There is a basic fact of physics sitting at the heart of the current satellite internet arms race: light moves through vacuum roughly 30 percent faster than it moves through fiber optic glass. The glass slows it down. Vacuum does not.

That gap — about 200,000 kilometers per second in fiber versus 300,000 in open space — is the entire premise behind optical inter-satellite links, or OISLs. It is why the two largest broadband constellations in orbit are each building a laser-based backbone in the sky, and why a small but consequential ecosystem of terminal manufacturers has grown up around them. Whether or not any individual operator can turn a profit, the underlying architecture is no longer a differentiator. It is becoming the cost of entry.

What the Mesh Actually Does

A LEO satellite without inter-satellite links is, in network terms, a relay. It receives traffic from a ground station, forwards it to a user terminal, and hands off return traffic back to the ground. Every packet must touch the earth twice. For a user sitting near a ground station, that is fine. For a user in the middle of the Pacific, it means routing through whatever gateway happens to have line of sight — which can introduce large detours.

OISLs change the topology. Satellites with laser links can hand traffic to neighboring satellites in the same orbital plane or across adjacent planes, building a persistent mesh that spans the globe. A packet from London to Tokyo no longer needs to find a ground station mid-route; it hops across the mesh and descends near its destination. The path is shorter, the routing is more direct, and the physics of propagating light through vacuum means each hop is faster than an equivalent length of fiber would be.

The practical payoff is most visible on long transoceanic and transcontinental routes — exactly the routes where fiber latency accumulates most. For financial trading platforms, real-time video, and military applications requiring tight timing, the difference is measurable and meaningful.

Starlink did not launch with ISLs. Early shells of the constellation relied on ground stations for all routing, which limited coverage over oceans and remote high latitudes. When SpaceX began integrating laser terminals into production satellites — initially for the polar orbital shell — it represented a quiet architectural upgrade, one that required no change to user hardware but substantially expanded what the network could do.

According to SpaceX’s own technology page, each Starlink satellite now carries three optical ISLs operating at speeds up to 200 Gbps, forming what the company calls a global internet mesh. The operational scale of the constellation means there are thousands of these links active at any moment, effectively making the Starlink network the largest free-space optical communications system ever deployed.

That scale also surfaced a subtlety. Research published by the Internet Society in 2024 found that ISL routing can, counterintuitively, increase latency for some customers, because traffic sometimes gets forwarded through distant ground stations rather than descending locally. Starlink engineers acknowledged the behavior: the routing system optimizes across the whole mesh, not just for the shortest path to the nearest gateway. The implication is that ISLs change but do not eliminate the complexity of LEO routing — they add more degrees of freedom, which requires more sophisticated traffic management.

Amazon’s All-In Commitment

When Amazon’s Project Kuiper launched its two prototype satellites in October 2023, one of the key experiments was validating OISLs in orbit before committing them to the full production design. The result, announced in December 2023, was a sustained 100 Gbps optical link between KuiperSat-1 and KuiperSat-2 across roughly 1,000 kilometers of space. Amazon called it the final validation needed to confirm the communications architecture.

The significant detail in that announcement was the commitment: every production Kuiper satellite will carry multiple optical terminals from launch. There would be no phased rollout, no early operational period without the capability. Amazon framed the 30 percent latency edge over equivalent fiber distances as a selling point, particularly for government and enterprise customers who care as much about path predictability as raw throughput.

That all-in posture makes strategic sense for a late entrant. Starlink has several years of operational data and consumer brand recognition that Kuiper cannot replicate quickly. But the architectural decision to ship OISLs universally means that the Kuiper mesh, once it reaches minimum viable density, will behave like a coherent global network from the start, rather than an uneven patchwork with gaps over the oceans.

The Vendor Ecosystem and the Standards Problem

Outside of the two mega-constellations, a specialized market has developed around laser terminal hardware. A handful of vendors — including Germany’s Tesat-Spacecom, Mynaric, CACI/SA Photonics, and Skyloom — have completed interoperability testing against the U.S. Space Development Agency’s Optical Communications Terminal standard, the closest thing the industry has to an interoperable interface specification. Kepler Communications validated an in-orbit link between its Pathfinder data-relay satellites using Tesat’s SCOT-135 terminals compatible with the SDA standard.

Mynaric, which began volume production of its CONDOR Mk3 terminal in early 2024, encountered yield and supply chain problems that delayed deliveries through mid-year — a reminder that manufacturing optical terminals at satellite volumes is genuinely hard. The physics of pointing a laser beam accurately between two objects each moving at roughly 7.5 kilometers per second, while compensating for vibration and thermal expansion, demands tolerances that standard electronics manufacturing was not designed to meet.

SpaceX complicated the picture in March 2024 when Gwynne Shotwell announced that the company would begin selling its laser link technology commercially to other operators. The announcement drew a pointed response from some corners of the industry. ALL.SPACE’s chief executive warned that adopting SpaceX’s proprietary terminal risked undermining the open interoperability frameworks championed by the SDA and the Consultative Committee for Space Data Systems. The concern is a real one: if multiple LEO constellations end up running incompatible laser protocols, the economic case for a genuinely interoperable space-based backbone weakens significantly.

Why Architecture Is Now a Differentiator — Until It Isn’t

The OISL race illustrates a recurring dynamic in infrastructure buildouts: a capability that starts as a performance differentiator gradually becomes a baseline expectation, at which point competition shifts to execution quality, pricing, and scale.

Starlink demonstrated that a LEO broadband constellation can deploy OISLs across thousands of satellites. Amazon Kuiper committed to shipping the capability universally rather than incrementally. Telesat Lightspeed, Rivada Space Networks, and others are each working through their own terminal selections. The question is no longer whether a serious LEO network needs laser links — it does — but whether the latency physics advantage translates into durable business value versus undersea fiber and terrestrial 5G.

The speed of light in vacuum is a constant. How operators turn that constant into revenue remains, for now, an open problem.

Sources