Can constellations of Internet routing satellites compete with long-distance terrestrial cables?

The goal will be to have the majority of long-distance traffic go over this network.
Elon Musk

SpaceX orbital path schematic, source

Four companies, SpaceX, OneWeb, Boeing and Leosat are working on constellations of low-Earth orbiting satellites to provide Internet connectivity. While all four may be thinking of competing with long, terrestrial cables, SpaceX CEO Elon Musk said "the goal will be to have the majority of long-distance traffic go over this (satellite) network" at the opening of SpaceX's Seattle office in 2015 (video below) and Leosat is focusing on high-end fast, point-point links.

Can he pull that off?

Their first constellation will consist of 4,425 satellites operating in 83 orbital planes at altitudes ranging from 1,110 to 1,325 km. They plan to launch a prototype satellite before the end of this year and a second one during the early months of 2018. They will start launching operational satellites in 2019 and will complete the first constellation by 2024.

The satellites will use radios to communicate with ground stations, but links between the satellites will be optical.

At an altitude of 1,110 kilometers, the distance to the horizon is 3,923 kilometers. That says each satellite will have a line-of-sight view of all other satellites that are within 7,846 kilometers, forming an immense mesh network. Terrestrial networks are not so richly interconnected and cables must zig-zag around continents and islands if undersea and other obstructions if under ground.

Latency in a super-mesh of long, straight-line links should be much lower than with terrestrial cable. Additionally, Musk says the speed of light in a vacuum is 40-50 percent faster than in a cable, cutting latency further.

Let's look at an example. I traced the route from my home in Los Angeles to the University of Magallanes in Punta Arenas at the southern tip of Chile. As shown here, the terrestrial route was 14 hops and the theoretical satellite link only five hops. (The figure is drawn roughly to scale).

So, we have 5 low-latency links versus 14 higher-latency links. The gap may close somewhat as cable technology improves, but it seems that Musk may be onto something.

Check out the following video of the speech Musk gave at the opening of SpaceX's Seattle office. His comments about the long-distance connections discussed here come at the three-minute mark, but I'd advise you to watch the entire 26-minute speech:

=====
Update 10/30/2017

Elon Musk has set a goal of having the majority of long-distance Internet traffic go over the SpaceX satellite network. The key to that is a richly connected, optical mesh linking their fast-moving satellites.

SpaceX is building a vertically-integrated organization -- rockets, satellites, ground stations, etc. are all being designed and manufactured in-house as opposed to OneWeb, which is working with strategic investors and partners. Based on this, I suspect that SpaceX is designing their own optical mesh network for inter-satellite communication.

I've not heard OneWeb talking about optical links between their satellites but, if they are planning for that, they might be seeking a partner and Mynaric may be a likely candidate. Mynaric says they have solved the problem of steering a narrow laser beam sufficiently accurately to keep it locked on a target only centimeters in diameter on a moving platform hundreds of kilometers away. OneWeb might satisfy their inter-satellite communication requirement by partnering with Mynaric.

Mynaric technology might be a good fit for OneWeb and other airborne platforms.

=====
Update 12/21/2017

As stated above, Elon Musk set a goal of having the majority of long-distance Internet traffic traverse the SpaceX satellite network. He based that goal on an inaccurate assumption about terrestrial networks and said nothing about competition from other LEO satellite networks. Let's look at both factors.

In his talk, Musk claimed that a route from Seattle to South Africa would involve 200 router hops. That is a major overstatement. I just ran a trace from my home to Durbin Technical University in South Africa and it was only 16 hops with an average latency of 350 ms. Musk also asserted that it would take only 2 or 3, perhaps four hops via satellite. My estimate to Chile, shown above, was 5 hops so that claim may have been a little optimistic. Regardless, it is far fewer than 200 hops.

Second -- Musk drew a comparison to terrestrial cables, pointing out that transmission is slower through a cable than in space and cables had to weave around geographic obstacles. That is true, but it seems that SpaceX will have at least one strong, satellite-based long-link competitor, Leosat. Leosat is focusing on the market for low-latency, secure, point-point links.

I can't believe I am contradicting Elon Musk (I am a big fan), but he may not reach his goal in the long-distance Internet service market.