When US forces began striking Iranian-linked targets under Operation Epic Fury at the end of February 2026, a Hangzhou-based AI company called MizarVision was watching. The company published annotated satellite imagery tracking roughly 2,500 American military assets across the Middle East \u2014 carrier groups, missile defence batteries, B-2 bombers, individual aircraft on their ramps \u2014 updated in near real-time, posted to Weibo and X for anyone who wanted to look. The Defence Intelligence Agency assessed that Iran’s Revolutionary Guard was using MizarVision’s datasets to refine strike planning. The US State Department sanctioned the company on May 8, 2026 under its legal name, Meentropy Technology (Hangzhou) Co., Ltd.<\/p>\n\n\n\n
MizarVision operates no satellites. Not one.<\/p>\n\n\n\n
The company sources imagery from commercial providers \u2014 the exact constellation of suppliers remains disputed, with European operators including Airbus implicated by technical analysis, Chinese constellation Jilin-1 named by other analysts, and both Maxar and Planet Labs formally denying any sales to the firm \u2014 and processes it through AI pipelines that annotate military hardware with the same systems that annotate container ships and agricultural fields. Whatever the confirmed sourcing, the entire operation ran on purchased commercial data. What the State Department framed as an espionage problem was, mechanically, a market transaction. MizarVision bought a product. The product tracked American forces. The suppliers were commercial satellite operators.<\/p>\n\n\n\n
This isn’t a story about Chinese aggression or Iranian ingenuity. It’s a story about what happens when a price drops and the institutions that depended on that price being high don’t notice until the drop is irreversible.<\/p>\n\n\n\n
For most of the Cold War and into the first decade after it, satellite reconnaissance was among the most expensive capabilities a state could possess. A KH-11 Keyhole reconnaissance satellite \u2014 the kind the National Reconnaissance Office operates for tasked imaging of specific sites \u2014 costs several billion dollars per unit. Recent-generation Keyhole satellites have been publicly assessed at well above $4 billion each. A government that wanted to know what was building up on the other side of a border had two options: build this infrastructure, or negotiate access to a government that had.<\/p>\n\n\n\n
That world is gone.<\/p>\n\n\n\n
Planet Labs operates more than 200 Dove satellites in low Earth orbit, each small enough to fit in a carry-on bag. Together they image the entire landmass of the Earth every single day. Planet doesn’t publish per-unit hardware costs, but industry analysis places manufacturing cost in the range of several hundred thousand dollars to under a million per satellite. The entire commercial constellation operates at a fraction of the cost of one KH-11 unit, and it’s subsidised by revenue from customers monitoring crop yields and tracking container ships. Maxar’s WorldView constellation, which offers sub-meter resolution commercially, sells imagery to insurance actuaries, municipal planners, and journalists on assignment. The default state of the Earth’s surface is now continuous, publicly purchasable surveillance \u2014 operated by companies whose primary interest is commodity logistics.<\/p>\n\n\n\n
In October 2021, Planet Labs’ SkySat constellation began capturing images of a significant buildup near Yelnya, Russia \u2014 roughly 160 miles north of the Ukrainian border. The 144th Guards Motorized Rifle Division, part of the Western Military District, was accumulating armour and self-propelled artillery at a tempo analysts could track against their own historical baseline. By January 2022, Maxar imagery showed additional equipment concentrations at Klimovo, a military storage facility roughly eight miles from the Ukrainian border. These images were not classified intelligence product. They were commercial satellite data, purchased by journalists and think tanks, published by news organisations for any reader with an interest.<\/p>\n\n\n\n
The Kremlin’s invasion buildup was visible to the world for months before it crossed the border. Not through espionage. Through a subscription.<\/p>\n\n\n\n
The institutional response is still catching up. In July 2024, seventeen NATO member states signed the memorandum of understanding establishing the Alliance Persistent Surveillance from Space programme, committing the equivalent of more than $1 billion over five years to leverage commercial and national satellite assets. The APSS reached initial operational capability in December 2025. These aren’t bets on hypothetical commercial potential. They are institutional acknowledgments that the commercial surveillance market has already exceeded dedicated military capability at certain tasks.<\/p>\n\n\n\n
None of which means commercial access is uncontrolled. The US government has demonstrated precisely what leverage it retains over American providers. Starting March 9, 2026, Planet Labs imposed progressive access restrictions on Middle East imagery at Washington’s request: 96-hour delivery delays at first, escalating to 14-day holds, culminating in an indefinite withhold announced April 5 covering Iran, Iraq, Israel, Lebanon, and the Persian Gulf states. In early March 2025, the Trump administration suspended Ukrainian access to Maxar’s Global Enhanced GEOINT Delivery programme \u2014 effectively cutting Kyiv’s access to high-resolution satellite imagery of Russian positions. Access was restored on March 12, after US-Ukraine talks in Jeddah.<\/p>\n\n\n\n
