Last May, SpaceX founder Elon Musk tweeted “6 more launches of 60 sats needed for minor coverage, 12 for moderate” and SpaceX President and CEO Gwynne Shotwell recently said they planned to be offering service in parts of the US in mid-2020, which would require six to eight 60-satellite launches. The first of those launches will be in the middle of this month on a thrice-flown Falcon 9 booster. (They will also need customer terminals and Elon Musk has used a prototype to post a tweet from his home).

Six to eight launches would bring them up to Musk’s “minor” coverage by mid-2020 and, if they maintain the same launch rate, they will achieve “moderate” coverage around the end of the year. But, what is meant by “minor” and “moderate” coverage? A simulation by Mark Handley, a professor at University College London, provides an approximation of the answer.

The first Starlink “shell” will have 24 orbital planes. Each orbital plane will have 66 satellites at an inclination of 53 degrees and an altitude of 550 km. Handley ran simulations of the first six and first twelve orbital planes—corresponding roughly to the SpaceX plan for 2020. Snapshots of the coverage area “footprints” from the two simulations are shown below:

Coverage with six and twelve 66-satellite orbital planes

The blue areas—around 50 degrees north and south latitude—are regions with continuous 24-hour coverage by at least one satellite. With six orbital planes, there will be continuous connectivity in the northern US and Canada and much of western Europe and Russia, but only southern Patagonia and the South Island of New Zealand in the sparsely populated south. Note that the financial centers of London and (just barely) New York will have continuous coverage, but, since these early satellites will not have inter-satellite laser links (ISLLs), SpaceX would have to route traffic between them through an undersea cable.

Coverage is continuous around 50 degrees north and south.

(At this point, you should stop reading and watch the video (6m 36s) of the simulation which shows the footprints moving across the surface of the planet as it rotates).

With 12 orbital planes, all of the continental US and most of Europe, the Middle East, China, Japan, and Korea will be covered. Shotwell says that once they have 1,200 satellites in orbit, they will have global coverage (with the exception of the polar regions) and capacity will be added as they complete the 550 km shell with 1,584 satellites. That should occur well before the end of 2021 since she expects to achieve a launch cadence of 60 satellites every other week.

Shotwell also said they planned to include ISLLs by late 2020, implying that around half of the satellites in this first shell will have them. Those ISSLs will give SpaceX an advantage over terrestrial carriers for low-latency long-distance links, a market Musk hopes to dominate. ISLLs will also reduce the need for ground stations. (Maybe they can lease ground-station service from SpaceX competitor Amazon in the interim)

All of this is cool, but what will it cost the user?

It sounds like SpaceX is serious about pursuing the consumer market from the start. When asked about price recently, Shotwell said millions of people in the U. S. pay $80 per month to get “crappy service.” She did not commit to a price, but homes, schools, community centers, etc. with crappy service would pay that for good service, not to mention those with no service. Some customers may pay around $80 per month, but the price at a given location will be a function of SpaceX capacity, the price/demand curve for Intenet service, and competition from terrestrial and other satellite service providers—so prices will vary within the U. S. and globally. In nations where Starlink service is sold by partner Internet service providers, they will share in pricing decisions.

Since the marginal cost of serving a customer is near zero as long as there is sufficient capacity, we can expect lower prices in a poor, sparsely-populated region than in an affluent, densely-populated region. Dynamic pricing is also a possibility since SpaceX will have real-time demand data for every location. “Dynamic pricing of a zero marginal cost, variable-demand service” sounds like a good thesis topic. It will be interesting to see their pricing policy.

National governments will also have a say on pricing and service. While the U. S. will allow SpaceX to serve customers directly, other nations may require that they sell through Internet service providers and some—maybe Russia—may ban Starlink service altogether.

The price and quality of service also impact long-run usage patterns and applications. Today, the majority of users in developing nations access the Internet using mobile phones, which limits the power and range of applications they can use. Affordable satellite broadband would lead to more computers in homes, schools, and businesses and reduce the cost of offering new Internet services, impacting the economy and culture and leading to more content and application creation, as opposed to content consumption.

Looking further into the future, SpaceX has FCC approval for around 12,000 satellites and they recently requested spectrum for an additional 30,000 from the International Telecommunication Union. Their next-generation reusable Starship will be capable of launching 400 satellites at a time, and they will have to run a regular shuttle service to launch 42,000 satellites as well as replacements since the satellites are only expected to have a five-year lifespan. (One can imagine Starships dropping off new satellites then picking up obsolete satellites and returning them to Earth).

This sounds rosy. As we said in the NSFNet days, what could possibly go wrong? SpaceX seems to have a commanding lead over its would-be competitors. Might they one day become a dominant Internet service provider in a nation or region and abuse that position? Also, before they launch 42,000 satellites—or even 12,000—SpaceX better come up with a foolproof plan for debris avoidance and mitigation. I hope they have a vice-president in charge of unanticipated side-effects.

Update Nov 5, 2019:

Speaking at an investment conference, Shotwell said that a single Starship-Super Heavy launch should be able to place at least 400 Starlink satellites in orbit. Doing so would reduce the per-satellite cost to 20% of today’s 60-satellite launches.

Update Nov 6, 2019:

Serge Eagleson informed me that since Mark Handley ran his simulations, SpaceX modified the configuration of the 550 km shell in order to sooner serve the southern U. S. The new configuration will have 72 orbital planes of 22 satellites rather than 24 orbital planes of 66 satellites—1,584 sats either way. The change will broaden coverage in the southern U. S., but thin overall capacity. Serge ran a simulation of the tentative configuration by around the middle of next year, 18 planes of 20 satellites:

U.S. overage with 18 20-satellite planes

The following shows global coverage with 18 and 36 planes of satellites:

Global coverage with 18 and 36 20-satellite planes

At 18 planes, the Earth outside of the polar regions is nearly covered. With 36 planes, it is fully covered and there is more capacity (deeper blue shading). The remaining 36 planes of the shell will further increase capacity.

Update Jan 16, 2020:

Gwynne Shotwell predicted a Starlink launch every two weeks, and they seem to be achieving that cadence. The next Starlink launch will be as soon as January 20—two weeks after the last one. That works out to about 1,500 launched by the end of this year. If they can maintain that rate, they will have completed their first 550-km shell around February 2021. (There will be some satellite failures, so that may take till March). At that point, they will have continuous coverage between about 53 degrees north and south latitude.

The financing also seems to be working out. Last May, Elon Musk said their recent fundraising rounds “have been oversubscribed” and “At this point, it looks like we have sufficient capital to get [Starship] to an operational level.” I don’t know what an operational level means, but if it means 400 satellites per launch, we are going to be seeing a lot of Starlink satellites in the next few years.