank’s liabilities are greater than its assets, it is insolvent.) There were people tweeting in November 2022 about risks in SVB’s portfolio. Those concerns came to a head this past week when Peter Thiel advised his portfolio companies to withdraw funds from SVB, precipitating a bank run.

The fastest bank run in human history

Overall, what happened at SVB was a classic, textbook bank run. But what struck me the most about this whole episode was the speed at which it happened.

A typical bank has lots of retail depositors that keep relatively small amounts of money in their bank accounts—certainly less than the $250,000 FDIC insurance limit for the vast majority of accounts. SVB was different, in that it had a relatively small number of accounts for the size of its deposits. This was because many of its customers were startups, who might have $80 million sitting in a single account.

Retail depositors have little to fear from a bank run if their deposits are under the FDIC limit. But startup founders have a lot to lose—I’ve seen numbers saying that 93% of SVB’s deposits were uninsured. And startup founders are all connected to each other over Slack/Whatsapp/Telegram/Signal—after Peter Thiel warned his portfolio companies about SVB, all of those startup founders probably texted their friends, and that was a straight up bank run. Apparently $42 billion of deposits were withdrawn from SVB in two days—almost a billion dollars an hour. Compare that to me warning my family about risks at Bank of America over our family Whatsapp group. Rather than opening new accounts at Mercury within two hours, my relatives might ask me about that message two weeks later, if they even noticed it in between the deluge of Internet forwards spam and videos of six-year-olds at karate competitions.

Said another way, having $1 million of deposits scattered over 1000 accounts with $1000 each is much less risky to the bank than having that same $1 million in a single person or company’s account.

And I don’t think this is the last lightning-speed bank run we’ll see. If startups can create new accounts at Mercury in two hours and deposit $100 million, they can also remove that $100 million in even less time. I’ll even extrapolate this trend of high-connectivity, rapid information spread, and perhaps even automated decision-making… right now you still need a human to withdraw funds from a bank, or press the red button that launches the nukes. But what does a world where ChatGPT makes these decisions look like?

Shouldn’t experienced bankers have known about the risk of rising rates to a long-term bond portfolio?

Yes. This part is really baffling to me. I can’t think of a good reason why a bank would have so much of its asset portfolio dedicated to long-duration securities after the Fed signaled an aggressive rate hiking path starting in March 2022. Well, other than incompetence, of course.

Would SVB have been in trouble even if they managed their bond portfolio correctly?

I think so. Nobody is really talking about the other half of SVB’s asset portfolio—their loan book. This part of the portfolio consists of consumer loans, like mortgages, to people who work in the startup community, and corporate lending, also to startups. Given that the tech sector was already in the process of blowing up, I have to think that all of these more traditional loans would also be at risk. If startups close down because they can’t reach breakeven or get enough traction, they obviously won’t be able to pay back their loans. And if they lay people off, some of those people may not be able to pay their mortgages. (Not only that, but I read that some of their personal loans were collateralized by stock options. Obviously, if a startup folds, stock options in it are worth nothing…) It’s the same as if a regional bank in Houston had heavy exposure to the oil & gas sector—if oil prices crash, I would expect much higher than normal loan losses in the Houston bank’s portfolio.

Reflexivity in SVB’s lending portfolio

A related point—SVB’s failure itself makes the non-bond portion of its asset portfolio worse. If a large number of startups have to furlough workers because their funds are frozen at SVB, or if they lose part of the funds entirely, that means they are more likely to shut down, and any loans SVB made to them obviously won’t pay out, employees can’t pay their mortgages, etc.

What will happen now?

The Fed and FDIC have been working this weekend to find a buyer to take over the remains of SVB’s business. I saw rumors on Twitter this morning that uninsured depositors will have access to 50% of heir funds by Tuesday and hopefully the rest of their funds over the next 3-6 months as the asset portfolio is gradually wound down.

I’m not so optimistic. I do think depositors will ultimately recover most of their funds, but I don’t think it’ll be 95%+ recovery like some commentators are saying. I think people aren’t recognizing how weak recovery in the venture debt portion of SVB’s portfolio is likely to be. SVB had $74 billion of loans outstanding in Q4 2022, with only $636 million of loan losses booked against them. That number is way too low. Debt isn’t a great way to fund startups, given how uncertain their cash flows and performance are. I think the next 18 months will show why debt is used so infrequently to finance startups.

Who is likely to be a buyer?

I’ve seen people suggesting that Goldman Sachs[1]or Morgan Stanley might be good potential buyers, as the franchise value of the SVB customer base might be attractive to their investment banking business. Maybe. If I was a consultant advising them, I would want them to think very carefully about whether or not they actually want to assume $74 billion of deposit liabilities in exchange for that $74 billion of loan assets (I don’t think those loans are worth $74 billion.)

What about contagion to other regional and community banks?

I think this concern is overblown. I’d imagine that very few banks have the kind of depositor base and distribution that SVB did. Regional and community banks tend to serve small businesses and retail depositors in their area. These types of deposits tend to be very sticky and the vast majority of them will be fully protected by the FDIC insurance limit.

There might be some small banks that mismanaged interest rate risk and as a result could be close to insolvency, but I don’t think their depositors will face anywhere close to the level of the prisoner’s dilemma that SVB’s customers had. They’re just too small.

What does all this mean for the tech sector?

My sense is that we are still in the middle innings of the tech collapse. These doomsday scenarios about how no startups will be able to make payroll next week because their funds are frozen at SVB are indeed scary, but I think the bigger effects are still to be felt. VC firms have yet to write down the book values of their investments in startups that will never reach profitability and SVB’s successor will ultimately have to recognize much more than $636 million of losses in its venture debt book.

I remember when I first moved to Austin and learned that there were something like 30 startup incubator/accelerator programs at UT, I was shocked and felt that 28 of them would be gone in five years. I think the same will be true of a bunch of weak venture capital firms.

Which other sectors are at risk?

I think this is just one example of how a rising rate environment affects industries that relied on money being free! We are going to see more. Startups with bad business models were able to gobble up huge amounts of money when interest rates were at 0%. Commercial real estate is another sector that is sensitive to interest rates and traditionally tends to be pro-cyclical. Some subsectors in CRE are enjoying strong tailwinds (industrial, healthcare, data) but office CRE is facing an extremely strong headwind due to work-from-home. I also wonder what kind of financing cliffs private equity firms will be facing in the next two or three years. Large firms with very large debt loads, like AT&T, have also been struggling.

[1]Goldman has already made a small foray into retail banking through their Marcus savings product. Both firms do wealth management already.

Subscribe to Negative Convexity

Don’t miss out on the latest issues. Sign up now to get access to the library of members-only issues.