When The Sky Goes Quiet, Who Governs The Machines? – Oped

Key Takeaways:

• A Reuters probe this week maps U.S.–China prep for orbital conflict; the author says the first pain may hit Earth machines. Day-one Ukraine: a cyber hit on Viasat’s KA-SAT left 5,800 Enercon turbines (11 GW) without remote control—they kept spinning locally. 24 July 2025 Starlink outage (~2.5 hours, software) paused Ukrainian drone missions. To the machine, attack and accident look the same: the sky went quiet.
• Satellites now feed clocks, GNSS, telemetry and soon onboard compute (Starcloud H100, SpaceX’s million-sat filing, Google Suncatcher). People improvise; machines follow stale code or freeze. One control plane can flip thousands at once. UN space treaties, ITU and NIS2 cover objects and spectrum, not what a tractor or grid controller may do when the link dies.
• Five contract-ready rules: declare the orbital dependency; keep a local safe floor and shrink authority as confidence falls; prove real diversity, not two dishes on one constellation; set priority and notice duties; share cross-border outage reports. Scale rules to harm. Line: no essential machine should take its authority from space unless it can lose the sky.

The space-to-machine economy is arriving before its rules: satellites increasingly supply timing, positioning, connectivity and soon computation to automated systems whose failures can cross borders in seconds.

The new space race is usually described by looking upward: launch rates, orbital maneuvering, counterspace weapons and competing constellations. This week, a Reuters investigation documented how the United States and China are preparing for conflict to extend into orbit. Yet the most immediate consequences of a space confrontation may first appear far below it – inside turbines, drones, ships, farms, power grids and financial systems that increasingly depend on machines in the sky.

On the morning Russia launched its full-scale invasion of Ukraine, a deliberate cyberattack disrupted Viasat’s KA-SAT network. The apparent military target was Ukrainian communications. The effects did not remain in Ukraine. Across Central Europe, 5,800 Enercon wind turbines with eleven gigawatts of capacity lost remote monitoring and control, while tens of thousands of other terminals were affected.

The turbines kept turning. They had been engineered to continue in an autonomous operating mode when the satellite link disappeared. What vanished was the ability of operators to observe them, adjust them or shut them down remotely. That distinction matters. This was not simply a communications outage. A weapon aimed at one country’s wartime network altered the control relationship between another country’s energy machines and the humans responsible for them.

Three years later, the same structural lesson arrived without an attacker. On July 24, 2025, Starlink suffered a global outage lasting roughly two and a half hours. Independent network analysis pointed to a centralized, software-defined control-plane failure. In Ukraine, Starlink went down across the front; drone missions were disrupted and some combat operations were postponed. The cause was internal software rather than hostile action, but the dependent machines experienced the same event: the sky stopped answering.

For a machine at the end of the link, attack and accident converge at the interface. It does not know whether silence came from an enemy, a failed update, congestion, a commercial decision or a government order. It knows only that the timing signal, position fix, data feed or command path on which its behavior depends is no longer available.

These incidents reveal the emergence of a space-to-machine economy. Satellites were built principally to serve human users: television audiences, telephone callers, navigators and analysts. Increasingly, their direct consumers are automated systems. Satellite timing disciplines clocks inside electricity grids, telecommunications networks and markets. Satellite positioning steers tractors, ships, drones and vehicles. Satellite links carry telemetry and control traffic for pipelines, mines, offshore platforms and remote industrial plants.

Machine dependence is different from human dependence in three ways. A person who loses a feed notices and improvises; a machine executes whatever contingency its designers wrote – or continues on stale information if they wrote none. A person’s loss is usually local; thousands of machines sharing one control plane can change state in the same second. And a person can explain what service was lost. A machine’s dependency may remain buried in an architecture diagram, absent from contracts, regulatory filings and national risk registers until a failure exposes it.

Now computation itself is beginning to move into orbit. Starcloud launched an Nvidia H100 GPU in November 2025 and says its spacecraft trained a small language model in orbit the following month. In August 2026, the company raised $250 million at a $2.3 billion valuation while targeting an eventual constellation of 88,000 satellites and twenty gigawatts of orbital computing capacity.

