The Network Overhead

On 26 February 2022, as Russian forces advanced into Ukraine, the country’s digital minister, Mykhailo Fedorov, wrote to Elon Musk on Twitter. While Musk tried to colonise Mars, he said, Russia was trying to occupy Ukraine, and he asked for Starlink stations. About ten hours later Musk replied: “Starlink service is now active in Ukraine. More terminals en route.” On 28 February Fedorov posted a photograph of the first terminals. He thanked Ukraine’s ambassador in Washington for help with licensing and certification, which suggests that most of the delay had been paperwork rather than engineering.¹

A network had been switched on over a country at war by a private company in response to a public message. No treaty, contract or licence had been negotiated in advance. Over the next four years the same network became the backbone of Ukraine’s military communications, a subject of dispute between governments, and a tool that could be geofenced, rationed and, in 2026, closed to the other side.

By the autumn of 2026 that network, Starlink, had 13.6 million subscribers in 174 markets and earned SpaceX $5.25 billion in a single quarter.² Trackers counted about 11,200 Starlink satellites in orbit, roughly two-thirds of all the working satellites around the Earth.³ Its owner had sold shares to the public in June in the largest initial offering on record.⁴ Meanwhile the International Telecommunication Union estimated that 2.2 billion people were still offline, 96 percent of them in low- and middle-income countries.⁵

This essay follows how internet access came to be delivered from the sky, and why it took eighty years from the first proposal to a mass market. It follows the decisions that gave satellite communications their shape: the choice of the high geostationary orbit in 1965, the failed low-orbit constellations of the 1990s, the high-capacity satellites with data caps, the reusable rocket, the dish sold below cost, a government’s purchase of a bankrupt rival, a rival’s choice of launch suppliers and a company’s choices about a war. Its deeper subject is what happens when the infrastructure that reaches the most remote house is a private network overhead: who owns the orbits it occupies, who decides where it works, and what it costs people who never signed up for it.

Relays in the sky

Long-distance radio in the first half of the twentieth century depended on the ionosphere, a layer of the upper atmosphere that reflects some radio waves back to the ground. Its behaviour changed with the time of day, the season and the sun, and television signals passed straight through it. In October 1945 the magazine Wireless World published an article by Arthur C. Clarke. Clarke had called it “The Future of World Communications”; the editor renamed it “Extra-Terrestrial Relays”. It argued that a station at a distance of 42,000 kilometres from the centre of the Earth would circle once every 24 hours. Placed above the equator, Clarke wrote, it “would revolve with the earth and would thus be stationary above the same spot on the planet.” Three such stations, over Africa, the Pacific and the Americas, could cover the whole globe.⁶

Clarke imagined crewed stations supplied by rocket and powered by the sun. He expected them to make long chains of radio relays on the ground obsolete within a few decades. He did not patent the idea; secondary accounts say a lawyer talked him out of it.⁶ The orbit he described, about 35,786 kilometres above the equator, is now called geostationary.

The first round: a high orbit wins

The Soviet Union launched Sputnik on 4 October 1957. Within five years there were two competing ideas of how to relay signals through space. NASA’s Echo 1, launched in August 1960, was a reflecting balloon 30 metres across. AT&T’s Telstar 1, built by Bell Labs and launched by NASA on 10 July 1962, was an active repeater in a medium orbit, and on 23 July it carried the first public live television across the Atlantic. Because Telstar moved across the sky, a ground station could use it only for minutes at a time, and continuous service would need many satellites and moving antennas. Hughes Aircraft took Clarke’s route. Its Syncom 2, launched in July 1963, was the first communications satellite in a 24-hour orbit, and Syncom 3, in August 1964, sat over one point of the equator and relayed the Tokyo Olympics to the United States.⁷

