To add on the costs (yeah I have nothing better to do)
- ground station: must be negligible compared to everything else. They are literally planting giant routers out in a farm, need to pay rent, power and cabling to an exchange. Zero other infrastructure costs
- internet bandwidth: despite Netflix's attempts, ISPs have peering agreements that are usually free, as does Starlink
- satellite production: nobody knows, but they have said the payload is cheaper than the launch, so somewhere between $250-500k each at start
- customer service: online only, seems like the dish is self-diagnosing, will likely end up being much lower than traditional broadband
> - ground station: must be negligible compared to everything else. They are literally planting giant routers out in a farm, need to pay rent, power and cabling to an exchange. Zero other infrastructure costs
Fair enough, though you're neglecting the costs for building this all over the world, if they want to get access to any size of market.
> - internet bandwidth: despite Netflix's attempts, ISPs have peering agreements that are usually free, as does Starlink
Actually, according to Cloudlflare[0], you still need to buy a significant amount of traffic through transit agreements - cloud flare pays for 40% of its traffic in Europe (at a price they don't disclose), 60% in the USA (same price as in Europe), 40% in Asia (at 7 times the European price), 10% in Africa (at 14 times European price), 40% in South America (at 17 times European price), and 50% in Oceania (at 17 times European price). Only the Middle East region is 100% peered for them.
So the cost to actually deliver internet is definitely not going to be 0.
> - satellite production: nobody knows, but they have said the payload is cheaper than the launch, so somewhere between $250-500k each at start
So that is another 15M$ per launch at their declared aspirational cost (or an extra 100M$ per launch if you believe the Starship numbers). Just maintaing the entire constellation today would cost ~100M$/year, assuming almost all satellites reach their 5 year life span (current failure rate is 6% after one year).
> - customer service: online only, seems like the dish is self-diagnosing, will likely end up being much lower than traditional broadband
You can't sell online only customer support for access to the Internet if you want to reach any kind of realistic market. Furthermore, customer service more broadly includes delivery, sales, marketing, training materials; and this will be required all over the world, since they must target customers in remote places with little or no acces to the internet today.
Fair points. Though I’m not sure cloudflare is a good reference - they are a content provider and will pay for a competitive advantage (latency). The post mentions 50% of their costs come from 6% of traffic, they could simply choose to not pay for transit and let these be rerouted elsewhere.
Once Starlink has the satellite “mesh” operational they would probably become a tier-1 ISP and have zero transit costs. Maybe even sell traffic to other ISPs.
Sure, that's a good point, and the Cloudflare article is from 2016, so the market may have advanced as well.
On the other hand, unless they somehow get the laser communication working, which is still a research problem despite their optimism, they have way too high latency to act as a tier 1 ISP - all traffic currently goes consumer - 1 satellite - ground station - Internet.
- ground station: must be negligible compared to everything else. They are literally planting giant routers out in a farm, need to pay rent, power and cabling to an exchange. Zero other infrastructure costs
- internet bandwidth: despite Netflix's attempts, ISPs have peering agreements that are usually free, as does Starlink
- satellite production: nobody knows, but they have said the payload is cheaper than the launch, so somewhere between $250-500k each at start
- customer service: online only, seems like the dish is self-diagnosing, will likely end up being much lower than traditional broadband