The economics of watching<\/strong>\n\nA KH-11-class reconnaissance satellite costs $4 billion or more per unit, is irreplaceable if destroyed or degraded, images on-demand rather than continuously, and belongs to an inventory of roughly a dozen active national technical means. Planet Labs' Dove constellation consists of 200+ individually replaceable units, built at commercial scale, imaging the entire Earth every day rather than selected targets on request. The commercial model that makes this viable is counterintuitive: the core customers are farmers, insurers, and logistics operators. Military-relevant surveillance is a byproduct of markets that were never designed for intelligence. And because those markets are paying the overhead, adversaries accessing commercial imagery are, in a narrow economic sense, being subsidised by the same commodity markets that the target country relies on for economic productivity.<\/code><\/pre>\n\n\n\nThe leverage is real. The problem is that it’s retroactive and partial. European providers were unaffected by the Planet Labs restrictions. More to the point: the imagery already existed before any suppression request arrived \u2014 already captured, already sold, already distributed across providers and jurisdictions that no single government’s request can reach. No instruction can un-capture what a constellation has already imaged. Gating one American company for one region for some defined period, while the broader commercial imagery market and its existing archives remain intact, is not a solution to the scarcity problem. It’s a local patch on a structural condition.<\/p>\n\n\n\n
The network that couldn’t be cut<\/h3>\n\n\n\n
On February 26, 2022 \u2014 two days into Russia’s invasion of Ukraine \u2014 Ukraine’s Vice Prime Minister Mykhailo Fedorov tweeted directly at Elon Musk requesting Starlink service. Musk replied the same day. Service was activated within twelve hours. Physical terminals arrived within 48 hours. By mid-March, more than 5,000 terminals were operational. By May 2022, 150,000 Ukrainians were using Starlink daily.<\/p>\n\n\n\n
The timeline alone is something military procurement cannot explain and doctrine doesn’t account for. A dedicated US military satellite communications programme runs years from requirement to deployment. Commercial service ran days from a tweet.<\/p>\n\n\n\n
The military consequences extended well beyond basic connectivity. Ukrainian combat units used Starlink to relay drone surveillance video to artillery batteries in near real-time, compressing what planners call the sensor-to-shooter loop. Both surveillance and strike drone operations ran over Starlink connections. When Russian electronic warfare teams targeted terminals with jamming, SpaceX pushed software updates that restored connectivity within hours. Russia had spent decades developing EW capabilities specifically designed to sever adversary communications in the opening phase of a conflict. Against a commercial LEO constellation distributing connectivity through thousands of small satellites, those capabilities hit a network that could be patched faster than any jammer could adapt.<\/p>\n\n\n\n
The cost comparison is instructive. The US Advanced Extremely High Frequency satellite communications programme \u2014 six satellites providing nuclear-survivable encrypted communications to US and allied forces \u2014 cost the programme an estimated $850 million per satellite in manufacturing costs alone, before launch, with total programme costs running to approximately $15.5 billion against an original estimate of under $6 billion. Starlink’s per-unit manufacturing cost is a fraction of that, subsidised by consumer broadband subscriptions. Ukraine received communications resilience exceeding anything its defence budget could have procured, not because a government engineered it for war but because a commercial company built it for rural broadband customers who couldn’t get cable.<\/p>\n\n\n\n
That’s the mechanism. Here’s the fault line it opened.<\/p>\n\n\n\n
In September 2022, Musk refused Ukraine’s request to extend Starlink coverage to Crimea for a naval drone operation targeting Russian ships in Sevastopol Bay. The Ukrainian military was running drone submarines packed with explosives toward the Russian Black Sea Fleet. Musk declined to activate service in the relevant area, citing concerns about nuclear escalation. Whatever his reasoning, the decision was his \u2014 not Ukraine’s government’s, not any treaty body’s, not any institutional process Ukraine could invoke or appeal. SpaceX subsequently developed Starshield, a separate military-grade satellite service for Pentagon use, while maintaining commercial Starlink under different access terms.<\/p>\n\n\n\n
