There’s a new paper out making the claim that Starlink, a satellite internet network mooted by SpaceX, will be hugely valuable for financial markets and thus make a fortune. The point is that high frequency trading becomes very much more profitable if it’s possible to trade over lower latency links. Starlink will be lower latency – therefore all the financial institutions will leap aboard and the network will be hugely profitable.
Well, yes and no really, yes and no. Latency is important, HFT exists and so on, but the relationship isn’t quite as the network topologist thinks.
The simulation predicts much faster round trips than over current networks, even faster than theoretical direct shortest route connection using fibre optics. Examples: 50ms round time trip from London-NewYork compared to theoretical 55ms from a direct connection, and 76ms that internet currently is capable of. This improvement is even greater for very long links.
OK, interesting enough. And yet:
As a result, it’ll bring in the $$ like you wouldn’t believe. Financial institutions in particular will pay through the nose for the fastest links, and the system will allow SpaceX a good amount of granularity and control to be able to set the bandwidth and charge accordingly.
Well, no, not really.
Much like electricity, the internet, and railroads went from kings of profit to marginal and tedious enterprises, it seems that HFT has gone from a nascent and fickle technology to a basic component of global infrastructure. As such, it is in no way, shape, or form “dead”. Just like step-change advancements in technology have forced service-based companies to upgrade or die, it seems that the availability of Starlink (or any comparable interconnected internet constellation) will create a massive imbalance between financial institutions that adopt early and those that do not or cannot.
If that ends up being the case, there will undoubtedly be an extraordinary surge in competitive financial infrastructure investment, with institutions desperately pursuing new ways to remain competitive (leveling the playing field vs. a shortcut to the front).
It’s the detail of the financial markets that matters.
The advantage of the Starlink system is low latency over long distances. It doesn’t change latency over short distances at all, has no advantage there. HFT s something currently conducted over short distances. Thus it’s of no great benefit, is it?
Sure, the short distance of HFT is partly to do with the existence of latency itself. So much so that exchanges can make good money by renting server space only metres from the hub of the exchange itself, rather than hundreds of metres away in someone’s trading room. Obviously, a system which generally reduces latency will eat into those location based profits. But they’re not large.
The structure of HFT is that any specific transaction is taking place only in the one place. On the one exchange. By design it’s all about what everyone else is doing in that place and at that time. Thus latency over long distances, as opposed to who gets to be metres instead of tens of metres, from the action doesn’t matter.
There is another idea entirely, arbitrage, where such distance does matter. Say, trading across the New York and London stock exchanges. Some stocks are listed in both places. Prices differ a little as changes are going to be based upon transactions in each place. There has long been arbitrage between the two. In the 1960s someone might actually phone up New York from London and if prices differ buy on one exchange, sell on the other. Thereby locking in the price difference and so also that profit. This is the essence of arbitrage, buying and selling the same thing in different places at the same time. This is different from our standard HFT, which doesn’t depend upon the differences in geography.
That reduction in latency would be of great interest to people who would wish to do such arbitrage across exchanges. The thing is, it wouldn’t be particularly profitable. It would start out that way, as HFT did. But as more people do it in greater volumes then the price differences across geography will be competed away and profits fall. Just what happened with HFT itself, that trading on the one exchange. Just as current internet technology has reduced those London and New York price differences. And the thing is, if you’re providing a service to people who are making little profit out of its use then you don’t get to charge them great and juicy prices.
Sue, Starlink would, on this basis, be economically valuable, as such arbitrage makes markets more efficient. Plus, more volume of trade means the buy/sell margin shrinks, to the benefit of all of the non-arbitrage members of the market. But that’s not the same as SpaceX being able to capture a significant portion of this value creation. They might well be able to gain good business from this possibility, but charging premium prices for it is much more unlikely.