Inter-Satellite Links: How LEO Constellations Route Data Without Touching Earth

The single most important architectural fact about a modern LEO constellation is invisible from the ground: its satellites talk to each other. Inter-satellite links — laser beams connecting spacecraft moving at nearly 8 km/s, thousands of kilometers apart — turn a swarm of relays into a routed mesh network that can carry a packet across an ocean without it ever touching Earth until the destination. Understanding inter-satellite links for LEO is understanding why the second-generation constellations are a categorically different product from the bent-pipe systems they replaced.

On this page

• Bent Pipe vs Mesh: The Architecture That Changed
• Why Lasers, Not Radio
• The Pointing Problem
• Mesh Routing: Coverage and Resilience Rewritten
• The Defense Standard That Made a Market
• The State of Throughput: 25 Gbps Now, 100 on the Roadmap
• Industry Implications
• What to Watch

This deep dive covers what an ISL actually is and why lasers won over radio, the brutal engineering of pointing a beam between two fast-moving satellites, how mesh routing rewrites coverage and resilience, the defense standard that turned a lab technology into an interoperable market, and the honest state of throughput today versus the roadmaps. The argument: ISLs are the feature that converts LEO from a coverage technology into an infrastructure one.

Bent Pipe vs Mesh: The Architecture That Changed

Traditional satellites are bent pipes: a signal goes up to the satellite and straight back down to a ground station within the same footprint, with the satellite acting as a mirror. That works only where you can build a gateway — which excludes most of the ocean, the polar regions, and every country that will not grant landing rights. A bent-pipe constellation’s coverage is really its ground-station map.

ISLs break the dependency. When satellites can hand traffic to their neighbors, a packet can hop across the constellation — satellite to satellite to satellite — and descend only where a gateway or the destination user actually is. The constellation becomes a router in the sky, and its coverage becomes the sky itself rather than the list of places an operator could build ground infrastructure. That is the whole difference between a relay and a network. [INTERNAL LINK: what is low earth orbit → the constellation architecture ISLs complete]

Why Lasers, Not Radio

Inter-satellite links can use radio, and some do — but the second-generation broadband constellations chose optical, for reasons that compound. Optical carriers operate at frequencies hundreds of thousands of times higher than radio, which means vastly more bandwidth per link. A tight laser beam concentrates its energy into a pencil rather than a floodlight, so it needs less power for the same data and is nearly impossible to intercept or jam. And crucially, laser links use no licensed radio spectrum, sidestepping the ITU coordination that governs every RF transmission — a regulatory free lunch in an industry where spectrum is the scarcest asset. [INTERNAL LINK: ITU spectrum allocation LEO → why avoiding spectrum coordination matters]

The price of those advantages is precision. A radio beam forgiving enough to spray across a wide arc is easy to point; a laser beam narrow enough to deliver its bandwidth advantage must hit a target the size of a dinner plate thousands of kilometers away, from a platform moving at orbital velocity. Optical won the bandwidth argument decades ago. It only won the commercial argument once the pointing problem got solved at production cost.

The Pointing Problem

Consider the geometry. Two satellites, each moving at roughly 7.6 km/s in different directions, must establish and hold a laser link across a gap that can approach 4,000 km. At that range, an aiming error of a tiny fraction of a degree misses entirely. The link must be acquired (the terminals find each other from their known orbits), tracked (each follows the other as the geometry changes continuously), and held stable against spacecraft vibration, thermal flexing and the relativistic point-ahead required because the target has moved during the light’s travel time.

This is why ISL terminals are precision instruments: fast steering mirrors, fine-tracking sensors and control loops that hold the beam within microradians. For years these were exquisite, expensive, low-volume devices — the reason ISLs stayed a demonstration technology through the 2010s. The breakthrough was not a new physics but manufacturing: building optical terminals on a production line, at unit costs and volumes that let a constellation put three or four on every satellite. Once pointing became cheap and repeatable, the mesh became affordable, and the affordable mesh changed the industry.

Mesh Routing: Coverage and Resilience Rewritten

Once satellites route to each other, three capabilities appear that bent pipes cannot offer. Coverage extends to anywhere the constellation flies: mid-ocean vessels, polar research and shipping routes, and aircraft over open water get service with no local gateway, because their traffic rides the mesh to wherever the nearest gateway sits. Resilience improves because the mesh reroutes around a failed satellite or a downed gateway in milliseconds, the way a terrestrial network routes around a cut fiber. And jurisdictional independence emerges: traffic can enter and leave the network at gateways an operator controls, reducing dependence on any single country’s landing permissions — a strategic asset in every market-access negotiation. [INTERNAL LINK: Starlink regulatory market access → how mesh routing changes licensing leverage]

