The first five installments mapped an internet made of cables, cuts, buildings, megawatts, and caches. The recurring promise against that map is an escape: an internet that arrives from the sky, above the ground layer entirely, or a mesh that routes around it from the ground up. This installment follows that promise to its physics and its ledgers. The counter-layer is real — and it is still the physical internet, with the bottleneck moved rather than removed.
The old satellite internet lives far away. A geostationary orbit is a circular orbit 35,786 km above Earth's equator, moving with the planet's rotation so it hangs fixed in the sky.[6]
The distance is the problem: "it takes about 240 ms for a signal to pass from a ground based transmitter on the equator to the satellite and back again."[6]
That is one leg; a full round trip through a geostationary system runs "approximately 550 milliseconds of latency round-trip time," in the physics alone, before any terrestrial network is involved.[2]
The propagation delay for a round-trip internet protocol transmission "can be over 600 ms" via a geostationary satellite — "but as low as 125 ms for a MEO satellite or 30 ms for a LEO system."[7]
Low Earth orbit is the geometry that fixes it: the LEO region is space "below an altitude of 2,000 km (1,200 mi) (about one-third of Earth's radius)."[5] The counter-layer's whole pitch is in that contrast — bring the satellite from 35,786 km down to a few hundred, and the physics stops dominating the ping.
The counter-layer is usually pictured as the internet arriving from the sky. The record is blunter. Starlink's network consists of satellites "that communicate with user terminals and gateway ground stations connecting the constellation to terrestrial internet infrastructure."[1]
The constellation does not replace the physical internet; it plugs into it. Every orbital network depends on ground stations, and the classic satellite architecture is a star: "the satellite operates a Star network topology where all network communication passes through the network's hub processor."[2] The sky layer is a door into the ground layer, not a parallel world.
The economics of constellations were tested before the rockets got cheap, and they failed in the open.[3]
Iridium launched its service on November 1, 1998; "On August 13, 1999, nine months after the launch of the organization, the founding company went into Chapter 11 bankruptcy."[3]
The fire sale is the monument: the company was sold from US bankruptcy court for $25 million in March 2001, and that "erased over $4 billion in debt."[3]
"The initial commercial failure of Iridium had a damping effect on other proposed commercial satellite constellation projects, including Teledesic."[3]
The lesson was not that the sky was impossible. The cost was named in the record: the handsets, "when compared to terrestrial cellular mobile phones were bulkier and more expensive," and "the cost of service dissuaded many potential users."[3]
A constellation's cost lands up front, while its subscribers accrue slowly.
OneWeb repeated the shape two decades later: it "entered bankruptcy in March 2020 after failing to raise the required capital to complete the build and deployment of the remaining 90% of the network."[4] The failure was not orbit physics. It was the capital curve: 90% of the network still unbuilt, and the money gone.[4]
What changed between Iridium and Starlink is on the launch ledger. Iridium's own next-generation campaign signed the deal that names the shift: a $492 million contract "designating the Falcon 9 as a major provider of launch services for the Iridium NEXT campaign, becoming the largest single commercial launch deal ever signed (simultaneously representing a benchmark in cost-effective satellite delivery to space)."[3] The campaign's "total expenditures of approximately $2.9 billion" were nonetheless "partially financed with debt."[3]
Starlink is the same idea after reusable rockets made it survivable: SpaceX estimated in May 2018 "the cost of designing, building, and deploying the constellation to be at least US$10 billion."[1] As of June 2026 the network held "approximately 10,413 satellites (10,397 operational) in low Earth orbit" and "more than 12 million subscribers," roughly "75% of all active maneuverable satellites in Earth orbit."[1] By the end of 2025 Starlink was "SpaceX's largest business segment, generating $11.4 billion in revenue and $4.4 billion in operating income."[1]
Even the winner, though, shows where the terrestrial dependency bites. The FCC awarded Starlink $885.5 million in rural-broadband subsidies, then revoked them in August 2022 "with the FCC stating that Starlink 'failed to demonstrate' its ability to deliver the promised service."[1] The satellite that was supposed to reach everywhere could not reach the specific ground it had promised.[1]
Escape narratives treat orbit as a frontier with room for everyone. Geostationary orbit is the opposite: satellites "must all occupy a single ring above the equator," and the requirement to space them apart "means that there are a limited number of orbital slots available, and thus only a limited number of satellites can be operated in geostationary orbit."[6] The scarcity is administered — the ITU allocates slots under the Radio Regulations, and the disputes are old enough to have names: in the 1976 Bogota Declaration, eight equatorial countries claimed sovereignty over the geostationary orbit above their territory.[6] The counter-layer has its own property law.[6]
The other escape route never leaves the ground. A wireless mesh is "in effect, a network of routers minus the cabling between nodes" — peer radios that forward packets hop by hop.[8] It is real: "a number of wireless community networks have been started as grassroots projects across the world."[8]
And the sky itself uses the trick — "some communications satellite constellations operate as a mesh network, with wireless links between adjacent satellites," which "allows for the constellation to operate with far fewer earth stations than would be required for an equal number of traditional communications satellites."[8]
The counter-layer does not escape the physical internet; it relocates the point where physics is paid. The satellite replaces the last-mile cable, then borrows everything else — gateways, spectrum, slots, terrestrial backhaul — from the layer it promised to bypass.
The series' map survives the sky: the internet is not a substance that can be delivered from nowhere, but a chain of bottlenecks, and the counter-layer is a new set of them. Iridium and OneWeb died on the capital curve; Starlink survived on a launch-cost curve that was itself physical — a rocket that lands and flies again.[3]
And the fastest escape is the most local one: mesh forwards in short hops, so it crosses a neighborhood, not an ocean.[8] A GEO satellite crosses the ocean but answers slowly;[6] LEO is the compromise that returned the internet to the ground, one gateway at a time.
[1] Wikipedia: Starlink
[2] Wikipedia: Satellite Internet access
[3] Wikipedia: Iridium Communications
[4] Wikipedia: Eutelsat OneWeb (OneWeb)
[5] Wikipedia: Low Earth orbit
[6] Wikipedia: Geostationary orbit
[7] Wikipedia: Satellite constellation
[8] Wikipedia: Wireless mesh network