Congress decided in 1962 that satellite communications would be run by a private, shareholder-owned corporation created by statute, COMSAT, with initial capital of $200 million, acting for the United States in an international consortium, Intelsat, created by agreements signed on 20 August 1964. COMSAT’s first satellite was a Hughes design. Early Bird launched on 6 April 1965 and began commercial service over the Atlantic on 28 June, with 240 telephone circuits or a single television channel. NASA’s history says it offered almost ten times the capacity of the submarine telephone cables for almost a tenth of the price.⁸

The choice of one high satellite over a medium-orbit constellation was sensible: it stayed in view of fixed dishes on both sides of the ocean, and nothing on the ground had to track it. For half a century satellite communications meant geostationary satellites selling capacity wholesale to telephone companies and broadcasters. The cost of that success came later. A satellite 35,786 kilometres up is far enough away that light takes about a quarter of a second to go up and down again. A request and its reply must make that journey twice, so no geostationary connection can respond in less than about half a second, and real consumer services measured 600 to 800 milliseconds.⁹ For a telephone call the delay was an annoyance. For the interactive internet, with its many small exchanges between computer and server, it would be a permanent handicap.

A finite ring

The geostationary orbit is a single ring above the equator, and satellites in it must be spaced so that their signals do not interfere. Both the positions and the radio frequencies are scarce. The International Telecommunication Union, a United Nations agency, coordinates them. For most satellite services its procedure has been first come, first served: a country files for a position and frequencies, and later filers must coordinate around it.¹⁰

That rule favoured the countries with satellites to launch. On 3 December 1976 a group of equatorial states, among them Colombia, Ecuador, Indonesia and Kenya, signed the Bogotá Declaration, which argued that the geostationary orbit exists because of the Earth’s gravity, so it is a natural resource rather than part of outer space, and claimed sovereignty over the segments above their territory. No other country recognised the claim, and it was largely abandoned.¹¹ The ITU went a different way. In 1977 a world radio conference drew up an advance plan for broadcasting satellites in two of its regions, reserving positions, usually six degrees apart, for countries that had not yet launched anything. The fixed satellite service kept first come, first served.¹⁰ The ITU’s constitution calls orbits and frequencies “limited natural resources” to which countries should have “equitable access”, taking into account the special needs of developing countries.¹² The tension between that principle and the practice of filing first has never been resolved.

Dishes on shops, then on homes

Satellites reached businesses before households. In the mid-1980s Hughes sold its first network of small dishes linking many sites to a hub to Walmart, which used it to track inventory and sales across its stores.¹³ In 1996 it launched DirecPC for homes. Downloads came by satellite at up to about 400 kilobits per second, and requests went back over a telephone modem. Ping times of 400 to 600 milliseconds were built in.¹⁴ DirecPC became DirecWay and then HughesNet. A rival, WildBlue, installed its first home customers in June 2005 using capacity on a Canadian satellite.

Internet in the sky

In the 1990s the internet’s growth persuaded several groups that satellites in low orbit, a few hundred to about 1,400 kilometres up, could offer broadband with little delay. The physics favoured them: a satellite 700 kilometres away is fifty times closer than a geostationary one. The economics did not. A satellite that low crosses the sky in minutes, so continuous coverage needs hundreds or thousands of them, and each user needs an antenna that can follow one satellite and hand over to the next.

Teledesic was the most ambitious. Work began in 1990, and in 1994 Craig McCaw and Bill Gates each put in $5 million. The first design called for 840 working satellites at about 700 kilometres, at an estimated cost of more than $9 billion. Contemporary coverage was sceptical. The FCC authorised the system in March 1997. That year Boeing became prime contractor and the design shrank to 288 satellites. In May 1998 Motorola abandoned its own planned constellation, Celestri, a 63-satellite system estimated at $12.9 billion, invested $750 million in Teledesic and replaced Boeing as prime contractor. A rival from Alcatel and Loral, SkyBridge, grew from 64 to 80 satellites while its estimated cost rose to about $6 billion.¹⁵¹⁶

The companies that had already launched did badly. Iridium, Motorola’s $5 billion satellite phone system, filed for bankruptcy on 13 August 1999, nine months after it began service, and was sold to private investors who relaunched it in 2001, a story this site tells in A Phone for Anywhere on Earth. ORBCOMM filed in September 2000 and Globalstar in February 2002.¹⁷ Each network survived under new owners who had not paid to build it.