There is even a latency dividend on long routes: light travels faster through vacuum than through fiber, so a mesh path can, on sufficiently long hauls, beat terrestrial cable outright. That regime is narrow but real, and it is the only case in networking where satellite is the theoretical latency optimum rather than the compromise. [INTERNAL LINK: LEO vs GEO latency comparison → the vacuum-advantage analysis in full]

The Defense Standard That Made a Market

Optical terminals were nearly useless commercially as long as every vendor’s product spoke a different dialect — a satellite with a Company A terminal could not link to one carrying Company B’s. The US Space Development Agency broke the logjam for the whole industry by publishing an Optical Communications Terminal standard and mandating it across its proliferated architecture. The SDA deliberately set a modest data rate — 2.5 Gbps — precisely to widen the vendor field rather than favor the fastest terminal, prioritizing interoperability over raw speed.

The effect was catalytic. With a common standard, terminals from different suppliers became interchangeable, defense demand underwrote production volume, and a dozen-odd Western vendors now ship optical terminals at industrial cadence — Tesat-Spacecom’s SCOT family, CACI’s long-flown CrossBeam, Mynaric’s CONDOR line among them. For any vendor selling into the US defense market, conformance to the SDA standard has become the gatekeeper; successive versions have tightened the interoperability mandate with each satellite tranche. Standardization did for optical terminals what it once did for shipping containers: turned a bespoke device into a commodity interface, and a commodity interface into an industry.

The State of Throughput: 25 Gbps Now, 100 on the Roadmap

Honest numbers matter here, because the marketing runs ahead of the fleet. Starlink, which builds its terminals in-house, specifies up to 25 Gbps per link at ranges near 4,000 km, with higher rates possible over shorter hops — and with roughly 9,000 of its satellites carrying three to four terminals each by mid-2026, it operates by far the largest optical mesh ever built. Commercial and defense terminals from the merchant vendors cluster lower, from the SDA’s interoperable 2.5 Gbps up through 10 Gbps-class LEO products, with 100 Gbps-class terminals firmly on development roadmaps rather than in wide service.

Terminal / class	Data rate	Status	Typical role
SDA OCT standard	2.5 Gbps	Fielded across defense tranches	Interoperable government mesh
Merchant LEO terminals (e.g. SCOT-class)	~10 Gbps	In production	Commercial and defense meshes
Starlink in-house	Up to 25 Gbps @ ~4,000 km	Fielded at fleet scale	The largest operational mesh
Next-generation (e.g. 100 Gbps-class)	~100 Gbps target	Development / roadmap	Future high-capacity backbones

The strategic reading: the current bottleneck is not any single link’s speed but the number of terminals per satellite and the routing intelligence that stitches them together. A constellation’s mesh capacity is a network-design problem — topology, terminal count, onboard routing — as much as a laser-power one, which is why the operators with the most software depth, not just the fastest optics, are pulling ahead.

One correction worth carrying into any technical conversation: the “100 Gbps per link” figure that circulates in industry shorthand describes the frontier, not the fleet. The operational mesh carrying real traffic in 2026 runs at 25 Gbps at the top end and single-digit-to-10 Gbps across the interoperable merchant market. The hundred-gigabit terminals are coming — the roadmaps are credible — but conflating a roadmap target with fielded capacity is exactly the kind of error that separates analysis from enthusiasm in this sector.

Frequently Asked Questions

What is an inter-satellite link?

A communications link directly between two satellites, letting them exchange data without relaying through a ground station. In modern LEO constellations these are laser (optical) links, which turn the constellation into a routed mesh network that can carry traffic across the globe in orbit.

Why do LEO constellations use lasers instead of radio for ISLs?

Lasers offer far more bandwidth, need less power for a given data rate, resist interception and jamming, and — critically — use no licensed radio spectrum, avoiding ITU coordination. The trade-off is pointing precision: hitting a small target thousands of kilometers away from a fast-moving platform, which is why the technology only scaled once terminals became mass-manufacturable.

How fast are inter-satellite laser links today?

Starlink’s in-house terminals specify up to 25 Gbps per link at ranges near 4,000 km. Merchant terminals range from the SDA’s interoperable 2.5 Gbps up to roughly 10 Gbps for LEO products, with 100 Gbps-class terminals on development roadmaps rather than in wide service as of mid-2026.

Do all satellite constellations have inter-satellite links?

No. Some operate as bent pipes, relaying traffic straight back to a ground station within the satellite’s footprint, which limits coverage to where gateways can be built. ISLs are what enable ocean, polar and gateway-hostile coverage — so their presence or absence is a defining architectural difference between constellations.

Why does the SDA optical standard matter commercially?

By mandating a common Optical Communications Terminal standard at a deliberately modest 2.5 Gbps, the US Space Development Agency made terminals from different vendors interoperable and underwrote production volume. That turned a bespoke lab device into a merchant-market component and seeded the dozen-odd suppliers now shipping terminals at scale.