Starlink : Markets and Marketing

[guckyfan]

Of course Elon rather explicitly mentioned that he's not even thinking about selling off Starlink right now, with really what should be a very hopeful statement for GEO sat operators - every LEO constellation has gone bankrupt, and his only goal at the moment is to not go bankrupt.

That is just Elon Musk style expectation management. Right before the first FH launch he said he will see it as a success if it does not blow up on the pad.

Read this statement as he is expecting a runaway success.   

[M.E.T.]

Looking at costs, and for the moment  ignoring SL’s obvious cost advantage in launch, a constellation of 41k sats at $150k (guesstimate) each will cost $6B. Three GEO megasats will be much less for only three launches and 2-3x the lifetime. With future launch costs showing every promise of dropping dramatically the equation morphs. Sats no longer need to be so robust... yada yada. And only three sats to manage.

That is not really a fair comparison.

Three Viasat 3 satellites are going to provide 3 Tbps of nameplate capacity for a build plus launch cost of $2.1 billion ($700 million each).

Starlink's 42,000 satellite plan will provide 840 Tbps of nameplate capacity for a build plus launch cost dramatically lower per Tbps.

Then it gets complex with utilization percentages while over the oceans and during night hours, etc. But on a cost to build plus launch capacity into orbit, Starlink is launching capacity into orbit at about 1/10 the cost of Viasat. And Viasat looks like the best cost numbers of the GEO satellite operators.

This. SpaceX is getting 20 Gbps per satellite and 60 satellites per launch, or 1.2 Tbps per launch. Viasat is getting 1 Tbps per launch. And SpaceX is getting launches for internal cost, probably less then $25M each. Viasat is paying 3 or 4 times that per launch.

So launch costs, once normalized for Viasat's 2-3x longer life, are still slightly in SpaceX's favor. And satellite costs are probably 20x or more in SpaceX's favor.

Ah. I’ve been looking for the Gbps number for each satellite. Assuming each user wants a speed of at least 20 Mbps, does that effectively mean Starlink is limited to serving 1000 customers per satellite at any one time?

If so, question 2 is how large an area does each satellite service - was it a radius of around 250km or something? So only 1000 customers within such a radius, until the later stages when multiple satellites will cover each area?

I like these types of numbers for rule of thumb revenue calculations.

[RedLineTrain]

Ah. I’ve been looking for the Gbps number for each satellite. Assuming each user wants a speed of at least 20 Mbps, does that effectively mean Starlink is limited to serving 1000 customers per satellite at any one time?

If so, question 2 is how large an area does each satellite service - was it a radius of around 250km or something? So only 1000 customers within such a radius, until the later stages when multiple satellites will cover each area?

I like these types of numbers for rule of thumb revenue calculations.

For the initial 1,584 satellites (550 km altitude down to 25 degrees from the horizon), the radius covered for each satellite is 940.7 km.  In the future, that will decrease to 573.5 km (550 km altitude to 40 degrees from the horizon).

You can see in Mark Handley's simulation, that even with the 400-satellite initial operating capability, a ground station in a covered area may see several satellites at once.

I can imagine the tricks that Starlink would utilize to maximize frequency reuse, but I don't know which are realistic given this initial generation of Starlink satellite and ground station chips.

[tbellman]

Assuming each user wants a speed of at least 20 Mbps, does that effectively mean Starlink is limited to serving 1000 customers per satellite at any one time?

Assuming 20 Mbit/s is the "advertised" bandwidth the customer buys, then no.  As an ISP, you always oversubscribe your capacity; you bet that only a portion of the users will be active, and using their full bandwidth, at any given time.  Or at least that that happens rarely enough that your customers don't abandon you in droves.  Similarly, the telephone systems are not dimensioned for all customers to make calls at the same time, and sometimes crash when people do anyway.

(20 Mbit/s sounds very low, by the way.  Is that common speed in the USA?  Here in Sweden, I think 100 Mbit/s is the most common speed people buy, and 1 Gbit/s is almost universally available to buy.)

I don't know what oversubscription ratios are common for residential ISPs, though.  Somewhere between 2:1 and 10:1, I would guess, but that's a rather wide span...  Likely depends on if there is effective competition for your customers as well; if customers don't have any practical alternative, the ISP can get away with much higher oversubscription and more dissatisfaction among its users.