Teledesic ended Motorola’s role in October 2000, hired Alenia Spazio to build a system of about 30 satellites for about $1 billion, and suspended satellite work on 1 October 2002, having raised about $900 million. The company’s explanation, reported by the Seattle Times, named the collapse of the telecom industry, capital markets closed to speculative satellite ventures, the bankruptcies of Iridium and ICO, which had found little demand beyond fishing boats and oil rigs, and land-based wireless networks that had advanced faster than expected and could be built city by city.¹⁵ The analyst Tim Farrar later argued that the low-orbit projects were a sideshow: the real story of satellite communications in those years was the steady improvement of geostationary satellites, and the slump of 2001 to 2004 came from too much capacity in that orbit.¹⁸

Judged by what was knowable in the 1990s, the plan was early rather than foolish. Launch on the Space Shuttle cost about $54,500 per kilogram to low orbit, by one NASA comparison, roughly twenty times what a Falcon 9 would later charge.¹⁹ The design changed with each new partner, from 840 satellites to 288 to 30, which suggests it was set by the money available rather than by evidence of demand. Nearly everything Teledesic proposed was built by someone else two decades later.

Capacity instead of speed

Satellite broadband at home survived in the high orbit, by making each satellite carry far more. Viasat bought WildBlue in December 2009 for $568 million, taking on about 400,000 subscribers. On 19 October 2011 it launched ViaSat-1, with about 140 gigabits per second of capacity, more than all the other internet satellites over North America combined. Its Exede service began in January 2012 at up to 12 megabits down and 3 up, from $50 a month; it signed more than 200,000 subscribers by early 2013 and did well in the FCC’s tests of whether providers delivered advertised speeds. Hughes answered with EchoStar XVII, launched in July 2012, and in March 2017 with HughesNet Gen5, the first American satellite service to meet the FCC’s broadband definition of 25 megabits down and 3 up, from $49.99 a month with household data plans of up to 50 gigabytes a month.²⁰

These satellites did what engineering could do within the orbit. Capacity rose with each generation; ViaSat-2, launched in 2017, carried 260 gigabits per second. The delay stayed the same, because it was set by distance. Capacity was shared among everyone under a beam, so companies rationed it with data caps and slowed heavy users. HughesNet had about 1.56 million subscribers at the end of 2020, about 1.19 million of them in the United States; Viasat reported 603,000 American subscribers in September 2020.²⁰ Against an FCC estimate in 2015 that 55 million Americans, including 53 percent of rural Americans, lacked a 25-megabit connection, those were modest numbers.²¹ Satellite was the service of last resort. In April 2023 Viasat’s largest satellite, ViaSat-3, reached orbit, and its main antenna failed to deploy properly; the company expected to get less than a tenth of the planned terabit of capacity.²⁰

The other three billion

One company found a middle path. Greg Wyler founded O3b Networks in 2007, named for the “other 3 billion” without internet access. It put 20 satellites in medium orbit, about 8,062 kilometres up, and sold capacity to island telephone companies, cruise ships and governments rather than to households. Its first satellites launched in June 2013, three years late, and SES bought it in 2016.²²

Two plans in one week

In June 2014 Wyler’s new company, WorldVu, acquired spectrum rights once held by SkyBridge, and in September he and colleagues left Google, which had invested in O3b, to pursue it. In January 2015 WorldVu, renamed OneWeb, announced backing from Virgin and Qualcomm for about 650 satellites at 1,200 kilometres.²³

