The beginning of the end of Silicon Valley Bank commenced on Friday, March 3, when the Goldman Sachs bond trading desk got an interesting inbound call: SVB, the venerable central nervous system of the tech industry, was seeking its assistance to negotiate the purchase, as a principal, of a $24 billion chunk of SVB’s portfolio of Treasury securities and mortgage-backed securities. The bonds were part of the investment portfolio that SVB had decided to sell in order to meet an unexpected rush of withdrawals from its uninsured depositors, such as venture-capital firms and their portfolio companies. 

Goldman is Goldman—the firm handles deals like this all the time, and well. The opportunity this time was fairly straightforward. Goldman would buy the portfolio at a significant discount and then sell the bonds over the course of the following weeks, providing SVB with liquidity while Goldman sought to arbitrage the spread. Working through the weekend, Goldman’s traders would have to value each of the bonds separately, taking into account their duration, their interest rates and their credit profile, among other things, including what Goldman thought each bond would fetch in the increasingly volatile market. 

But the SVB deal was slightly unusual in a couple of ways. Not only was the SVB bond portfolio huge, it was also complicated, comprising some 291 different CUSIPs, or individual bonds, at low interest rates, bought at the top of the market. Goldman divided the 291 bonds into seven different segments and analyzed them in these buckets. And then there was another difference, but one that may not have been so readily apparent on this Friday: serious time pressure.