> I'm curious how no other company has managed to even directly compete with Musk when it comes to a service like Starlink.
Simple: to provide maximum speed to the end users, you want a large amount of small cells so you can better use the allocated frequency spectrum. However, launches used to be really expensive, so operators had to go for high orbits covering large areas - Iridium, for example, cost 5 billion dollars to build for 77 satellites, meaning a cost per satellite of ~64 million dollars (hardware+launch).
In contrast, SpaceX is at ~250k-500k dollars for the satellite and anything from 30-60 million dollars per launch of 60 satellites, so total cost of ~750k-1.5 million dollars per satellite [2]. That's a whole order of magnitude less cost than Iridium.
And on top of that, SpaceX's lower orbit also means better service quality due to lower latency and better signal quality. Established satcom providers physically cannot compete with that.
>> Starlink is also going to be seeing competition in the coming years thanks to companies like OneWeb and Telesat, which plans to create smaller constellations that will offer service by 2021. Tech giants like Amazon and Samsung have also announced plans to deploy their own constellations, which would consist of 3,236 to 4600 broadband satellites, respectively.
SpaceX has a huge advantage in launch costs here. Mass produced satellites will be lower cost per unit than most others, and launching on used rockets will give them the cheapest deployment costs ever. And they are the only company that can reuse rockets today. I don't see the competitors really having a chance.
> with a much smaller cost per unit for build, launch and maintenance
Somebody further down in the comments estimates $250k in launch costs for each Starlink satellite^[1]. Elsewhere, I found a claim that SpaceX execs said in 2019 that the per-satellite cost was "well below" $500k^[2]. So let's say $400k per satellite including manufacturing and launch, just to pick a number.
Once it's in orbit, you need to pay for operations, but not maintenance (except on your ground-based infrastructure). And multiple factors (e.g. cheaper launch) stand to bring that price down.
Realistically, how much do we expect one Starlink satellite's worth of telecom UAS to cost on an ongoing basis, including the maintenance and operations costs that don't apply to Starlink? Where is the analysis? OP's article makes the "cheaper" claim with (AFAICT) absolutely nothing to back it up, and the fact is that full size aircraft are not cheap.
The service density flexibility and possibility for additional radio bands Starlink can't provide might be a good argument for this system to exist, but all this talk of it being cheaper seems very vague and presumptuous. What am I missing?
AT&T doing the same number of launches would generate over $1B in profit.
And again, Starlink is already competing with land-based cellular networks, and so would AT&T. So the bulk of the AT&T satellite customers could come at the expense of other land-based cellular networks rather than Starlink -- Starlink has ~2M users compared to >140M for Verizon Wireless (much less carriers in other countries), so who are they really competing with?
> Starlink has like 1700 of the planned 42,000 satellites in orbit now.
I'm not sure if anyone believes that 42k number. They are launching ~60 satellites at a time - that would mean ~700 launches. There is no way that will be economical for the relative handful of people (500k? 1-2M?) who could realistically be interested in this.
Not to mention, the lifespan of these satellites in orbit is tiny, just a few years. They would have to be constantly launching new satellites to keep up.
The current state is probably more or less the best Starlink will ever offer - as more people will join the network, coming closer to Musk's 500k number, bandwidth will significantly diminish, even if the number of satellites is maybe doubled.
And if federal funds dry up, I expect the whole venture will quickly go bankrupt, or remain alive with a handful of satellites and a huge price spike.
> I can't remember the numbers now, but I think that meant costs in the $10's of millions / week
$500 million per year is a drop in the bucket for ~600Tbit of worldwide low altitude high speed satellite Internet. That’s ~$800/Gbit at 20Gbit/satellite and 30,000 satellites.
Starship will carry 240 satellites at a $2 million launch cost. That’s $250 million in launch costs every 5 years to fully cycle 30,000 satellites. If the satellites themselves will cost ~$250k each or $7.5b for the full constellation, so annual costs overall would be $1.5 billion, but I would guess at that scale per-satellite cost is actually much lower.
If you do the math, a fully operational Starlink will print vast sums of money ($10s of billions annually) until if/when there is a viable competitor.
The beauty is that just the marginal costs to Starlink directly fund SpaceX’s R&D budget — basically the cost side of the equation pays for Starship, while the substantial profit is just pushing Elon further up the Richest List to the point where he’ll have to find another trillion-dollar problem to throw money at.
> Kessler syndrome is not going to happen because unlike their competitors SpaceX intentionally put all their satellites in very low orbit where the atmosphere naturally cleans up any debris within a couple of years. That's something uniquely enabled by SpaceX's low launch costs because it limits the useful life of the satellites, requiring frequent replacement.