The same week, on 16 January 2015, Musk presented a plan to an invited audience in Seattle: about 4,000 satellites, a cost of perhaps $10 billion to $15 billion, latency of 20 to 30 milliseconds, a new Seattle office to build them, and revenue to help pay for a city on Mars. Days later SpaceX confirmed that Google and Fidelity had invested about $1 billion. Reports said Musk and Wyler had discussed working together before going their separate ways.²⁴

Rockets that come back

What separated SpaceX from Teledesic was a rocket. A Falcon 9 booster first landed on 21 December 2015 and first flew again on 30 March 2017, carrying a satellite for SES.²⁵ A company that recovers and reuses the most expensive part of its rocket pays far less per launch than one that throws it away, and SpaceX was both the launcher and the customer. SpaceX has never published its internal launch costs, and outside estimates vary widely, but the advantage over rivals buying launches at market prices was plain.²⁵ It also built the satellites, designed the user terminal and sold the service, so no supplier’s margin stood between the cost of a launch and the price of a subscription.

The FCC approved 4,425 Starlink satellites in March 2018, at about 1,110 to 1,325 kilometres, with half to be launched by March 2024. In November it approved 7,518 more at about 340 kilometres. Then SpaceX asked to move 1,584 satellites down to 550 kilometres, and the FCC agreed in April 2019, noting that at that height a failed satellite would fall out of orbit within about five years, where debris at the higher altitude could last a thousand years or more, by one expert’s estimate.²⁶ The lower orbit also reduced latency and, because each satellite saw a smaller area, required more of them. Two test satellites flew in February 2018. On 23 May 2019 a Falcon 9, on a booster flying for the third time, carried the first 60 production satellites.²⁷

Better than nothing

In October 2020 SpaceX invited users into a public beta. It called it the Better Than Nothing Beta and warned of brief outages. Service cost $99 a month plus $499 for a kit, and promised 50 to 150 megabits per second with latency of 20 to 40 milliseconds.²⁸ The price was high by the standards of city broadband, and the name set expectations for people with few alternatives. Rural users were the natural first customers, because a satellite beam is shared by everyone beneath it and is less crowded where people are few.

The dish was the hardest part, because it has to follow satellites that cross the sky in minutes. Early on each one cost SpaceX about $3,000. In April 2021 its president, Gwynne Shotwell, said the cost had fallen to about $1,500 (other reports gave $1,300) and that the company was not charging customers what the terminals cost to make. In September 2023 SpaceX said it no longer lost money on them.²⁹ From September 2021 new satellites carried lasers to pass traffic between them, which let the network serve high latitudes and the open ocean, far from any ground station.³⁰

The subsidy bought scale. By SpaceX’s own announcements, relayed by the press, Starlink passed one million subscribers in December 2022, two million in September 2023, four million a year later, nine million in December 2025 and twelve million in June 2026.³¹ United announced in September 2024 that it would offer free Starlink Wi-Fi on all its aircraft,³² and by the end of September 2026 Starlink was on more than 3,800 aircraft.²

A bankruptcy and a ministerial direction

OneWeb, meanwhile, solved the problem Teledesic never reached. A joint venture with Airbus was building two satellites a day by January 2020, and the company had launched 74. But it needed more money, SoftBank, its largest investor, reportedly declined to put in more after about $2 billion, and on 27 March 2020, blaming the turmoil of the pandemic, OneWeb filed for bankruptcy and laid off most of its staff.²³

The British government and the Indian company Bharti Global bought it, for $500 million each, in July 2020. The decision was unusual enough to leave a record. On 26 June the acting permanent secretary of the business department, Sam Beckett, wrote to his minister that there had been no time for a full business case, that the UK Space Agency’s technical assessment had found “substantial technical and operational hurdles” and a high likelihood that more money would be needed, and that although one scenario offered a good return, “the central case is marginal”. He could not satisfy himself that the purchase was value for money and asked for a formal direction. The business secretary, Alok Sharma, gave it the same day.³³ The government’s stated aims included a sovereign satellite capability and rural broadband. Commentators and a parliamentary committee questioned whether a broadband constellation served them.²³

