I dont know if it will be starlink but I do expect(hope) ground based telecom to go the way of floppy disks in the coming decades
Satellites however move so you need to accept the network sucks in moderate density areas or you have vastly more coverage in low density areas than you need.
Wired connections inherently provide a lot more bandwidth, and in dense urban environments the last mile isn’t a mile it’s within a building.
Some of that is provided for free when covering higher density areas which satellite networks like Starlink simply can’t handle at any kind of reasonable constellation size.
A quick search suggests that the range of a 5G tower, operating only at low/mid-band spectrum (so in other words, below peak speed - but at higher range) can only operate in the 1 to 3 mile range. [1] We'll say 2. That's an area of pi*2^2 = ~12.5 square miles, we'll say 13. The area of the US is 3.8 million square miles. So your number would provide coverage for (10,000 * 13) / 3.8 million = 3.4% of the US. That maybe enough to cover the most exceptionally dense urban locations, but you're missing a lot of people there.
And, again, this is just for the low/mid-band stuff. And then you need to regularly maintain those towers. While you could get global coverage with relatively few satellites that can just be trivially remotely launched/decommissioned. A quick search there [2] turns up a current practical (not peak/theoretic) bandwidth for Starlink in the 100+ Mbps range + ~50ms latency. I have difficulty seeing a logical argument for ground based telecom, beside as a hedge against WW3 when probably the first thing that will happen is a huge chunk of all satellites going poof.
[1] - https://dgtlinfra.com/cell-tower-range-how-far-reach/
[2] - https://www.pcmag.com/news/starlink-speed-tests-2023-vs-2022
Before you ask if 10,000 * 2,000 = 20 million square miles is high, handoffs require you to be in range of multiple towers so there’s a lot of overlap and hills, ocean, etc that reduce useful range.
Ultimately there’s 142,100 cell towers in the US, but that’s including density from urban areas and redundancy from multiple cell networks. Anyway, the point I was making was that Starlink is targeting low density areas by necessity they simply can’t target NYC density for any reasonable constellation size. However, if you’re a cellphone company and you’re already covering anywhere in the US with 50+people per square mile extending that to anywhere with 5+ or even 0.5+ people per square mile and killing Starlink just doesn’t take that may towers.
https://www.benton.org/headlines/us-cell-towers-and-small-ce...*
Each upgrade with telecoms entails a sacrifice. You can have really fast signals that can't go far and have difficulty penetrating obstacles like walls/buildings/hills/etc, or you can have really far reaching and high penetrating signals that can't go fast. So for instance Verizon's max speed towers can only reach 1500 feet [1], so I think my estimate of ~2 miles was a pretty reasonable meet in the middle.
All that said I agree with you in principle. Obviously space based telecoms are much better for less populated areas than heavily populated, but I'd argue that that space based can scale much more easily. The ground based telecoms aren't just those 140k towers, but also the other 450k nodes on top. And that's to cover a pretty small geographic area. And each of those nodes not only needs land and construction permits, but they also need to be be regularly maintained, and so on. It's a pretty big deal. For space based coverage, you can just launch your satellites from Texas and have them providing coverage on the other side of the world in a matter of minutes.
Put another way - imagine we were creating a civilization from scratch and these technologies were all 'unlocked.' I don't think we'd be using ground based stuff much at all. In the present when the infrastructure already exists, there's no reason not to take advantage of it, but in general it just doesn't scale so well.
[1] - https://www.verizon.com/about/news/how-far-does-5g-reach
The thing about high density places with a ton of infrastructure is that wired will always be the best because you have close access to infrastructure and its likely to already be built in. For rural areas or even suburban areas the equation starts to tip to orbital wireless for the reasons you state above and also geographic realities make ground based wireless unreliable in places where there are mountains valleys canyons etc.
There’s a 4 orders of magnitude difference between high density areas and low density ones. So no you don’t need millimeter wave everywhere. You can increase bandwidth per tower, but you can also the number of cell sites.
Further every frequency you add removes users from other frequencies. IE: At 10 miles you can use a subset of frequencies, but those frequencies don’t need to cover for people 100m from the cell tower because those are on 5G.
Thus double the number of cell sites means there’s an extra circle of people on mm wave frequencies around the new towers. Thus you more than double effective bandwidth in low density areas when you double the number of towers.
Meanwhile the reverse happens with satellites. For a given number of satellites there’s some areas where you have sufficient capacity for the density at those area. Suppose you have enough satellites for ships and aircraft over the ocean, add new satellites to handle higher density and the time those satellites are over the ocean isn’t getting you new customers. IE the percentage of time the average satellite is at 90+% capacity drops when you add more satellites.
You can’t just put 50 satellites next to each other over a suburb and call it a day you need a ring(s) of satellites circling the entire globe to reach whatever your target density is along their full orbit. Unfortunately, most land has really low density North Dakota only averages 11 people per square mile, while Florida a mostly empty state sits at 422.
Target 10 people per square mile (adjusting for household size and rates percentage of people signing up) and just about all your satellites are useful across the entire US.
But Pick 100 people per square mile 90% of your time over North Dakota is wasted. Worse large chunks of Florida are also nearly empty as most of its population is along the coastline in places like Sweetwater where 8,800 people per square mile live. So your wasteful 100 people per square mile in ND still only covers a small fraction of the population in Florida.
