In early 2025, the terminals went quiet. Not destroyed — throttled. Reports surfaced that Starlink service to Ukrainian frontline units had been restricted, degraded, or cut, apparently without prior notice to the military that had come to depend on it. Ukraine's presidential adviser confirmed the disruption. Poland's foreign minister — whose government helps foot the bill — publicly questioned Elon Musk over what he described as political manipulation. No formal explanation followed. No scope was published. A constellation of thousands of satellites, owned by one company and answerable, ultimately, to one man, had reached into an active war and turned a dial.
The detail that should stop you cold is not the restriction itself. It is the silence around it. A critical military capability was adjusted the way a cloud provider reshapes a pricing tier — quietly, unilaterally, and with no channel for the affected party to object.
I have spent a decade watching this pattern migrate across systems. In 2017, as a student in Copenhagen, I manually audited the tokenomics of failed ICOs and learned that control is rarely where the whitepaper says it is. What Ukraine is now discovering, at the cost of battlefield efficiency, is the same lesson in a harder register: the infrastructure carrying your most important signals is only as sovereign as the entity that can switch it off. The lesson keeps arriving in new costumes, and we keep failing to recognize it.
To understand the stakes, you have to understand what Starlink became. It was never designed as a weapons system. When SpaceX activated coverage over Ukraine in 2022, it was a commercial broadband product — consumer terminals, a monthly subscription, a promise of low-latency connectivity anywhere under the sky. But wars consume whatever is available. Within months, Starlink had become the communications spine of the Ukrainian military: drone control links, artillery correction, command coordination, the quiet data layer beneath almost every modern battlefield function. The military came to depend on a commercial low-earth-orbit constellation because the alternatives — military SATCOM systems like WGS or AEHF — were scarce, expensive, and simply not available at the density Ukraine needed.
What emerged was a trap assembled from good intentions. Three parties, three roles, no alignment. Poland helps pay for the service. Ukraine uses it. SpaceX provides it. The payer has no control over the terms. The user has no alternative to fall back on. The provider has no obligation to continue. This is not a supply chain. It is a dependency wearing the costume of a partnership, and every dependency has an off switch somewhere.
The reflex in my own community is to treat Ukraine as a vindication. See? Centralized infrastructure has a kill switch. Decentralize everything. I understand the reflex. I have felt it. But it is too easy, and it is partly wrong. Vindication is a feeling, not an argument, and this moment deserves the argument.
Consider what actually happened. A private operator adjusted access to a service. The adjustment was legal, within the operator's contractual discretion, and technically trivial. No code was hacked. No protocol was forked. The system worked exactly as designed. That is the horror of it. The kill switch was not a bug that slipped through. It was a feature present since the first launch, waiting for the moment someone chose to use it.
The mechanism itself is worth understanding, because it is almost insultingly simple. A satellite constellation is not a magic field of connectivity; it is a network with a ground segment, a control plane, and an authentication layer. Terminals are provisioned, geofenced, and rate-limited by software running in data centers far from the battlefield. Restricting service does not require deorbiting a single satellite. It requires changing a configuration — the digital equivalent of a landlord changing the locks while you are still inside the apartment. The most consequential infrastructure of the twenty-first century can be turned against its users by a settings update.
This is where the slogan finally breaks against reality. Code is law, until the law breaks the code — and here, the law did not even need to try. The operator simply declined to run it. Decentralization, in the sense that matters, is not about whether a system can be shut down; everything physical can be shut down. It is about whether the people who depend on a system have any standing when someone decides to shut it down. Ukraine had none. Poland, which paid, had none. The users were, in the most literal sense, customers — and customers do not get a vote.
When I audited lending protocols in 2020, I learned to ask a question most token analyses skip: not "is this decentralized?" but "who holds the switches no one advertises?" Admin keys. Upgrade proxies. Multisig thresholds. Emergency pause functions. Every serious protocol has them, and most bury them three layers down in the documentation. The honest answer is that almost nothing is as decentralized as its marketing. The difference with Starlink is only that its switches are visible, because there was never any pretense of hiding them.
And here is the harder truth for anyone who believes blockchain will fix this. The kill switch on a satellite constellation does not live in software. It lives in physics, ownership, and jurisdiction. You cannot tokenize your way out of the fact that someone owns the satellites, someone holds the launch contracts, someone answers to a national regulator. A decentralized ledger can record who is allowed to speak. It cannot guarantee that the channel stays open. The layers beneath the ledger — spectrum, hardware, orbit, launch — are exactly where power concentrates, and they are precisely the layers crypto tends to wave away as "solved by the base layer." They are not solved. They are rented.