The rescue worked as engineering. When Russia invaded Ukraine in 2022 it cancelled a launch and kept 36 OneWeb satellites at Baikonur; OneWeb switched to SpaceX and Indian rockets and completed its first generation, 618 satellites, in March 2023. It merged with the French operator Eutelsat that September. In 2025 Eutelsat raised about €1.5 billion, led by the French state, which ended with about 29.65 percent. The British stake fell to about 10.89 percent.³⁴ Eutelsat ordered 440 replacement satellites from Airbus, and in the year to June 2026 its low-orbit revenue rose 69.5 percent to €297 million.³⁴ Europe has a constellation of its own. The question the 2020 letter raised, what Britain was buying and how anyone would know whether it had succeeded, was never answered in public.

The rival that bought rides

Amazon entered later and with more money. The FCC approved its Project Kuiper constellation of 3,236 satellites in July 2020, by five votes to none, on condition that half be in orbit by 30 July 2026. Amazon pledged more than $10 billion. In April 2022 it announced the largest commercial launch purchase in history, up to 83 launches from United Launch Alliance, Arianespace and Blue Origin. None was from SpaceX.³⁵

The choice was defensible. Buying launches from your main competitor gives it information, revenue and control over your schedule. But the rockets Amazon chose were new, and they were late. Its first two prototypes flew in October 2023, its first production launch in April 2025, and by the time it renamed the service Amazon Leo in November 2025 it had about 150 satellites in orbit and had bought Falcon 9 launches as well. In January 2026 it asked the FCC to extend the July deadline by two years. On 5 June 2026 the FCC instead waived it with conditions: satellites launched after the milestone lose priority for their spectrum until March 2028, and the whole system is still due by July 2029. By October 2026 Amazon had nearly 400 satellites in orbit and had not begun commercial service, though its chief executive said it would start that year.³⁵³⁶ In April 2026 Amazon agreed to buy Globalstar, Apple’s satellite partner, for about $11.6 billion.³⁷

Others are further behind. Canada’s Telesat chose MDA to build 198 Lightspeed satellites in 2023, financed by C$2.54 billion in government loans, and now expects service in 2028.³⁸ China’s two state-backed constellations, Qianfan and Guowang, began launching in 2024 and had a few hundred satellites each by the autumn of 2026, with no confirmed broadband service.³⁹

A network in a war

After the first terminals reached Ukraine in February 2022, governments and SpaceX disagreed about who was paying. The US aid agency USAID announced in April that SpaceX had donated most of 5,000 terminals; reporting by the Washington Post and a later inspection by USAID’s inspector general found that the agency had bought about 1,500 of them. In September 2022 SpaceX wrote to the Pentagon that it could not keep donating terminals or funding service indefinitely, and by its own account about 85 percent of some 20,000 terminals had been paid for wholly or partly by the United States, Britain, Poland or others. Poland said it had supplied more than half of the terminals Ukraine received and paid for their subscriptions.⁴⁰

The episode most often told about Crimea needs care. In September 2023 a CNN excerpt from Walter Isaacson’s biography Elon Musk said Musk had turned off Starlink near Crimea in September 2022 to stop a Ukrainian drone attack on the Russian fleet at Sevastopol. Isaacson corrected this within days: Ukrainian forces had believed coverage extended to Crimea, but it did not, and Musk was following a policy already in place. Musk said that SpaceX “did not deactivate anything” and that it had refused an emergency request to activate coverage there because doing so would have made SpaceX complicit in an act of war.⁴¹ The documented decision is a refusal to extend coverage, made by a company, under its own geofence.