(Based on the amount of complaints I see, I suspect ISPs in the USA generally oversubscribe much more than ISPs here in Sweden, for example.  But municipal Fiber-To-The-Home is super-common here, with 10-30 ISPs to chose from over that FTTH connection.)

[Barley]

You can see in Mark Handley's simulation, that even with the 400-satellite initial operating capability, a ground station in a covered area may see several satellites at once.

One figure of merit is coverage area divided by the number of satellites divided by the area each satellite covers.

With a 400 satellite constellation and 900km radius every spot on Earth is covered by an average of 2 satellites.  So every spot in the coverage area certainly is.  Since the initial constellations doesn't cover the whole earth the density of the coverage in the zones around 40 degrees north and south will be quite a bit higher.  I'd eyeball it at 10 or more.  A more accurate estimate could be made by subtract the uncovered area from the Earth's total surface area.

I don't know what oversubscription ratios are common for residential ISPs, though.  Somewhere between 2:1 and 10:1, I would guess, but that's a rather wide span.

The oversubscription ratio is an important number.  My guess is that it's higher than 10:1 and could be pushing 1000:1, at least at the level of a hub serving tens of thousands of residential customers.  Anybody have some actual data?

What are residential customers doing with GigaBit links?  Doing full disk backups multiple times a day?

[LiamS]

The oversubscription ratio is an important number.  My guess is that it's higher than 10:1 and could be pushing 1000:1, at least at the level of a hub serving tens of thousands of residential customers.  Anybody have some actual data?

What are residential customers doing with GigaBit links?  Doing full disk backups multiple times a day?

generally its 100:1

[gongora]

Remember that video is not the current plan for the constellation.  It's still informative, but no longer accurately portrays the satellite layout within the constellation.

[Asteroza]

Considering the issue of saturation of a single satellite in more urban areas, I wonder if Starlink might be willing to partner with conventional broadband providers, in a fashion. If the router is intelligent enough to do connection tracking, send low latency sensitive stuff through Starlink, but offload bulk CDN stuff to wireline providers?

What are residential customers doing with GigaBit links?  Doing full disk backups multiple times a day?

Multiple people doing video streaming from a single line goes over 100Mbps with ease if you aren't careful. Perception-wise, 4K almost feels like being there so the whole remote telepresence feeling is important as well.

As for backups, gigabit speed is roughly equivalent to a single spinning HDD in performance, so to have relatively transparent remote disk access/backup reduces subjective slowness of a computer. There are a number of things that get unlocked when you can get to gigabit speed. (note this also applies to other things as well, famously USB 2.0 was 400Mbps but real world performance is 200Mbps max, so the advent of USB 3 finally allowed relatively seamless use of external hard disks irrespective of if they were SSD or HDD)

There's also the old axioms "increased capacity will always be used", and "the street finds its own uses for things"

[Hummy]

(20 Mbit/s sounds very low, by the way.  Is that common speed in the USA?  Here in Sweden, I think 100 Mbit/s is the most common speed people buy, and 1 Gbit/s is almost universally available to buy.)

No, the average speed in the US and Sweden is basically the same: 135 Mbps.

[tbellman]

The oversubscription ratio is an important number.  My guess is that it's higher than 10:1 and could be pushing 1000:1, at least at the level of a hub serving tens of thousands of residential customers.  Anybody have some actual data?

When I guessed at less than 10:1, I was thinking of aggregation close to the customer.  Like the uplinks from a 48 port ethernet switch.  Higher up, you can have more oversubscription, but 1000:1 sounds extremely high at any level.  For example, the city I live in has about 160k population (entire municipality; city proper ~110k).  Assuming customers buy 100 Mbit/s on average, 1000:1 would serve the entire municipality on a single 10 Gbit/s link.  I rather expect the larger ISPs have at least 2×100 Gbit/s to our city, quite possibly more.

What are residential customers doing with GigaBit links?  Doing full disk backups multiple times a day?

Watching cat videos rocket launch videos in 4K resolution, of course.

But, the vast majority of people having gigabit to their home will not be using that the entire day, just short bursts.  But it might make the difference if something takes 10 seconds or a minute.  Or five minutes or half an hour.

