Blockchain vs. Human.

If you’d like to tell me that blockchain is really just a glorified database, you would be mostly right, except that it can remove human factor from a process and that is pretty big because the world works on relationships. Anyway, here is a fun story:

There was this nice churchgoing trader and philanthropist, who used to work in a very famous hedge fund in an industry where prestige is sort of a big thing, so when he called up banks to have a couple of meetings to open prime brokerage accounts with them for his new fund, things must have gone relatively smooth.

Archegos (the fund) then started making very very big bets. They entered into swap agreements with the banks to go long (bet the price will go up) on a small number of stocks (Viacom, Tencent, Discovery). If the price would go up, the banks would be paying Archegos and vice versa. These swaps with one of the banks (Credit Suisse) were static, meaning that the fund had to post collateral only once at the time they opened the position. If the stocks went up a lot and Archegos’ winning bet increased in value, they did not have to post more in collateral. *Say that the fund opened $1B long position on Viacom posting 20% collateral ($200M or just about Bill Hwang’s own money), but the price of Viacom went up and Archegos’ long position was now worth $2.5B. Their margin requirement didn’t change from the $200M. Archegos could have called up Credit Suisse and ask to pay them $1.5B of profits.

Here is where the fun part comes: they absolutely did call Credit Suisse and ask for the money and Credit Suisse paid, even though that at that point in March 2021, the bank knew very well that Archegos could blow up and kept asking them to move onto dynamic margining (so the fund would have to keep topping up its collateral each time their position increased) and to post more collateral in general.

For some time, this story was presented in media, as banks not seeing Archegos’ total exposure with all of them, when in fact, a report prepared by a law firm contracted by Credit Suisse shows that the banks have known very well and in detail what is happening, but they still didn’t make that call. Someone (and it would be someone from the sales, not from the risk department) was supposed to pick up the phone and tell the client to pay up. It would be an awkward and unpleasant call and the client would most likely take their business elsewhere? You are a good risk manager in a bank when you calculate that a very levered fund should pay up, but you are most likely not a good salesperson at the same bank, nor is it good for your future career and your relationships when your client goes away. Instead, they kept pinging each other emails; gentle requests for a call from Credit Suisse account managers, but the client was busy and promised to get back.

So when Archegos asked for their profit mid-March, Credit Suisse paid them $2.4B. At the end of that same month, when things blew up because Archegos positions kept plummeting, and Credit Suisse issued margin calls for $2.8B from Archegos, the fund said it is broke.

The system worked, everyone knew absolutely everything they should know to make a correct decision, but .. human factor. Seems messy when people interact; the calculation is not exact, it includes emotions and relationship biases.

Here is a thing and sorry it is stupid: you could say that blockchain really is removing this human factor because it is rewriting into code subject to consensus outside of human control everything we have seen above. I mean, a blockchain derivatives platform like this allows you to go long or short any asset in any size, with the exception that if you were wrong your margin call would be instant, automatic and there won’t be any relationships hurt or careers burned. It works something like: you put up a collateral in the native token of the platform. You can issue anything ..sorry I mean you can create out of a thin air any synthetic asset, e.g. a synthetic Tesla stock. Blockchain (uh ok, oracles, which are smart contract-based price discovery protocols) is connected to outside internet to keep checking the price of real Tesla. If it goes up, you can trade in the profit. If it drops, you are doomed.

It is stupid because you would absolutely NOT be able to do what Bill Hwang did above there since every DeFi platform asks for a collateral that is far exceeding the position you are creating. So for that platform above, it is 750%. Your collateral has to be 750%(!!!) to go long or short anything.

The sexy thing in the world of finance is leverage and masquerading that leverage to be something else. There is no leverage in DeFi. But there could be. You are interacting with a code that is non-discriminatory. On some platforms you can borrow and then plug the borrowed thing right back in as a collateral and borrow more. This creates a long on the borrowed thing that is multiple times levered. Slowly, like a giant money lego, these things are coming together.

I am also here and here. My writing is just a humble redux of Matt Levine’s column.