It has been a few years now since the topic of high-frequency trading (HFT) has garnered mainstream attention, or even much coverage in the financial press. My last blog was almost two years ago, which reflects the relative maturity of the technology and business models underlying the practice of algorithmic HFT. However, the press’s favorite “frenemy” Elon Musk may soon launch HFT (pun intended) back into the mainstream with his global satellite internet network — Starlink.

HFT— what is the point?

The fundamental premise of HFT is to take advantage of low-latency connections to a stock exchange’s market-price and order data (i.e., faster connections with less time lag) in order to 1) have the most up-to-date price signals possible, and 2) be able to act on those price signals by placing orders and quotes faster than your competitors. Traders with lower-latency connections to order books and exchange prices will have an informational edge over their competitors and an operational edge in terms of acting on that information. Both good and bad side effects can arise from this speed race, and these are discussed in one of my previous blogs.

With physics dictating that time = distance / speed, there are only two ways of reducing latency — increase the speed of your signal or decrease the distance it needs to travel. Both techniques have been used by HFT firms. One early example, made famous by Michael Lewis’s book Flash Boys, was a connection built by Spread Networks using a tunnel that was bored through the Allegheny Mountains in Pennsylvania to create a slightly straighter route for fiber-optic cables between the Chicago and New York stock markets. This upgraded route would eventually shave three-thousandths of a second, or three milliseconds, from the light’s journey between the two markets.

It quickly became apparent that 1) laying fiber-optic cable is expensive and time-consuming, and 2) the speed of light in glass fiber-optic cables is around 30% slower than its theoretical maximum (obviously, the speed of light is constant; it is rather the “speed over ground” that varies as the light signal “bounces” through or, more correctly, refracts through a medium such as fiber-optic cable or air).

Beyond the speed of light (in glass).

Luckily there was an obvious solution — send the light signals through the air between microwave towers. The speed of light in air is almost the same as its speed in a vacuum. The previously mentioned cities — New York and Chicago — had been connected by an AT&T microwave network comprising 34 towers as long ago as 1949. The main issues with microwave towers are that they require line-of-sight because of the radio horizon, and that because microwaves are a higher frequency than visible light they are susceptible to disturbance from the weather. The latter problem is particularly true at the higher end of the microwave spectrum that would be used for high-bandwidth applications.

As a result of the bandwidth and interference issues, fiber optics dominate the backbones of most networks. However, microwave towers have a niche in high-value, long-distance connections. The 30% speed advantage that air has over fiber-optic cables becomes increasingly relevant when 1) distance increases, and 2) it is the latency, not the bandwidth, that is important. This set of priorities almost uniquely describes high-frequency trading signals.

Given that these microwave networks typically serve high-value applications, their existence is relatively unknown and not widely advertised — although several routes between major trading centers in the United States and Europe exist (see here and here). Often, one hears about these networks only when local residents complain about permission being given to construct one of the microwave towers.

Going into space to cross the Atlantic.

To date, microwave towers were land-bound and thus could only connect regional trading centers such as London and Frankfurt, or New York and Chicago. However, latency gains increase with distance because the relative speed advantage of microwaves has farther to play out, so connecting New York and London or, even better, London/ New York and Asian markets would be very significant. While some firms have looked at various solutions to a trans-Atlantic microwave network, these efforts have not translated into reality.

Enter SpaceX, which is in the process of launching the first test phase of its Starlink satellite-based global internet system. Starlink will comprise several thousand small communication satellites that each provide uplinks and downlinks with receivers on Earth as well as the ability to relay signals to and from the others. The phased array technology of the ground  links  and the laser communications between the satellites are very interesting; but for our purposes what Starlink does, if it works, is to solve the two main challenges of microwave networks. First, radio horizon and land-use permission issues become irrelevant because the constellation of satellites ensures line-of-sight laser communications across the globe, largely independent of any NIMBYs. Second, the network exists in a near-vacuum, so while the speed advantage over atmospheric transmissions is not that great, the reliability improvements afforded by having no weather to deal with should be significant.

While the system is being touted as a way to provide high-bandwidth, blanket coverage for the entire world, it has quickly become apparent that the system would be relatively bandwidth-limited, and almost certainly very expensive. It is simply not possible to match the bulk bandwidth of trunk-route fiber-optic cables with a constellation of satellites at a competitive cost.

Fast and overpriced? HFTs will buy it.

So at first glance we have a system that will be more expensive and bandwidth-limited than existing systems, which would not be a relevant product offering for customers wanting to watch Netflix in 4k. On the upside, it will have the potential to provide low latency across long distances, although again this is of little relevance outside of online gaming — or perhaps another niche market. Several commentators quickly caught onto the fact that an extremely expensive network whose main selling point is long-distance, low-latency coverage has a unique chance to fund its growth by addressing the needs of a wealthy market that has a high willingness to pay — high-frequency traders.

As the first operational phases of Starlink launch in the coming years, don’t be surprised if the HFT speed race, now dormant for several years, sparks back into life for one, last, glorious blast.

That is, until someone cracks quantum entanglement. . .