Some things that use a lot of bandwidth:

* If your household has two adults and 2-3 teenagers, all wanting to watch (different) 4K videos at the same time, 100 Mbit/s will be pretty tight.  (250 Mbit/s is probably enough, though.)

* Gaming.  Several modern computer games are pretty large.  Apparently there was an update/expansion to Call of Duty a few days ago that was close to 100 Gbyte in size, one to Rainbow Six Siege that was 80 Gbyte.  At 100 Mbit/s, downloading that will hog your line for 2-3 hours, but at 1000 Mbit/s the download could finish in half an hour, and still leave some free capacity.

* Cloud backups.  If e.g you have photography as a hobby, you can easily come home with 10 Gbyte of photos after a day, or 50+ Gbyte after a week's vacation trip.

* You are a computer nerd, who occasionally (re-)install Linux machines (including doing major version upgrades).  That's downloads of several gigabytes.  With a gigabit link, that installation will take ~10 minutes, with 100 Mbit/s it will take more than half an hour.

* Updates to MS Windows or Apple MacOS can also be fairly large.

* You are a researcher in some data-intensive field (e.g, quantum chemistry, molecular dynamics, or computational fluid dynamics).  Analysis and simulations are of course done at some HPC centre, but sometimes those generate visualization output files that can be many gigabytes in size.  You want to download them to your laptop/workstation to view also at home, not just in the office at your university.

[tbellman]

No, the average speed in the US and Sweden is basically the same: 135 Mbps.

Interesting!  I seem to hear so much complaints from Americans that I assumed the situation was much worse there.  (Of course, those who are satisfied don't write about it, and since USA has 30× the population of Sweden, one would expect to see a lot more Americans than Swedes complaining.  Plus, I probably hang out more in English-language forums and mailing lists than Swedish, so that will further bias what I see.)

Great information, thanks!

[Semmel]

In terms of the Starlink Spinoff, Shotwell said something interesting:

https://mobile.twitter.com/joroulette/status/1237453859900948480?p=v :

"We still have a lot to do to see whether this is gonna work," Shotwell said. "It was just a way to potentially get employees... I'm not saying we're not gonna do it, but I'm just saying it is not in our thought process right now. It should not be news."

I read this as: getting employees that are not US citizens. Since SpaceX is building rockets, they have a high barrier to hire non US citizens. So a spinoff of Starlink could open the option to hire people from abroad easier.

[RonM]

No, the average speed in the US and Sweden is basically the same: 135 Mbps.

Interesting!  I seem to hear so much complaints from Americans that I assumed the situation was much worse there.  (Of course, those who are satisfied don't write about it, and since USA has 30× the population of Sweden, one would expect to see a lot more Americans than Swedes complaining.  Plus, I probably hang out more in English-language forums and mailing lists than Swedish, so that will further bias what I see.)

Great information, thanks!

High speed broadband is available in urban areas but the United States is a very large country. Tens of millions of Americans in rural areas don't have access to broadband. That's a big market for Starlink.

[gongora]

No, the average speed in the US and Sweden is basically the same: 135 Mbps.

Interesting!  I seem to hear so much complaints from Americans that I assumed the situation was much worse there.  (Of course, those who are satisfied don't write about it, and since USA has 30× the population of Sweden, one would expect to see a lot more Americans than Swedes complaining.  Plus, I probably hang out more in English-language forums and mailing lists than Swedish, so that will further bias what I see.)

Great information, thanks!

Most americans don't have gigabit connections, and many do have to live with 20 Mbit/s or less.  I don't think there are many service tiers between 100Mb and 1Gb commonly offered in the US, so a single gigabit connection in their test would drag up a whole bunch of lower speed connections to create their "average".  I'd like to see the median value.

[wannamoonbase]

No, the average speed in the US and Sweden is basically the same: 135 Mbps.

Interesting!  I seem to hear so much complaints from Americans that I assumed the situation was much worse there.  (Of course, those who are satisfied don't write about it, and since USA has 30× the population of Sweden, one would expect to see a lot more Americans than Swedes complaining.  Plus, I probably hang out more in English-language forums and mailing lists than Swedish, so that will further bias what I see.)

Great information, thanks!