I haven't researched Starlink much before, but if the target satellite count is 10k (some say 30k?) and replacements are needed every 24 months, this suggests that to keep Starlink running they need to deploy an average of ~415 satellites every month, forever. Some quick searching suggests a cost of $250k/satellite[0], for a total monthly bill of over $100M. I must be misunderstanding, because that seems pretty pricey, and it excludes launch costs, which they say are ~$30M. Maybe everything will be cheaper by the time the constellation reaches 10k satellites.
Starlink is promising well under a million per satellite, and OneWeb is reportedly already at a million. Satellite costs might end up lower than launch costs, and even if they're higher that doesn't seem like they'll dominate.
> Yes, if it's directional, and the directionality is pretty narrow if you want to get any reasonable performance. I don't think it's feasible for each town to have a dedicated antenna for each house within a 50 mile radius.
You're asking the wrong questions.
So lets say Starlink is 4000-customers per satellite (10x oversubscription of a 40Gbps underlying, 100Mbit per customer, seem fair to you?). It costs $1000 to $3000 for the Starlink Satellite dish right now.
The total costs of Starlink, per customer, is therefore 1/400th of a satellite launch + $1500 or so (depending on how cheaply the dish can get).
Can you run an antenna from a town center to a house 50-miles away for less than $1500 + SatelliteLaunch/4000 ??
EDIT: 4000, not 400.
I admit that I don't know how much a Satellite Launch costs, but Elon Musk is asking $30 Billion for a reason. I assume the satellite launches are a good chunk of those costs.
> Starship could drive down launch costs, but it can't reduce the cost of the satellites themselves. You'll hit a floor where it could reduce costs, but not remotely enough to justify it.
You're assuming that satellites are so expensive that the difference between Falcon 9 and Starship is not relevant but we don't know this. Satellite hardware is usually very expensive but SpaceX is building them internally as a series product so they might be able to drive the marginal cost/unit much lower.
It's safe to assume constellation bandwidth (and potential revenue) scales linearly with total payload mass, meaning it is proportional to ($satellite_cost_per_keg + $launch_cost_per_kg). Unless satellite cost is much higher than launch cost there are benefits from switching to a cheaper launcher.
Since Falcon 9 design is mostly frozen and still requires throwing away an upper stage for every launch it has a price floor of its own. I honestly wouldn't be surprised if launch is already more expensive than satellites.
> That’s ~$800/Gbit at 20Gbit/satellite and 30,000 satellites.
This is pretty optimistic because the shells at different altitudes have overlapping beams and can not all use the same spectrum at the same time (OK, technically, two satellites can share the same spectrum if they use different polarizations, but, anyway, SL plans to have 3 orbital shells according to WP.) LEO bandwidth scales sublinearly with the number of satellites because of this. Also, this doesn't take into account inefficiency due to timing and coordination between satellites, nor the fact that 70% of the total bandwidth is over water and even the bandwidth over land is mostly not over population centers. The actual cost per bit is estimated to be about 2.5x-10x higher (see https://twitter.com/Megaconstellati/status/13103869669917040...).
> Starship will carry 240 satellites at a $2 million launch cost.
I think the operative word here is "will". In 2019, SX charged around $2700/kg to launch (not sure to what altitude, but let's take that figure, if you will.) A SL satellite is around 260 kg. If I round the numbers down to $2500/kg and 250 kg/satellite, I get $37.5 MM to launch 60 at a time. Suppose Starship can overcome the laws of physics and aerodynamics and launch 240 satellites with the same amount of fuel, so the cost drops by 75%. That's still roundabout $9.3 MM/launch, or, around $1 B in launch costs every five years.
Nevertheless, as you pointed out, that's not much compared to the cost of the satellites themselves. My understanding is that $250k/satellite would be very good, even at scale, but let's take that number. That brings the cost of making and launching the constellation to around $8.5 B / 5 years.
But, that's just the satellites. You need an entire ground segment to make the whole thing work. As a rough estimate, say that's another $8.5 B, but it'll last twice as long, so, $4.25 B / 5 years. OK, now we're at around $2.5 B / year just for the machinery, not including the CPE, landing rights, or personnel to keep things humming.
> If you do the math, a fully operational Starlink will print vast sums of money...
I think that's the big question everyone wants the answer to: Who will pay for this? SL will cost around $2.5 B / year in just the ground segment and the space segment. At $100/month, SL needs a bit over 2 MM customers just to cover the cost of equipment. But those 2 MM customers are not spread out evenly around the Earth, where all of the bandwidth is.
