A satellite that will never fly has a name, an orbital position, frequency bands, and a coverage map. There's no metal in it. Lawyers call such objects paper satellites — and they live in the registry longer than some real ones.
A digest on reusable rockets had a phrase that engineers rarely pass by: low Earth orbit is a finite resource, "first come, first served." The framing was about orbits, but something else clicked for me: in space you can "claim" without a rocket. You just file an application with the International Telecommunication Union — and the filing date secures your priority for frequency bands and orbital positions long before the first piece of hardware exists.
I checked the publication archive: no topic on paper filings, ITU only appeared as a regulator in pieces about laser communications and O3b. I dug in — and it turned out this isn't a quirk from a space law textbook. It's a functioning economy where the most active participant was a kingdom with a population just over a hundred thousand, and in 2023 SpaceX unexpectedly showed up in the credits of this story.
The mechanics are embarrassingly simple. A national administration submits advance publication information to the ITU — a description of a satellite it intends to launch. Then three stages: publication, coordination with those who might be interfered with by the new system, and notification — entry into the Master International Frequency Register, the main ledger of frequency assignments. Priority works on "first-come, first-served": whoever has the earlier receipt date gets to demand that late neighbors accommodate them.
Geostationary orbit is a ring at 35,786 km altitude, circumference about 265,000 km, and that looks like a lot until you do the math: satellites in the same frequency band need to be separated by roughly two degrees. That's about 180 positions for the entire planet. Today there are around 550 active spacecraft sharing these positions through frequency planning. In 1966 all eleven geostationary satellites were American; in 1982, out of 74 spacecraft on the ring, four belonged to developing countries.
Hence the politics. In 1976 eight equatorial states signed the Bogotá Declaration, claiming sovereignty over the segments of geostationary orbit above their territory — the idea was rejected almost unanimously. But by the 1980s developing countries achieved something else at the ITU: geostationary orbit was recognized as a "limited natural resource" and slots were reserved for those who were late to the distribution.
Reservation was the door everyone walked through later.
In 1987 American Matt Neilson came to Tonga to retire and instead proposed to King Tāufaʻāhau Tupou IV a scheme: the kingdom files applications for orbital positions over the Pacific Ocean and leases them to those who have rockets. The state spends not a dollar, the company — Friendly Islands Satellite Communications, aka Tongasat — gets half the profit. Registration took place February 13, 1989, the startup million came personally from Neilson.
March 23, 1990, Tongasat requested sixteen positions out of 180 — those very last useful ones that nobody had claimed yet. Planned lease price — $2 million per position per year, which would give the kingdom a budget boost of about 20%. Actually it came to around $700 thousand per transponder, and by September 1994 the company's capitalization was estimated at $45 million.
The reaction was harsh. Intelsat, which by then held nearly a monopoly on commercial satellite communications, declared that Tonga was turning orbit into private property and speculating on a resource that had been given limited status. The State Department got involved. Intelsat's grievance sounded almost like a legal joke: formally Tongasat violated not a single point of the Radio Regulations. It violated the "spirit" — what lawyers call the foundational ideology of the regime: space is the province of all mankind, not a place where a small state runs a financial operation.
The compromise result: in March 1991 the ITU agreed to six positions instead of sixteen, Tonga got a seventh soon after, two more later. The working positions were 14.0°E, 70.0°E, 83.3°E, 130.0°E, 134.0°E, 138.0°E, 142.5°E, 170.75°E and 257.0°E.
Then the real story began: Unicom licensed two positions and couldn't find money, Rimsat launched three satellites — in August 1993 a Russian Gorizont under the name Tonga Star One at 134.0°E, then Rimsat One and Rimsat Two. In 1995 Rimsat went bankrupt, Russia confiscated the two remaining spacecraft and ignored American court decisions. By 1997 Tonga had five satellites on positions, soon only two remained. Neilson was fired in February 1994 — an audit showed he owned a stake in Rimsat, meaning he licensed positions to a company that was his own.
Tonga is still remembered as "the mouse that roared": a small country stepped on a superpower's paw without formally violating anything. And infected others by example — Gibraltar, Papua New Guinea, Bermuda later followed its scheme.
The rules require: an assignment must be brought into use within seven years — but not necessarily by the declared satellite. It's enough for a spacecraft to operate on the frequency from the position for ninety consecutive days.
Hence a craft with its own name — satellite parking. A satellite living out its years is moved to the needed spot to check in with the registry. In November 2024 Thai Thaicom saved position 50.5° east longitude this way: the license expired November 27, the new spacecraft was only on order, and they drove old Eutelsat W7, launched back in 2009, to the position.
