The old idea
In 1880 Alexander Graham Bell transmitted speech on a beam of sunlight, using a vibrating mirror and a selenium receiver. He called the photophone his greatest invention, greater, he said, than the telephone. It was useless at the time, because the atmosphere and the technology were not ready, and it took a century for the laser and the optical fibre to prove him right. Optical communication in space is the same idea with the atmosphere removed, which is the one place it was always going to work best.
The physics
A beam spreads at roughly the wavelength divided by the aperture. That single relationship explains why this layer went optical.
The price is that you must hit a 15-metre spot from 1,000 kilometres away, from a vibrating spacecraft, while both
ends are moving at 7.5 km/s. Optical terminals are precision mechanisms with lasers attached, and that is why they are a product rather than a component.
The constraint nobody budgets for
Figure 14.1 — A data centre that cannot reach the ground is an archive
Chapter 7 showed that an orbital compute node’s traffic is modest in bandwidth terms. The problem is not how fast you can move data. It is how often you are allowed to. A node at 650 km passes over a given mid-latitude ground station for roughly ten minutes at a time, a handful of times a day. That is on the order of four percent of the day. A polar station sees a sun-synchronous satellite on nearly every orbit, which helps considerably. A ten-station global network gets you to something like a third. Only an inter-satellite relay mesh, where your traffic hops across the constellation to whichever node currently has a ground contact, gets you to continuous service. So the relay layer’s value is not bandwidth. It is duty cycle, and duty cycle is the difference between selling a cloud service and operating an archive. That is a twenty-fold multiplier on the usefulness of the same hardware, and it is why this layer is strategically more important than its revenue currently suggests. The other half of it is capital nobody counts. A geographically diverse ground network, sited against cloud cover, licensed, staffed, backhauled is a real infrastructure programme. When someone quotes you dollars per GPUhour in orbit, ask whether the ground segment is in the number. It usually is not.
Who is attacking it
Kepler Communications operates a deployed optical relay constellation in low Earth orbit, running as an IP mesh, with compute and storage on the relay satellites themselves and a roadmap to substantially higher link rates. Commercial orbital data-centre nodes have already ridden and used it.12 It is simultaneously the backhaul for orbital compute and, because its relays carry accelerators, a distributed compute platform of its own. Skyloom and the other terminal suppliers sell the mechanism. Starlink operates by far the largest laser mesh in existence but does not sell open relay to third parties and that, rather than any technical risk, is the live commercial question in this layer.
The investment stance
This is the cleanest picks-and-shovels position in the sector: a deployed network, real switching costs, revenue from customers who are not dependent on orbital compute succeeding, and a service every operator needs regardless of which operator wins. The risk is specific and worth stating rather than glossing: if SpaceX opens its laser mesh as a commercial service, pricing in this layer compresses immediately. It would not eliminate the incumbents, defence traffic
will not route through a commercial mesh, and sovereign customers will pay for alternatives but it would change the margin structure permanently. I hold the layer with that risk explicitly priced rather than assumed away.
What to watch
Any signal that Starlink laser links become purchasable by third parties. Operational 100 Gbps-class optical relay, not demonstrated but in revenue service. Whether orbital compute operators build their own ground networks or buy contact time. That decision tells you where this layer’s margin ends up.
Chapters 15 to 17, in-space assembly and servicing, debris and insurance, and innovation outside the United States, complete Part IV in the next instalment.