The old vulnerability \u2014 adversaries physically destroying communications infrastructure \u2014 is at least legible within existing doctrine. Armies know how to harden command nodes, build redundancy, route around damage. The new vulnerability doesn’t have a doctrinal name. Military communications infrastructure is now materially dependent on commercial decisions made by people who answer to shareholders and have their own views on acceptable risk. Military doctrine was built to address an adversary that might destroy your communications. It has no framework for a supplier who might restrict them.<\/p>\n\n\n\n
When cheap weapons hit like expensive ones<\/h3>\n\n\n\n
GPS is not a commercial product. Worth stating plainly, because this section’s argument depends on the distinction.<\/p>\n\n\n\n
GPS is a US military navigation system, built by the Department of Defence beginning in 1973, currently operated by the U.S. Space Force from Schriever Space Force Base. Ronald Reagan committed to civilian access after the 1983 Soviet shootdown of Korean Air Lines Flight 007 \u2014 a civilian aircraft that strayed into prohibited airspace due to navigational failure, killing all 269 on board \u2014 and the government eliminated Selective Availability, the deliberate accuracy degradation imposed on civilian GPS receivers, on May 1, 2000. The precision revolution that followed was enabled by government infrastructure. But it was commoditised through commercial manufacturing, and that commoditisation produced the same economic outcome as if the infrastructure had been commercial from the start.<\/p>\n\n\n\n
The Joint Direct Attack Munition demonstrates the scale of the shift. A JDAM guidance kit bolted onto an existing unguided iron bomb costs roughly $25,000 to $40,000 per unit. It converts a $5,000 dumb bomb into a GPS-guided weapon with a circular error probable of approximately five metres. Before GPS guidance, achieving comparable all-weather precision required weapons costing $500,000 to $700,000 each \u2014 the AGM-154 JSOW, the AGM-84 SLAM-ER. Laser-guided predecessors were cheaper but required clear weather and a human operator illuminating the target throughout the weapon’s entire flight. Boeing has manufactured over 550,000 JDAM kits. Precision went from a strategic asset, deliverable only by specific aircraft under specific conditions, to a logistics item ordered by the pallet. If you wanted to know what it meant that GPS went from military-exclusive to civilian infrastructure, the JDAM supply chain is the answer.<\/p>\n\n\n\n
Ukraine’s experience with the M982 Excalibur GPS-guided artillery shell then demonstrated both the precision revolution and its counter-economy, simultaneously.<\/p>\n\n\n\n
Excalibur costs between $68,000 and $112,000 per round and gives a standard 155mm howitzer near-surgical accuracy against a target the size of a car at ranges exceeding 40 kilometres. When Ukraine first received the shells in 2022, documented hit rates ran around 70 percent. By January 2023, as Russian GPS jamming systems \u2014 the R-330Zh Zhitel prominent among them \u2014 spread across the front, the documented hit rate had fallen to approximately 55 percent. By August 2023, it was approximately 6 percent. Ukraine stopped receiving Excalibur by mid-2024, not from supply shortage but because the cost-per-successful-strike had risen from roughly $300,000 to approximately $1.9 million as the hit rate collapsed. The weapon had lost its combat utility.<\/p>\n\n\n\n
The floor of accessible precision has still risen permanently \u2014 GPS access plus commercial manufacturing has made guided munitions available to any military that can buy the kits. But the ceiling is actively contested. Adversaries have demonstrated they can degrade the precision at a cost far lower than the cost of the precision itself. The arms race hasn’t ended; it’s migrated to a different layer of the stack. From expensive precision versus expensive targets, to cheap guidance versus cheap denial.<\/p>\n\n\n\n
And the commercial pressure is now working on the GPS infrastructure itself. In March 2026, Xona Space Systems closed a $170 million Series C funding round to build the Pulsar constellation \u2014 a commercial positioning network in low Earth orbit with signals roughly 100 times stronger than GPS, designed from the start for resistance to jamming and spoofing. Initial commercial service is targeted for 2027. The same economic logic that produced Planet Labs and Starlink is now applying to GPS: government infrastructure, once exclusive and monopolistic, is being commercially replicated, hardened, and improved by private capital that needs precise navigation for autonomous vehicles and precision agriculture far more than it cares about military doctrine.<\/p>\n\n\n\n
The doctrine gap<\/h3>\n\n\n\n
US Joint Publication 3-0, the Pentagon’s governing document for joint campaigns and operations, still lists surprise as a principle of war \u2014 achievable through operational security, deception, and information management. NATO Allied Joint Publication 3.3 still frames communications disruption as a first-phase offensive objective. FM 3-0, the Army’s capstone operations manual, still treats the ability to shape the information environment as something a military can reliably do.<\/p>\n\n\n\n
These documents describe a world the commercial satellite market has already dissolved.<\/p>\n\n\n\n
