[JumpCrisscross]

> "But why are you only assuming a usage time of 3 years?"

Your entire analysis rests on this point, which you fail to demonstrate. (You also cite zero sources, which isn't encouraging.)

(EDIT: This assumption is conservative, but reasonable.)

Was this AI generated?

> The cable itself costs virtually nothing

Did you attempt to look up the cost of laying new fibre trunk?

> due to their Satellites now being massively overbooked, speeds went down months to months

Then this isn't a remote location. Starlink's economics have been pretty obvious for anyone who has been on a plane, boat or train in the last decade. They're also terrifically useful for remote mining, observation and military operations.

> people noticed that it is actually cheaper to run 10 KILOMETERS / 6 Miles of Fiber cable through the jungle

Well sure, if you ignore negative exernalities a lot of stuff is cheap.

[Fischgericht]

Wow. Well, I believe that YOU are a bot, not me. Are you Grok?

Anyway, yes, I am a human.

And it is not that hard to find the sources for this point:

https://en.wikipedia.org/wiki/List_of_Starlink_and_Starshiel...

v1 constellation was completed in 2021, and decommissioned from 2024. v2 deployed from 2023, but the sat-to-sat communication is not working, so all of them, will need to be replaced by v3, too.

[wmf]

The sat-to-sat laser links are used to provide connectivity on the open ocean and in remote parts of Australia and Argentina that are beyond the range of any ground station. They're definitely working but AFAIK they are only used when necessary so if you're within range of a ground station your traffic will never use laser links.

[Fischgericht]

Oops, forgot one important thing: Sure, why do additional hops if you can see the base station. But what about shared state? Why do you definitely still get a completely new session when moving to the next sat? If the laser links are working, that state should be shared between neighboring sats.

[lxgr]

Inter-satellite links simply provide additional (time-variant) paths, which doesn't inherently relate to shared state.

You seem to be under the impression that inter-satellite links somehow imply a self-organizing mesh topology that preserves terminal-to-gateway associations at any cost (including that of extra in-space hops), but that does not necessarily follow from the existence of ISLs.

In other words, your observation of occasional routing instability causing higher-layer issues is perfectly compatible with working ISLs.

[Fischgericht]

Accepted.

[inemesitaffia]

Starlink switches beams every 15 seconds and satellites every 120 seconds.

You keep your sessions through both. Lasers or no lasers

[JumpCrisscross]

> Why do you definitely still get a completely new session when moving to the next sat? If the laser links are working

Imagine Amazon 10x'd its ingress/egress fees between regions.

[inemesitaffia]

You're not getting new sessions period.

[Fischgericht]

I will not disagree as I can not verify this claim. Have you tested it yourself or have a source which has some tech proof on that one?

[hughes]

Why do you believe the inter-satellite links are not working?

[Fischgericht]

[Due to the part of the spectrum I am on, I do not have believes or opinions.]

The laser based inter-links still not working has been subject on various conferences like AngaCOM etc.

But in my case: I have simply tried it *). And every Starlink user can do it, too: Use traceroute. And if you think "they might be hiding the hop-to-hops between Sats!", you can dig deeper using MTR behind the modem or simply rooting the modem itself.

Last time I have connected to a v3 Sat however was ~6 months ago. Maybe an active user reading this can try today?

[lxgr]

You're equating occasional dropouts (which can happen for all kinds of reasons even in bent-pipe topologies) with the absence of inter-satellite links. That makes no sense.

The empirical way to test for the existence of ISLs would be to go to the middle of an ocean, safely out of reach of any ground station, and see what happens. If you get a connection, that can only be due to ISLs.

It seems like your actual complaints are with network/routing stability, and you're drawing invalid conclusions from there.

[niwtsol]

Do you have a link to a blog or writeup regarding the inter-links not working? Hard to find it without getting lost in "Troubleshoot your starlink device" SEO hell.

[JumpCrisscross]

> Do you have a link to a blog or writeup regarding the inter-links not working?

The simpler answer is intra-constellation communication is a bleeding-edge technology. It's an extraordinary challenge for which extraordinary proof is needed to show success, not the other way around. SpaceX has solved most of the gating technical problems. But getting it to work reliably enough that it becomes more economic than ground-based backhaul will take time.

[lxgr]

> intra-constellation communication is a bleeding-edge technology

Iridium has been successfully doing it for a quarter of a century now.

[inemesitaffia]

An even simpler answer is Starlink is available in locations too far from ground stations.

Ergo they are served via laser.

Cook Island

Ascension Island

Iran

Venezuela

Cuba

Galapagos/Easter Islands

Vanuatu

Eastern Ukraine

Syria

Lebanon

Iraq

Iqaluit

Antarctica, as South as the South Pole

Tristan de Cunha

The range of the ground stations are under 1500 miles and I really don't know where people are getting the idea that the lasers don't work.

