Last August, a global Starlink outage knocked out communications with two dozen Navy robot boats drifting off the California coast. For nearly an hour, the drones — designed to operate without a crew aboard — had no way to receive commands or report back. That is the single point of failure at the center of the Pentagon's autonomous weapons strategy, and it happened in plain sight.
The outage, documented in internal Navy safety reports reviewed by Reuters, is the strongest evidence yet that the U.S. military has built part of its next-generation drone fleet around a dependency it cannot easily replace. The drone boats involved were made by Saronic, an Austin-based startup, and BlackSea, a Maryland company — both of which advertise their systems as capable of extended autonomous operations at range. The problem is that "autonomous" in this context still means: dependent on a commercial satellite network owned by one company, with no proven backup system.
The network is Starlink, operated by SpaceX. It has close to 10,000 satellites in low-earth orbit and provides the kind of global coverage no other system can match right now, according to Reuters. The U.S. military has leaned into it heavily — not just Starlink's consumer product but also Starshield, SpaceX's national security-focused variant built for government customers. The coverage is global and the price is right. That is also the problem.
Elon Musk, SpaceX's CEO, has already demonstrated what that dependency looks like in practice. During Ukraine's 2025 counteroffensive, soldiers panicked and surveillance drones went dark after Musk ordered Starlink shut off over territory Russian forces had retaken, Reuters reported. The decision affected Ukrainian troops in the field, not some distant policy debate.
SpaceX is now preparing for a public offering that Reuters reports could value the company at $2 trillion — the largest in history. When that happens, the Starlink dependency goes from a back-channel defense procurement issue to a permanent, public fact in the investor documents. The Pentagon has accepted this risk, and accepted it knowingly.
Bryan Clark, who studies autonomous warfare at the Hudson Institute, argues the trade-off is rational. "You accept those vulnerabilities because of the benefits you get from the ubiquity it provides," he told Reuters. The U.S. has no alternative that matches Starlink's coverage right now. Amazon closed an $11.6 billion deal for Globalstar in April 2026, according to Reuters, but building a comparable global network will take years. Clayton Swope, who runs the Aerospace Security Project at CSIS, put it plainly: without Starlink, the U.S. government would not have access to a global low-earth orbit communications constellation at all.
The August outage was not the first sign of strain. Navy tests in California in April 2025 also reported connectivity problems — Starlink struggled to maintain solid network connection when multiple drone boats and flying drones were running simultaneously, with high data demand from coordinating all the systems at once. That is a separate problem from a global outage: the network works, but it degrades under the weight of the very fleet it is supposed to support.
What the Navy bought, in other words, is a drone fleet that cannot function without one company's satellite network — and that company's CEO has already cut service during an active war. The IPO will make that dependency public and permanent.