MizarVision is the sharpest available illustration. The US military runs OPSEC procedures designed to limit adversary knowledge of deployments, movements, and force compositions \u2014 procedures refined over decades, codified in regulation, practiced across every major exercise. While those procedures were running, a company in Hangzhou with no satellite assets and no intelligence apparatus was publishing the locations of carrier groups, missile defence batteries, and individual aircraft in annotated, publicly accessible updates during an active US combat operation. The DIA assessed that Iran was using the data for strike planning. The information asymmetry didn’t merely shrink. It inverted. The actor with the $850 billion defence budget \u2014 the FY2025 DoD request was $849.8 billion \u2014 had less information security than adversaries with a commercial satellite subscription and an off-the-shelf AI pipeline.<\/p>\n\n\n\n
The government’s response \u2014 sanctioning MizarVision and pressuring Planet Labs into a regional imagery blackout \u2014 addresses the incident without addressing the structure. Meentropy Technology (Hangzhou) Co., Ltd. is sanctioned. Dozens of comparable operators across other jurisdictions are not. European providers were unaffected by the Planet Labs restriction. Sanctioning one company when the underlying capability is a feature of a global commercial satellite market is equivalent to banning a single printing press after the technology has already spread to every continent. The response is accurate about the proximate instrument and wrong about what produced it.<\/p>\n\n\n\n
The lag here is structural, not a failure of intelligence. Military procurement cycles run a decade or more. Commercial satellite companies iterate in months. The gap between what commercial systems can do and what doctrine assumes is widening with each new launch cadence.<\/p>\n\n\n\n
The dangerous corollary: a military that plans for surprise and discovers at execution that its buildup has been visible for months doesn’t simply acknowledge the failure and adapt quietly. It compresses timelines dangerously, front-loads force to compensate, or escalates rather than accept the constraint. A military that plans for communications dominance and discovers it cannot sever an adversary’s commercial LEO constellation may decide to destroy the constellation \u2014 which means attacking private companies’ assets in orbit, treating commercial firms as de facto belligerents, and entering legal territory that no existing framework covers. The Outer Space Treaty was written in 1967 for government-operated space programmes. No NATO article specifies alliance obligations triggered by an attack on commercial space infrastructure providing dual-use services to a belligerent. The legal framework for commercial space as military infrastructure is, for practical purposes, absent.<\/p>\n\n\n\n
What the ledger says<\/h3>\n\n\n\n
Return to MizarVision \u2014 not the company’s specific conduct but what it represents as a market outcome.<\/p>\n\n\n\n
The institutions dealing with the fallout are responding to a series of security incidents. One company tracked American forces; it was sanctioned. The company that supplied the imagery was pressured to stop. The response framework is incident-based: identify the proximate cause, apply the available remedy, move on.<\/p>\n\n\n\n
But the proximate cause is an effect. The cause is that the commercial satellite market solved the cost problem for Earth observation \u2014 not for intelligence, but for agriculture and insurance and logistics \u2014 and military surveillance is a byproduct of those markets. Those markets are too large, too internationally distributed, and too commercially entrenched to be sanctioned into non-existence. No classification regime can restore scarcity to a commodity. No sanction reaches the archives of imagery already distributed across dozens of operators and jurisdictions.<\/p>\n\n\n\n
What looked like an intelligence problem when MizarVision’s annotated maps first circulated on Weibo was actually a market condition that had been building for years. Treating it as an intelligence problem means reaching for the wrong instruments \u2014 counterintelligence, sanctions, access restrictions \u2014 which address individual incidents but leave the structural condition entirely intact. The commercial satellite business will keep growing. The costs will keep falling. The imagery will keep accumulating. And the institutions built around the assumption that space-based visibility is scarce and controllable will keep discovering, one incident at a time, that those assumptions were wrong.<\/p>\n\n\n\n
MizarVision’s imagery came from commercial satellites operated by companies whose core business was agricultural productivity, insurance underwriting, and logistics monitoring. Infrastructure built, licensed, and commercially developed for economic purposes became the infrastructure used to track American forces in a combat operation. Exactly which companies supplied it is disputed, and will likely remain so. That the commercial satellite market supplied it is not.<\/p>\n\n\n\n
That’s not an espionage case. It’s not a Chinese competition story. It’s not even really a technology story. It’s a story about what happens when a price drops, and the institutions that were built when that price was high haven’t examined the assumption since.<\/p>\n\n\n\n