Tonga had its own champion for these purposes. In April 2002 Tongasat bought satellite Comstar 1D, launched by NASA February 21, 1981 — twenty-one years in space — renamed it Esiafi 1 and placed it at 70° east longitude. The author of an analysis in The Space Review wrote in plain text: an outside observer would conclude that a spacecraft with minimal residual capabilities is being used as a squatter on expensive real estate until someone willing to pay for the position shows up. The position was held by hardware manufactured in the year when most of Tongasat's current employees didn't yet exist.
There's also a more elegant way — not launching anything at all. In 2011 the ITU Radio Regulations Board deleted Iranian position 34° east from the registry. Zohra-1 did not exist. With Zohra-2 it was more interesting: the administration notified the union of bringing into use, but there was no proof of the spacecraft's existence, yet there was reason to believe that different foreign satellites took turns operating on the frequency from 26° east, creating the appearance of occupancy. Second in line for this position was France — and it insisted Iran had lost its place. The proceedings have dragged on for years precisely because the ITU doesn't launch satellites, doesn't inspect factories, and can't order a country to turn off a transmitter. It has no police.
To prevent the flow of applications from turning into an eternal warehouse, the ITU invented two instruments.
The first is money. Cost recovery under Council Decision 482: filing API for non-coordinated NGSO costs a fixed 570 CHF; coordination request — from 5,560 CHF plus 150 CHF per unit of complexity (ceiling 20,560 / 24,620 / 33,467 CHF depending on the number of coordination forms); notification — from 15,910 CHF and up. Each member state has the right to one free publication per year, and you can only withdraw an application without payment in the first fifteen days. Don't pay within six months — the publication gets canceled.
Scale for comparison. In September 2025 SpaceX paid about $17 billion for EchoStar's AWS-4 and H-block spectrum licenses — half cash, half stock, plus interest payments on EchoStar's debt. A filing in Geneva costs thousands of Swiss francs. The arithmetic where it's profitable to flood the union with paper arises on its own.
The second instrument is milestones. Under Resolution 35 (WRC-19, revised at WRC-23) after the seven-year bringing-into-use period, reporting begins in three steps: after two years — 10% of the declared constellation, after five — 50%, after seven — 100%. The penalty for failure is built into the text: if by the first milestone less than 10% is deployed, the final number of satellites in the filing gets cut to ten times the actually deployed. Meaning one launched spacecraft turns a filing for ten thousand into a filing for ten. After the second milestone the multiplier is two.
Elegant. But between filing and the first milestone nine years pass, and the first satellite can be launched in the seventh year — meaning spectrum can theoretically sit idle for up to fourteen years. Exactly what researchers pointed out back in 2023. And the weak link sits in reporting: all information about deployment is submitted by the administration itself, and the Iranian 2011 case control showed what this reporting is worth when parties are motivated to lie.
The scale changed by three orders of magnitude. In September 2021 the government of Rwanda submitted filings for two constellations, Cinnamon-217 and Cinnamon-937 — 337,320 satellites in 27 orbital shells at altitudes 550–640 km. Rwanda's connection to the story is thin: the country launched one spacecraft, RwaSat-1, in 2019, and the filing was linked to French company E-Space by Greg Wyler — the same person who previously built OneWeb together with the Rwandan Icyerekezo project. In June 2023 E-Space re-registered its holdings through French agency ANFR: Semaphore-C constellation, 116,640 satellites in 810 planes. The Rwandan filing for 300,000 spacecraft formally ceased to be primary, though the company stated it's not abandoning it.
The work by Falle, Wright, Boley, and Byers, published in Science in 2023, counted from the "as received" database: from 2017 to 2022 over 300 constellations were filed totaling more than a million satellites — 115 times more than were operating in orbit at the time. The authors explicitly name Tongasat as the historical precedent.
In the last week of December 2025 China submitted filings for approximately 203,000 satellites at once in fourteen constellations. The main part — 193,428 spacecraft — was proposed by the Radio Spectrum and Digital Technology Innovation Institute (RSDTII), aka 无线电创新院, registered in the new Xiong'an district December 30, 2025. The filings went out the day before the institute's own registration. The institute, as PolicyTracker noted, doesn't even have a website. The two largest constellations, CTC-1 and CTC-2 — 96,714 satellites in 3,660 planes each. Official comment: routine following of procedure, "a ticket to strategic reserve," and the 5G, 6G, and autonomous vehicle market will digest it all anyway.