I'm originally from Canada and it's ever larger and more sparsely populated than the US. 

My brothers satellite internet has improved a lot over the last decade.  But it is still slow and can't stream anything.

Starlink in areas of Canada will do very well with customers.

[abaddon]

Most americans don't have gigabit connections, and many do have to live with 20 Mbit/s or less.  I don't think there are many service tiers between 100Mb and 1Gb commonly offered in the US, so a single gigabit connection in their test would drag up a whole bunch of lower speed connections to create their "average".  I'd like to see the median value.

Charter offers 200Mb (what I have) and 400Mb as a "business" tier.  It's not exactly "uncommon", but I don't disagree with your overall assessment.  I also have AT&T DSL (no fiber where I live) as a backup, which is more like 14Mb down.  Plenty of people have DSL only...

That's down, up is generally far more constrained (10Mb/1.5Mb up respectively for my links), unless you're on fiber.

[Rondaz]

Musk's SpaceX Looking to Compete for $16 Billion in Federal Broadband Subsidies.

March 12, 2020

Source: Wall Street Journal

Author:
Ryan Tracy
Brody Mullins

https://www.benton.org/headlines/musks-spacex-looking-compete-16-billion-federal-broadband-subsidies

[Brovane]

Most americans don't have gigabit connections, and many do have to live with 20 Mbit/s or less.  I don't think there are many service tiers between 100Mb and 1Gb commonly offered in the US, so a single gigabit connection in their test would drag up a whole bunch of lower speed connections to create their "average".  I'd like to see the median value.

My Local Service Provider, Cox Communications has speed offerings of 150 Mbps, 300 Mbps, and 1Gb. 

[gongora]

Musk's SpaceX Looking to Compete for $16 Billion in Federal Broadband Subsidies.

March 12, 2020

Source: Wall Street Journal

Author:
Ryan Tracy
Brody Mullins

https://www.benton.org/headlines/musks-spacex-looking-compete-16-billion-federal-broadband-subsidies

I don't have a WSJ subscription anymore to read the article, but I was looking at this a couple weeks ago.  For anyone interested I believe they must be talking about the Rural Digital Opportunity Fund proceedings.  Some relevant links:

Information about the upcoming auction:
https://www.fcc.gov/auction/904

the Report and Order for the auction adopted at the end of January:
https://www.fcc.gov/document/fcc-launches-20-billion-rural-digital-opportunity-fund-0

Docket 19-126 has the related filings from everyone, including SpaceX:
https://www.fcc.gov/ecfs/search/filings?limit=100&proceedings_name=19-126&sort=date_disseminated,DESC

The LEO constellations qualify as "low latency" service for this proceeding, which sets the bar at <100ms.  There are various speed tiers defined.  The speeds offered and latency will be factors along with pricing when picking the winners.

[su27k]

Musk's SpaceX Looking to Compete for $16 Billion in Federal Broadband Subsidies.

March 12, 2020

Source: Wall Street Journal

Author:
Ryan Tracy
Brody Mullins

https://www.benton.org/headlines/musks-spacex-looking-compete-16-billion-federal-broadband-subsidies

https://old.reddit.com/r/spacex/ now has a link to WSJ that allows you to read the entire article, not sure if it's kosher so I'm not going to post the link, just go to reddit and click the link "SpaceX Looking to Compete for $16 Billion in Federal Broadband Subsidies".

Basically it looks like SpaceX is successful in convincing FCC to allow them to participate the auction in the low-latency tier (well, sort of, it says "FCC left open the possibility", whatever that means). Apparently the previous version of the auction proposal forbids satellite internet from being qualified as low-latency, which SpaceX tried very hard to change.

Also the FCC auction proposal is not final and is currently seeking public comment, and FCC could still forbid Starlink from participating because it's a new technology.

Anyway this latest FCC change got terrestrial telecom very worried, they're asking congress to pressure FCC to stop Starlink from being able to participate in the auction.

“While this is just a proposal, if adopted, it literally could allow satellite providers to win the entire auction,” a NTCA lobbyist wrote in an email reviewed by the Journal.

This confirms Starlink is really competitive in rural areas.

Edit: The FCC auction proposal in question is located at https://www.fcc.gov/ecfs/filing/030214207017