[Fischgericht]

Here is an example thread of someone having done the measurements of v3 vs mini:

https://www.reddit.com/r/Starlink/comments/1eg4e4d/starlink_...

Have a look at the downtimes of the system.

A simple way to verify that their inter-sat links are not working and/or are not used is to simply sit and wait: If you are switched from one Sat to the next, you get new "session" and previous NAT state is lost. If this would be a meshed backbone, that would not happen.

[inemesitaffia]

This is ridiculous.

How's service delivered to the South Pole?

Iqaluit?

As long as your traffic is terminated at the same POP, you won't get any session terminations.

And Starlink tells you when your public IP changes anyway

[Fischgericht]

Erm. That person has posted a detailed explanation on how he has measured.

How can this be ridiculous? Is it ridiculous because the data does not match your believes...? Confirmation bias?

It's Data. And it hints, amongst other things that they have seen the same that I am seeing on every single Starlink installation I got my hands on so far: There is no active handover, and no shared state between Sats.

And you are referring to the wrong layer, talking about the ground station. Of course that does not move, and does not forget about your IP. Wrong layer.

It's about the Satellites (!) not doing an active handover and not sharing L2 state, like it would the case for any meshed network, no matter if cellular or WiFi. The analogy here would be a WiFi access point or a cell tower, and you roaming from one to the next while having a phone call, not having any drop-outs. That's the industry standard for Wireless. Starlink isn't there (yet).

If you don't think that true data is true, check ARP table of the MAC of your gateway IP changing after handover.

You appear to be a happy Starlink user - so do you care to share some 24h benchmark with us to prove your claims? I would highly appreciate that!

So far sadly none of the "But it works!" people has been able or willing to provide a benchmark on their own setup. `

Again: I am not here to win and argument. But to change my conclusions, I need data that hints at my conclusions potentially being wrong. As explained elsewhere in this thread, due to lack of serious benchmarks, most of this is based on anecdotical data points.

[inemesitaffia]

There's lots of users in laser only areas doing RTC via VOIP or video. Some of them on Ships in the Persian Gulf (look up SEA-2. https://www.wired.com/story/us-navy-starlink-sea2/)

That person has posted an install in a moving vehicle with the antenna inside.

Do you think I can't read or what?

The ground stations don't handle user IP addresses or even IP packets.

They are strictly layer 2 from the user perspective and traffic is terminated at POP's.

Here's a scientist that does actual work, not the nonsense you've been posting as "data".

https://www.reddit.com/user/panuvic/submitted/

I gave you his email.

I've pointed out to you where you can travel to test laser service yourself.

I'm not saying "but it works"

I pointed out to you places that can't be served without laser connectivity.

It's like asking me to prove the earth isn't flat.

[lxgr]

That person is measuring periodic packet drops (which can also be caused by e.g. having an incomplete view of the sky), and you're drawing unsupportable conclusions on "session drops" from that. (The word "NAT" does not even occur in the observation thread you've linked!)

So, yes, confirmation bias.

[JumpCrisscross]

> I believe that YOU are a bot

I don't believe you were a bot, but there were one or two phrasings that gave me pause. (If I believed you had written that with AI, I'd have just asked that and not bothered engaging.)

> v1 constellation was completed in 2021, and decommissioned from 2024. v2 deployed from 2023, but the sat-to-sat communication is not working, so all of them, will need to be replaced by v3, too

Fair enough. $3.6mm on $2mm--assuming $100,000 per month revenue and $2mm paid up front, which is unrealistically conservative--yields a 22% annualised. Take that out to the increasingly-attained design life of 5 years and it jumps to 25%. To put it bluntly, these are both incredibly high telecom returns.

You've already incorporated launch, maintenance, disposal, et cetera in TCO. So the remainder is customer service (usually 5 to 10% of revenue) and cost of capital. Even assuming 10% WACC, which is on the upper end for a leveraged telecom play, we're still comfortably generating excess return.

Where the comparison fall apart is in respect of fibre. Laying physical infrastructure is hard. You have long periods between capital outlay and return. Also, you have to right scale up front--you can't just launch more birds in a few months as demand scales (or hold them back if it doesn't).

You're not going to replace fibre with Starlink. But the economic case for the latter doesn't fall apart with 20%+ operating returns.

[Fischgericht]

Well, on purpose I have given Starlink very optimistic numbers, yes. :)

And yes, 22% yield sounds nice, but if someone would hand me their pitch deck and give me a SWAT analysis I would just laugh them away: The risks are far too high.