A decade of precedent supports the fear. When Amazon removed Parler from its servers in 2021, it demonstrated that cloud hosting is a political instrument. When payment processors cut off platforms they disliked, they showed that financial rails carry the same discretion. Starlink is the same logic escalated to the level of national defense — the difference is only that the consequences are now measured in artillery coordinates rather than app store revenue. The pattern is consistent: as more of the world's critical functions run on privately owned infrastructure, more of the world's politics is decided in private rooms.
This is why the Polish foreign minister's question matters more than the answer. He was not asking about satellites. He was asking who is accountable when a private actor holds a public good. The uncomfortable reply is: no one, or at least no one the affected party can reach. The provider is not a signatory to the alliance. The payer is not a party to the decision. The user is not a party to either. We built the temple, but forgot who the god is — and in this case, the god has a board seat, a launch manifest, and a personal account on a social network.
I have spent the past year building bridges between AI developers and the blockchain world, arguing that zero-knowledge proofs could let models train on sensitive data without exposing it. I still believe that. But Ukraine is a reminder that privacy at the software layer means nothing if the pipe itself can be closed. The most sophisticated cryptographic guarantee is downstream of a satellite that someone can choose not to point at you. Faith in the protocol is not faith in the people who run it, and it is certainly not faith in the people who own the ground stations.
There is a symmetry worth naming. In one arena, writing code is treated as a crime — the Tornado Cash sanctions set the precedent that a developer can be prosecuted for what a protocol does. In another, switching code off is treated as a routine business decision. Both reveal the same uncomfortable truth: code was never outside politics. The question is only who holds the pen. When the pen belongs to a private operator, and the paper belongs to a nation at war, the nation does not get to write.
The contrarian move here is not to celebrate decentralization. It is to admit how much of our own industry has quietly rebuilt the same kill switch, only with better branding.
Bitcoin was designed to be peer-to-peer electronic cash. Post-ETF, it is largely a Wall Street instrument whose price is set by custodians and flows that answer to the same institutions capable of freezing a bank account. The chain resists censorship; the on-ramps do not. Most "decentralized" protocols run on foundation-controlled treasuries, upgradeable contracts, and governance that a handful of whales can swing. The grant committees that fund public goods, in my experience, tend to track relationships more than merit — with rare and instructive exceptions, like Optimism's RetroPGF, which funds outcomes rather than applications and remains the closest thing this industry has to a mechanism that genuinely resists capture. The exceptions prove how much the norm fails.
Much of what passes for decentralization is theater — a token, a Discord, a multisig dressed up as a DAO. The theater works until it doesn't, and it fails precisely when it matters, under stress, when the people holding the switches decide that their interests diverge from the protocol's stated values. Starlink did not betray anyone. It behaved exactly as a centralized system is expected to behave. The betrayal was in the belief that it wouldn't.
So before we lecture SpaceX about centralization, we should audit ourselves. We traded the temple for a ticker. We traded soul for speed, and called it progress — and now we are surprised when the speed belongs to someone else.
The real lesson of the Starlink restriction is not that centralization is bad. Everyone in this industry already knows that slogan; it is the cheapest sentence in crypto. The lesson is that accountability and decentralization are not the same thing, and we have spent a decade confusing them. A system can be fully decentralized and still leave its users with no recourse. A system can be centralized and still be accountable — if the parties have contracts, courts, and alternatives. Ukraine had none of the three. That is the failure. Not the architecture, but the absence of any lever the dependent party could pull.
Accountability, after all, is a legal technology, not a cryptographic one. It requires a counterparty you can sue, a jurisdiction that will hear you, and a substitute you can switch to. Cryptography gives you none of those. It gives you verifiability, which is powerful but partial. You can prove that a message was sent. You cannot compel anyone to receive it.
The market has begun to price this. Europe's IRIS² constellation, national military LEO programs, and the scramble for sovereign launch capacity are no longer science-fair projects; they are insurance policies. When I look at where capital is quietly repositioning during this sideways stretch, the signal is not in tokens — it is in the boring layers: satellite manufacturing, ground stations, spectrum licenses, launch cadence. Those are the assets that answer the question Ukraine just asked. The chain will record the transaction. Someone else will own the sky.
So here is the question I cannot shake. If a single company can decide how much of a war a country is allowed to fight, then decentralization is not a technical preference. It is the difference between being a customer and being a citizen. We have built remarkable machinery for recording truth. We have not yet built the thing that matters more: infrastructure no one can quietly switch off. Until we do, every ledger we ship is a temple with a landlord — and the rent comes due at the worst possible moment, usually without notice.