If you have any insight as to who will pay for this, I'd genuinely like to know because I'd be happy to buy into the SL IPO and retire early.
> OneWeb has over 600 satellites in orbit right now, with paying customers.
OneWeb is B2B only. I assume, but don't know that it's because they can't get the cost of their user terminals low enough.
> Amazon's Kuiper constellation is planned to have over 3000 satellites; they've got regulatory approval and launch contracts signed, and their pockets are deep. Once those come online properly, it'll be hard for Starlink to charge anything like proper monopoly rents.
It'll certainly be interesting to see how that goes. They went with the "hire everyone but SpaceX" launch strategy, which means that none of the rockets they've signed with are operational yet. And they have a looming FCC/ITU deadline. There's widespread belief that if they make a good effort, they can get a variance, but it's certainly not guaranteed.
In general, I agree with you, though. There's not much evidence that SpaceX is trying to squeeze customers. In part because there are often plenty of competing options, and in part because Elons companies tend to keep prices low even when they do have a defacto monopoly (like right now with space launch).
>If I understand correctly, the number of bright satellites will be proportional to the launch rate, not the total quantity in orbit.
Because the constellation needs constant replenishment, the launch rate (and hence the # of brighter satellites) will have to reach a steady state that is directly proportional to the size of the constellation.
We can extrapolate that in the future this rate will actually be significantly higher than it is currently:
They currently have permission to launch around 12,000 sats. They're launching around 250 per year. The current launch rate is only sustainable if each satellite lasts for 50 years.
If you expand that to the proposed 42,000 constellation, 250 new satellites per year is only sustainable with a MTBF of around 150-200 years per satellite, which is nigh impossible in low Earth orbit. Using a lower (but still very generous and optimistic) MTBF of 10 years, Starlink will need to launch 4,200 satellites every year, about ~15x higher their current launch cadence.
> the majority of the remaining satellites are destined for much higher (and dimmer) orbit
Just about all of the satellites they've orbited so far are hanging out around 550km. SpaceX initially got permission to go as high as 1300km, but they've since changed their mind. The new plan (still pending FCC approval I believe) is to keep all of the satellites between 300-550km. So the future satellites will be as low or lower than the current ones.
Taking what Iridium did and scaling it up 500x does not make it novel.
Iridium did all of the things Starlink does - it just did them with 2G bandwidths, instead of 5G bandwidths.
Furthermore the opex of keeping a constellation that large in orbit, has not been proven to be cheaper than operating a ground based wireless network. Starlink operates with an assumption, that has not yet been proven to be true - that the cost per kg to launch something into LEO will go down as time goes on. If this prediction does not prove to be true, then it has no hope of making money, much less being self-sustaining.
Beyond this, the shorter lifetime of their sat fleet compared to a geosynchronous sat, means they must be replaced sooner, but its not obvious that the sats are significantly cheaper.
The appearance of novelness is what allowed SpaceX to get the spectrum to operate what is effectively another cellular network - well a hybrid between fixed wireless and cellular.
> It gives a bandwidth efficiency ratio, which is directly tied to cost since the more efficient per bit the system is, the cheaper it is to run.
The source you linked said Telesat had a 4x Gbps/satellite bandwidth over spacex which is not directly tied to cost. Telesat has no mention of satellite weight and spacex plans to launch 40x the amount of satellites as Telesat for their initial constellation and 100x at full capacity if I recall correctly.
> Path to profit.
Spacex has reusable first stage rockets. I don't see how any of the other mass satellite constellation companies that you mentioned can have better margins than spacex.
> They need upwards of 500 of these in the air just to compete with satellites that have been up for multiple years already.
Do those 5 companies you just mentioned somehow don't?
> Dishy is the first phased array antenna widely available to consumers, as far as I know.
Dishy is the first phased array antenna that was willing to take a huge loss on each sale to consumers. This goes back to the point of getting VC and FCC money to subsidize the service. The estimated cost of the dish is still $1100-$2000 since SpaceX doesn't have any phased array technology that's proprietary, so at $500 they're taking a loss.
> Starlink is the first low orbit massive satellite constellation
Massive is a property of the launch cadence and not anything specific to the satellites.
> The older geostationary communication satellites are a distinctly different system.
What is different about the satellites compared to something like Iridium, O3b, Intelsat, etc?
Emphasis on up to. They only have 3000 right now and it already works. If they go above 5000 it's because they think they will make more money that way, not because they need more than 5000.
