With the right type of architecture, low-earth satellite could offer better latency than terrestrial.
terrestrial internet goes through optic fiber in which the speed of light is maybe half the speed of light in a vacuum. Already HFTs use microwave links to shave off a few milliseconds between New York and Chicago. From Boston to Seattle the extra 600 miles up and down from the constellation would be overcome by the fast path along that route.
Note that requires switching from one satellite to the next, not the "bent pipe" architecture to your local rent seeking wireless company that StarLink and other LEO constellations plan.
I’ve read the paper that started this idea of Starlink being a low latency alternative for hft (http://nrg.cs.ucl.ac.uk/mjh/starlink-draft.pdf) There are a couple of problems with the paper. The first is that they establish a range of acceptable latency as existing between the great circle route and the fastest ping time on the internet between two points. For example they found that NY to London was 75ms round trip on the internet. However the long time standard bearer for a pure fiber path had been 64ms until the latest gen cables arrived in 2016. At that point the latency was sub 60 ... very comparable to the 55ms great circle route. But it’s important to understand that the starlink point to point route will be constantly changing — so only in one or two instances will that latency even come close to the existing terrestrial routes. More often than not the latency would trend closer to the 70ms range - well beyond anything considered competitive either in the current generation of routes much less the prior generation. It’s as if the authors simply had no idea about the hft market and what latencies might be needed (all of which has been published publicly). But there’s another problem with the study. For the solution to work, spacex would need to dedicate one of its laser links to enable communications across other satellites and dedicate that capacity to this application. And further, the study did not address the issue of optronic delays within the satellites. Most microwave manufacturers care little about hft latency tolerances and intermachine delays can be 1-2ms per hop. Assuming starlink to not to be much different (although I don’t know this to be true or not) that would definitely add even more latency to the equation. So all in all while it’s interesting and will do amazing things for consumers and enterprises, I don’t see starlink being a competitive hft solution unless the route in questions is really long and a lot of other engineering stuff can be made to work to enable it. Call me a skeptic but I don’t think we will see it.
terrestrial internet goes through optic fiber in which the speed of light is maybe half the speed of light in a vacuum. Already HFTs use microwave links to shave off a few milliseconds between New York and Chicago. From Boston to Seattle the extra 600 miles up and down from the constellation would be overcome by the fast path along that route.
Note that requires switching from one satellite to the next, not the "bent pipe" architecture to your local rent seeking wireless company that StarLink and other LEO constellations plan.