The larger proposals are even more striking. In February, the US Federal Communications Commission accepted for filing SpaceX’s application for an orbital data-center system of up to one million satellites. Google’s Project Suncatcher envisions solar-powered satellites carrying its AI accelerators and linked by lasers; two prototype satellites are slated for launch in early 2027. These numbers are ambitions, not deployed capacity, and formidable problems of heat rejection, radiation, orbital debris, maintenance and cost remain.

Not every proposal will survive contact with physics or economics. The direction nevertheless matters. Terrestrial machines may soon depend not only on data relayed through orbit but on decisions computed there. A lost link would then interrupt more than communication; it could remove part of the decision infrastructure itself. At the same time, the leading systems will sit under the jurisdiction of a small number of states and the control of a small number of companies. For countries unable to build their own constellations, convenient access can become strategic dependence long before it is recognized as such.

Existing rules address important pieces of the problem, but not the interface as a whole. The United Nations space treaties allocate state responsibility and liability around national space activities and damage caused by space objects. The International Telecommunication Union coordinates frequencies and harmful interference. Europe’s NIS2 cybersecurity directive reaches operators of ground infrastructure supporting space-based services. These regimes are not irrelevant. They are simply centered on space objects, operators, networks and spectrum – not on what an automated system on Earth is permitted to do when its orbital dependency degrades.

That leaves an unowned governance gap between the satellite provider and the machine operator. Five rules would begin to close it, and governments do not need to wait for a new treaty before putting most of them into procurement contracts and sector regulation.

First, disclose the dependency. Every consequential automated system should identify which orbital services it uses, the functions they support, the maximum tolerable interruption and the behavior triggered by loss or corruption of the signal. A dependency that is not declared cannot be stress-tested, diversified or governed.

Second, require an autonomous floor and authority contraction. The German turbines continued safely because local controls did not require a live satellite connection. That should become a design requirement for critical machines, not a fortunate engineering choice. Safe mode must also mean that authority narrows as confidence falls. A vehicle operating on uncertain positioning, or an industrial controller working from stale telemetry, should not retain the same permissions it held when every input was trusted.

Third, test diversity rather than merely purchasing it. Critical functions need an alternative service, a terrestrial path or enough local capability to bridge an outage. Two terminals are not redundant if both depend on the same constellation, control plane, timing source or jurisdiction. Failover should be demonstrated in exercises under realistic disruption, not described in a slide deck.

Fourth, establish priority and continuity duties. When capacity degrades, someone decides which users remain connected. If the dependents include emergency services, military units, grids or transport systems, that choice cannot remain an improvised and invisible exercise of vendor discretion. Governments should establish priority classes in advance and require providers serving critical machines to maintain resilient control planes, report incidents, give notice of material changes and publish realistic recovery objectives.

Fifth, create a cross-border incident layer. The 2022 attack demonstrated that a disruption aimed at one country can alter machine control in another without crossing a physical border. States need a shared reporting and coordination channel for loss of essential space services, including enough technical information to identify common-mode dependencies. Over time, international law will also need a clearer account of service-loss harm: not only what happens when a space object falls, but what happens when an indispensable space service disappears and machines below act badly as a result.

These duties should be proportional. A consumer navigation app and a power-grid controller do not warrant identical rules. The trigger should be consequence and scale: whether loss, corruption or withdrawal of the orbital input could produce serious physical, economic, military or public-service harm, especially across many systems at once. The goal is not to regulate every satellite connection. It is to make common-mode machine dependence visible before it becomes a national emergency.

The geopolitics of space is still discussed mainly in terms of who owns the satellites, who can disable them and who controls the orbit. The next question is what authority those satellites quietly exercise through the machines that depend on them. A country can retain legal sovereignty over its territory while losing practical control of essential functions when a foreign constellation, provider or control plane goes silent.

The German turbines supplied one encouraging answer: keep operating locally, within safe bounds. Ukraine supplied another: diversify before an outage, not after it. Orbital computing raises the stakes because the sky may soon provide not merely the connection but part of the judgment flowing through it.

The rule should be simple: no machine essential to public life should receive authority from space unless it has been designed to lose the sky.