> 5000 or whatever towers ain’t insurmountable
The US has hundreds of thousands of towers, and that number gets a lot bigger with 5G.
And while that covers "a huge percentage of the population", not all of that is fast, and reaching the rest of the population gets harder and harder.
> replacing on a 5 year cycle
> The tower probably even lasts longer than 5 years.
Let's say a cell tower lasts 20 years, which is a bit longer than two cell technology generations. So that means starlink needs to launch about a thousand satellites per year to maintain their network over the whole planet. And cell providers need to rebuild 20 thousand cell sites per year just in the US, more if they try to cover everyone.
Starlink can launch that many satellites with about $300M. Let's round way up to $500M a year to keep satellites and base stations running.
Each year cell companies are pulling in hundreds of billions and putting tens of billions into towers, just in the US.
> It's not a full sized aircraft. 1 small autonomous plane is cheaper than 1 small satellite in general - in almost all respects. The article didn't claim anything beyond this:
From the picture in the article I'm not sure what else to call it other than "full size". Wiki claims it has a ~90 foot wingspan^[1]. Wiki also claims it has an 11 lb payload capacity, so one imagines a practical version of this concept would need to be be substantially larger in order to carry a useful telecom payload.
(A Starlink satellite is a couch-sized slab weighing over 500 lbs, and one can only assume how much of that is telecom equipment, but it's probably a lot more than 11 lbs. So it seems plausible that you'd need on the order of 10 aircraft of this size, or a single much larger and much more expensive aircraft, to equal the throughput of a single Starlink satellite. None of that sounds cheap.)
Being a UAS doesn't necessarily make it cheaper, either. The MQ-1 Predator, a borderline obsolete ~1500 lb airframe pushed by a ~100 hp air-cooled Rotax, seems to cost governments on the order of 10 million dollars to procure. I'm sure there's a lot of pork in there, but it seems far from obvious that relatively exotic UASs like this one can be procured and pressed into production use for, say, two orders of magnitude less than that.
> This is fairly obvious and uninteresting, what is interesting is cost per GB/user/km2, in other words the end result, which is hard for anyone to say with much certainty at this point even with a thorough analysis because nether business has reached scale yet.
Starlink is already a service you can buy and use, subject to whatever "beta" rollout scheme they're using. They still have a long way to go to realize their scale ambitions, but I think their economics are far more settled than other schemes (like this one).
I do agree that cost per user is the interesting number here, but--as above--it seems far from obvious that this UAS scheme can even be competitive with Starlink, let alone cheaper in any reasonable sense.
"tens of thousands satellites, are all flying above OUR heads"
Do you feel threatened by them, or just resent the fact that they will alter the look of the night sky, especially around sunset and sunrise?
I can understand the latter, but communication satellites in orbit aren't a safety problem for people on the surface, unlike, say, every drone that someone flies in your proximity.
"4G works awesomely."
This is very local, I can't even get 4G coverage at some places literally on the border of Prague, a major European capital. Not a wilderness by any means.
"You are completely delusional."
Stop that shit. And do not attack my math, I actually graduated in algebra and number theory.
Seriously, to calculate returns on Starlink in the future, you need to assess several variables such as launch cost, manufacture cost, personnel cost and total # of customers. These are not yet known with necessary precision into the future, so we can only guesstimate. If we take Musk's admission that Starlink needs 30 billion USD to survive, 10 million paying end customers worldwide should be more than enough. That is not that much, compared to the world population.
And we do not know how much extra money are getting from customers such as the US Army which is really intrigued with Starlink.
"Those 20 years serving ARE still made today! Your comparison is invalid. "
So yes, they are. Do they provide a service on 2021 quality level? There are steam engines still in operation, but that does not mean that the technology isn't obsolete.
Simple: to provide maximum speed to the end users, you want a large amount of small cells so you can better use the allocated frequency spectrum. However, launches used to be really expensive, so operators had to go for high orbits covering large areas - Iridium, for example, cost 5 billion dollars to build for 77 satellites, meaning a cost per satellite of ~64 million dollars (hardware+launch).
In contrast, SpaceX is at ~250k-500k dollars for the satellite and anything from 30-60 million dollars per launch of 60 satellites, so total cost of ~750k-1.5 million dollars per satellite [2]. That's a whole order of magnitude less cost than Iridium.
And on top of that, SpaceX's lower orbit also means better service quality due to lower latency and better signal quality. Established satcom providers physically cannot compete with that.
[1] https://en.wikipedia.org/wiki/Iridium_satellite_constellatio...
[2] https://www.nextbigfuture.com/2019/12/spacex-starlink-satell...
reply