Cellular is the reverse the first 10k towers are largely “dead weight” that cover few people per tower, but the rest of the 130k are really useful because you optimize locations for density. Swap that to satellites initially the constellation has very high utilization, but the ratio keeps getting worse as you add more satellites.
PS: Starlink could try to vary speeds or prices more based on density, but people really want predictable results for their money.
By contrast SpaceX: build satellites, launch satellites, done. They can launch tens (and soon hundreds if not thousands) from their base in Texas with a single launch. There's still some bureaucrazy they have to deal with, but this is overall just a many orders of magnitude greater difference in terms of scalability and overall ease. And when satellites start hitting end-of-life - no problem, just deorbit them and continue expanding the swarm.
That's not really a problem, and is already the case over most of the world (particularly oceans.)
So, for any square mile of land, you can have at most some small number of subscribers. Sure, you can cover a huge surface, but only with a very low density. Conversely, the vast majority of the world's population lives in huge clumps in small areas.
Starlink should (annd perhaps anlready do) approach commercial high rises with a single uplink for the building, shared through the building network. Even as a backup data system, still very valuable system.
Starlink has ~1 million customers in the US from ~6,000 satellites so you’d think they could do 10x that with 10x the number of satellites. But much of the US is low enough density that their current constellation is already sufficient and effective bandwidth per satellite is maximum bandwidth * percentage of orbit in useful locations. Which means 10x satellites are closer to 3x useful bandwidth and it gets much worse the higher density you’re aiming for.
Ahh you might think just offer lower bandwidth per customer in urban areas, but people will pay less as the bandwidth drops and Starlink is already fairly slow.
Outside the US there is major major FTTH rollout going on everywhere. UK will get to ~90%+ penetration within a few years (it's at 66% now, 83% if you include HFC DOCSIS3.1 and growing at ~1% a month, including some very very rural areas).
All of Europe is basically like this, and will have/already has overwhelming FTTH coverage. The same is happening in Asia and even in the more developed parts of Africa.
So every month the addressable market for Starlink IMO declines. There will of course be places that are extremely rural that won't get FTTH for a long time (but I think they will eventually), and underdeveloped countries will struggle to roll it out for a long time too - but there is a lack of capacity in these underdeveloped places to pay for starlink.
What Starlink as actually amazing at is bringing the price of fixed line connections down. A lot of countries have ridiculously high fixed line/mobile data costs (I would assume some level of corruption is happening to keep competition out). Starlink will push those prices down and force providers to offer unlimited data packages in those areas. However, I'm not convinced Starlink will see the benefit of a lot of that.
Don't get me wrong - Starlink is an awesome service that really benefits humanity. However, I think the long term economics for it are poor for it to grow substantially more (this may be ok as I believe it is EBITDA positive now). And I think churn will be a problem in developed countries as more and more of them get FTTH.
They are just one of many hopefuls in the industry.
There's a big difference. Virtually all rocket startups are doomed regardless of whether their rockets succeed or fail because of the competition from fully reusable Starship and/or Terran R and/or Stoke.
Relativity and Stoke actually have a chance of success. It's small, but non-zero.
I was speaking with someone from US Cellular at a conference recently and they were saying their whole advantage over T-Mobile was spectrum in rural areas. Which explains why they were bought out by T-Mobile.
T-Mobile is also in the process of rolling out fiber home internet to compete with AT&T because 5G isn't that great.
T-Mobile's network might be good enough in suburban areas but it's lacking in rural areas. This is where Starlink is competitive, and also places like the middle of the ocean.
The issue is it’s rolling out enough places to start hurting Starlink because speeds are generally higher and it costs 50$ vs 120$ / month.
I probably would have been fired if I said that out loud in front of the client.
The $/kg doesn't matter that much in practice in the real world. Maybe in the future where people are launching only fuel. But not now.
The reality is Spin launch requires people to build sats 'for' launching on Spin launch. And that's simply not happening. No major consumer is gone do that.
Spin launch is borderline a grift.
A cellphone experiences higher peak acceleration when hitting a floor when falling off a table but they are generally fine. So yes some sats would need modifications, but not necessarily significant ones and you can trivially test those conditions prior to launch. A Starlink competitor would happily design for such if it reduced their launch costs by 90%.
This simply isn't something that anybody would do unless there is a huge benefit to it. Spin Launch still requires a rocket, and one that will likely not be reusable one anytime soon. So the idea that its going to massively reduce cost is a fantasy.
To even have any cost saving at all they need a huge number of launches. There simply isn't a market for that many tiny sats. There is a reason all the constellations have moved to larger sats that are not SpinLaunch compatible. Micro launch is a tiny market and companies like SpinLaunch are just lying to their investors, just like Astra did.
If we are looking at things like consumables, it would take 500 SpinLaunch launches and that doesn't even match a single Starship launch. How anybody could think this would be cheaper is beyond me.
Starlink is becoming a core component of their business model which presents its own issues. The service is attractive today, but cellular or wired connections have inherent advantages in most situations and could seriously eat into their customer base fairly quickly. T-Mobile/Verizon/etc 5G wireless home internet for example is generally a better choice for most people.
Which isn’t to say SpaceX is in any kind of imminent danger, rather the company has significant long term risks.