(See for example the article that this very thread is about.)

Of course you can only guess based on that, but it looks that in real life things are worse:

https://arstechnica.com/space/2025/02/starlink-profit-growin...

These data points might be interpreted as "Starlink is getting 40% of their revenue from tax money".

And while "7 million subscribers" might sound impressive on first sight: This is the number of DSL connections subscribed to in the tiny country of Belgium. But for magical reasons Starlink is valuated at a price higher than if you would buy all of Belgium ;)

Your point in regards of laying physical infrastructure is valid for a lot of western countries. But not all of them. Some countries in the EU for example years ago created laws that say that whoever opens the street for any reasons has to put in empty tubes for someone to later put in fiber before closing the street again.

So: This is a regulatory subject really, not physical cost. Fiber is dirt cheap if you are allowed to use existing power poles for example (which is unlike with copper obviously not a problem in regards of signal integrity), or existing underground pipes, or just throw it from house roof to house roof.

[JumpCrisscross]

> I have given Starlink very optimistic numbers

Your revenue figures are consumer only. And while you're generous on utilization factor, we capitalised the TCO up front while amortising revenue, and then reduced asset tenure to worst case observed during development.

Flex up to 4 years, let $1mm TCO be paid up front and the rest amortised, and reduce utilisation to 80% ($80k/month revenue) and IRR shoots up to 73%. Take TCO to $3mm ($1mm up front, $2mm amortised), reduce utilisation to 75% and we're still over 20%.

> while "7 million subscribers" might sound impressive on first sight: This is the number of DSL connections subscribed to in the tiny country of Belgium. But for magical reasons Starlink is valuated at a price higher than if you would buy all of Belgium

Well, yes. Starlink connections are more profitable and you can't scale selling internet to Belgium into a Starshield defence contract. Or selling to airlines and cruise ships and yachts and mining operations, all of which pay more than a Belgian.

> some countries in the EU for example years ago created laws that say that whoever opens the street for any reasons has to put in empty tubes for someone to later put in fiber before closing the street again

Starlink doesn't sense in densely-populated areas of the EU or Asia. (And the equivalent for SpaceX would be ridesharing Starlink on someone else's flight.)

> Fiber is dirt cheap if you are allowed to use existing power poles for example

If you have the scale. You're underestimating the risk that comes from having to place infrastructure up front.

Your analysis is pretty solid. But I don't think it's taking into account the fact that you can build multibillion-dollar telecoms business on a few tens of millions of high-paying customers.

[Fischgericht]

I guess we can agree that the comparison between Sat internet and physical links depends a lot on the physical situation in the target region, and the regulatory frame work.

And please keep in mind that while you are right that there is a risk investing into physical infrastructure also applies to Starlink. It's worth remembering here that all Sat Internet companies prior to Starlink had failed and needed to be rescued with tax payer money.

I don't have exact numbers, and it's a bit muddy due to state subsidiaries, but in Germany the average cost to connect a subscriber in a medium density town with fiber, with given that nothing was prepared and you have to open the street etc appears to be in region of €/$ 2,000 or so.

I don't know if that is done in the US, but also in Europe we now do "trenching". It has some downsides and pitfalls, but this reduces the upfront infrastructure cost for fiber massively.

[JumpCrisscross]

> while you are right that there is a risk investing into physical infrastructure also applies to Starlink

Absolutely. It's why I think assuming the WACC of a highly-leveraged telecom (around 10%) is appropriate.

> this reduces the upfront infrastructure cost for fiber massively

Fibre makes sense where there is density. It's higher capacity and cheaper. That doesn't mean it makes sense everywhere. And a lot of that everywhere will pay a lot of money for connectivity.

The global telecom market generates trillions of dollars of annual revenue [1]. There is a lot of fruit for the picking.

[1] https://www.grandviewresearch.com/industry-analysis/global-t...

[stronglikedan]

> Was this AI generated?

It's crazy to me that people use AI to generate comments for social sites of all things, but here we are.

[Fischgericht]

I find it even more crazy that you no longer can comment on HN without someone trying to invalidate valid points by claiming you not being human. :)

To be honest, while I took it lightly, others might feel pretty insulted by such claims. De-humanizing someone stinks.

[JumpCrisscross]

> you no longer can comment on HN without someone trying to invalidate valid points by claiming you not being human

I made this mistake, but I'll defend it by pointing out that I've gone a few comments deep on HN, thinking through and citing and engaging in good faith, only to realise I wasn't talking to a human but to a bot. (Then the commenter gets defensive about using a bot, hallucinations and all.)

Instead of taking it as a personal insult, maybe interpret it as your comment having inspired someone to engage effortfully with what you said.