Now look at the metal. According to Look Up Space's September barometer, there are 16,819 active satellites in orbit — 29% more than a year earlier. Of these Starlink is 11,041 spacecraft, almost two-thirds of the entire sky. China — 1,434 satellites, growth of 26% in a year. In the same year 1,412 objects were deorbited, including 914 satellites, of which 507 were Starlink. Meaning the paper-to-metal ratio is now about sixty to one, and that's with 2026 filings not yet in this statistic.
And here's the twist worth digging for. In October 2023 SpaceX registered with the ITU through Tonga the Esiafi II constellation: 29,988 satellites in 288 orbital planes at altitudes from 350 to 614 km, W-band range (75–110 GHz). The kingdom that in 1990 first figured out how to trade orbital positions now trades its flag for an operator with half the world's metal in orbit. The third party in this deal, by the way, is also familiar — the Esiafi name is inherited from that very 1981 satellite.
The asymmetry worked both ways. In February 2026 SpaceX filed with the FCC for a constellation of up to a million satellites for orbital data centers, in July — for 100,000 Gen3 spacecraft in W- and D-bands (shells 323–327.5 and 473–477.5 km). Amazon Leo missed its deadline: under license terms by July 30, 2026 it needed to have deployed half of 3,232 spacecraft, 396 were on hand. The FCC waived the deadline, but with a caveat: spectrum priority for satellites launched after the deadline is temporarily lowered and automatically returns by March 2028. Industry publication SatNews summarized the rule change without sentiment: launch manifests are becoming spectrum policy.
In parallel another fight is flaring up in the States — over the same limits, only from the other side. January 9, 2026, the FCC gave SpaceX temporary relief: Gen2 can operate in the US with power above EPFD limits, those very "volume limiters" from Article 22 that were written in the late 1990s so LEO wouldn't jam geostationary networks. Viasat called the approach "bad science," SES proposes periodic review of limits, and SpaceX in response writes that the current system "over-protects GSO operators," while Michael Calabrese of New America testified the opposite at Senate hearings in March 2026. The resolution formally lies with WRC-27 in Shanghai — agenda item 1.13 on improving coordination procedures, planning and due diligence for large constellations.
First. Public retellings of milestones contradict each other. SatNews writes: 10% in nine years, 50% in twelve, 100% in fourteen from filing. The official ITU release says: 10% two years after the end of the seven-year period, 50% after five, full deployment after seven. The text of Resolution 35 is unambiguous — the reference point is precisely the end of the bringing-into-use period, ceilings of "tenfold" and "twofold" are written in separate clauses — so SatNews arithmetic is correct. But in the Chinese explanation for television quoted by Beijing Scroll, a phrase slips through "10% within seven years," and that's a different reference point. A detail while discussing hardware; for lawyers counting milestones by calendar, a two-year difference is a matter of dispute.
Second, and this is more serious. The claim that Chinese filings "reserve interference budget" and force Western engineers to design hardware around "phantom noise" from paper satellites — I couldn't confirm with primary sources. It relies on the logic of Article 22: EPFD limits are calculated for the cumulative emissions of an NGSO constellation, and the ITU program calculates interference from declared parameters, not from actual spacecraft presence. The logic is plausible.
But in 2026 the specific EPFD conflict unfolded not between China and the West, but within the US: the FCC gave relief to SpaceX, and Viasat, SES and geostationary operators disputed it. I found no legal or regulatory case where an actual constellation was blocked by a paper one — neither recent nor historical. Tonga in 1990 blocked Intelsat diplomatically rather than procedurally. So "paper blockade" looks like a warning for now, not a documented fact.
Tonga's money also doesn't add up to one figure. RNZ in 2018 reported two payments from China totaling $49.9 million — compensation for Beijing putting its satellite on Tonga's 130° east position in 2006 without asking. Matangi Tonga wrote about "over $46 million," and AidData breaks the sum into tranches: $24.45 million as second tranche for the orbital position. The Supreme Court then ruled that Tongasat received this money illegally and had no parliamentary approval for it; the company pays the government 60% of gross revenue over $15 million per year, and in 2018 it was found guilty of transferring money from a Chinese enterprise to the princess instead of the state. The discrepancy in figures is simply explained — several tranches and several different rulings are involved — but I found no single verifiable sum.
Another fact I honestly don't know where to fit: in 2023 the same ITU database has a SpaceX filing for 29,988 satellites through Tonga, and simultaneously Tonga as a state since 2019 has been purchasing satellite communications from Kacific for its outer islands under a fifteen-year contract. The kingdom leases out its orbital position and buys satellite internet from others for itself — while having not a single satellite of its own. This is probably the best illustration of what the registry actually trades: not hardware, but queue position.