Last August, a global outage across SpaceX's Starlink satellite network lasted nearly an hour. For most users it was an inconvenience — a lost Zoom call, a buffering video, a momentary lapse in connectivity in some rural county that rarely gets it anyway. For the U.S. Navy, it was something else. Unmanned surface vessels participating in tests off the California coast went effectively dark, their communications severed, their autonomous operations halted. The drones sat in the water, waiting, until Starlink came back online.
The incident was not classified. It surfaced in reporting by Reuters and was subsequently confirmed by defense sources. And while a few hours of lost connectivity during a test exercise is far from a catastrophe, the episode crystallized something the Pentagon has been managing quietly for years: the U.S. military has built significant operational dependencies on a commercial satellite network it neither owns, controls, nor can replicate on any timeline relevant to near-term conflict.
How the Dependency Grew
The military's entanglement with Starlink did not happen through a single procurement decision. It accumulated gradually, driven by Starlink's extraordinary performance advantages over any government alternative. With more than 10,400 satellites in low Earth orbit as of mid-2026 — a number SpaceX continues to grow — Starlink offers low-latency broadband connectivity in locations where no military communications satellite could realistically serve. The terminals are cheap, rugged, and increasingly small. Starship-launched batch deployments have compressed the cost per satellite to a fraction of any government program.
Ukraine demonstrated what Starlink can do in a contested environment. SpaceX delivered terminals within days of Russia's full invasion in February 2022, and the network became the backbone of Ukrainian military communications in the early months of the war — enabling drone operations, artillery coordination, and command communications across a front line where terrestrial infrastructure was being systematically destroyed. The U.S. military watched this performance closely, and what it saw was a capability it could not replicate through its own systems.
Commercial Starlink contracts have since expanded across multiple service branches. The Army has integrated terminals into ground vehicle programs. The Air Force uses Starlink for airborne connectivity. The Navy's unmanned surface vessel programs — the Ghost Fleet Overlord initiative among others — have been architected around commercial low-Earth-orbit connectivity. SOCOM has fielded Starlink terminals in special operations units. The accumulation is vast, and it happened faster than any formal dependency analysis could track.
The Strategic Exposure
The problem is not that SpaceX is unreliable. By commercial satellite standards, Starlink has an impressive availability record. The problem is structural. Starlink is a private company with a single controlling shareholder whose public statements on foreign policy, military support, and government cooperation have introduced uncertainty into what should be a predictable defense relationship. Elon Musk's 2022 decision to limit Starlink coverage near Crimea to prevent a Ukrainian submarine drone attack on Russian ships — a decision he disclosed publicly and defended as deliberate — demonstrated that SpaceX retains operational discretion over its network's use in ways that no military commander should have to factor into planning.
There is also the question of adversary targeting. China and Russia have both demonstrated anti-satellite capabilities against higher-orbit systems, and both have invested in directed energy and electronic warfare systems explicitly designed to counter low-Earth-orbit constellations. Starlink's distributed architecture — thousands of satellites rather than a handful of expensive high-orbit systems — provides significant resilience against kinetic ASAT attacks on individual nodes. But the ground stations, the user terminals, and the radio frequencies Starlink uses are all targetable. A coordinated electronic warfare campaign against Starlink terminals in a conflict zone would not need to destroy a single satellite.
The Navy drone test outage was caused by a software issue in SpaceX's ground network — not a cyberattack, not a hardware failure, but the kind of routine software deployment problem that any complex distributed system occasionally produces. In peacetime testing, the consequence was a disrupted exercise. In a shooting war, the same scenario could mean unmanned vessels unable to receive commands at a tactically decisive moment.
What the Pentagon Is Doing About It
The DoD's response is multi-layered, though critics argue it is not moving fast enough. The Space Development Agency's Transport Layer — a government-owned low-Earth-orbit communications constellation — is being built explicitly to provide military-controlled broadband connectivity that replicates some of Starlink's capabilities under Pentagon authority. Initial operational capability was projected for 2024 and has slipped, but the program continues. Several dozen Transport Layer satellites are now on orbit, with hundreds more planned.
The military is also pursuing resilience through redundancy: contracting with multiple commercial providers — Amazon's Project Kuiper, Viasat, and others — to avoid single-vendor dependency. This approach accepts that commercial networks will remain central to military communications while attempting to ensure that no single provider's outage or policy decision can be operationally decisive.
Space Force has established a Commercial Space Office specifically to manage the growing entanglement with commercial providers, develop standardized contractual frameworks, and assess vulnerabilities in commercially-dependent systems. The office is new, and its influence on procurement decisions is still developing.
The Harder Question
None of these measures fully resolves the underlying tension: the U.S. military needs Starlink's capabilities now, today, and no government program will deliver a comparable system on any timeline that matters for near-term competition with China and Russia. The Transport Layer will eventually mature. Kuiper will eventually launch. But in the interim, the military will continue flying unmanned drones, coordinating special operations, and processing battlefield data through a network that one man in Starbase, Texas ultimately controls.
The California drone test outage was a minor incident. The lesson it carries is not minor at all.