In February 2023 Gwynne Shotwell said Starlink was never meant to be weaponised, that its use on drones went beyond the agreement, and that SpaceX could and did limit it.⁴² In June 2023 the Pentagon confirmed it was buying Starlink services for Ukraine. Reuters later reported that SpaceX’s government arm, Starshield, was building spy satellites under a $1.8 billion contract signed in 2021 with the National Reconnaissance Office.⁴³ In February 2025 Reuters reported, citing three sources, that American negotiators had raised Ukraine’s access to Starlink during talks over minerals; Musk called the report a lie. In March, after Musk posted that Ukraine’s front line would collapse if he turned Starlink off, Poland’s foreign minister, Radosław Sikorski, said Poland paid about $50 million a year and would look for other suppliers if SpaceX proved unreliable. Musk told him to “be quiet, small man”. Secretary of State Marco Rubio said no one had threatened to cut Ukraine off, and Musk later said Starlink would never turn off its terminals.⁴⁴

Russian forces had meanwhile acquired terminals through third countries. SpaceX said it did no business with Russia. In early 2026, after attacks by Russian drones attributed to Starlink connections, SpaceX first limited speeds and then admitted only terminals on a list of registered users, which Ukraine updates daily. Ukraine’s defence ministry said on 5 February that Russian terminals had been cut off; some Ukrainian units were disrupted too.⁴⁵ In October 2026 the Wall Street Journal reported that a speed cap had been imposed on military use in Ukraine; who ordered it and why is disputed.²⁴⁶ The pattern across four years is consistent. A company decided where its network worked, for whom and at what speed, and governments that depended on it found out what those decisions were after they were made.

Switched on, switched off

Elsewhere the same power ran in other directions. In September 2022, during the protests in Iran, the US Treasury widened the licence for communications exports, and Musk said he was activating Starlink; it has never been licensed there, and terminals were smuggled in. Iran criminalised its use in 2025, shut down its internet for three weeks during the protests of January 2026 and for 88 days after the war that began on 28 February 2026, jammed Starlink signals, and called the service “a tool of war”. The Wall Street Journal reported that the United States had sent about 6,000 terminals into Iran after the January crackdown.⁴⁷ In October 2025 SpaceX said it had disabled more than 2,500 terminals near suspected scam compounds in Myanmar, where Starlink is not licensed either, which showed it could locate and disable terminals in a country where it had no permission to operate.⁴⁸

National governments pushed back through licences. In Brazil in 2024 a Supreme Court justice froze Starlink’s accounts to collect fines owed by X, another company Musk controls, and Starlink complied with an order to block X after first refusing; the accounts were released after about $3.3 million was transferred.⁴⁹ India licensed Starlink in 2025, but by mid-2026 it had still not launched there, pending spectrum, pricing and security clearance. South Africa’s rule requiring 30 percent ownership by historically disadvantaged groups was still unresolved in late 2026.⁵⁰

A crowded shell

In 1978 two NASA scientists, Donald Kessler and Burton Cour-Palais, showed that collisions between objects in orbit could produce fragments that cause more collisions, until a belt of debris sustained itself.⁵¹ On 10 February 2009 the working Iridium 33 and the defunct Russian Kosmos 2251 collided at about 790 kilometres over Siberia, producing well over a thousand trackable fragments.⁵² In December 2021 China complained to the UN that its space station had manoeuvred twice that year to avoid Starlink satellites; the United States replied that the approaches had not met the threshold for warnings and that China had not contacted it or SpaceX.⁵³

On 29 September 2022 the FCC voted 4–0 to require satellites in low orbit to be removed no later than five years after their missions end, replacing a 25-year guideline.⁵⁴ Starlink’s low shells meet the rule almost automatically. The risk changed shape rather than disappearing. By SpaceX’s reports, its satellites made 148,696 avoidance manoeuvres between June and November 2025 and 207,152 in the following six months, and they manoeuvre at a far lower threshold of collision risk than the industry norm.⁵⁵ In December 2025 a Starlink satellite suffered a propulsion failure and released debris, and another broke up at about 560 kilometres on 29 March 2026. SpaceX announced that it would lower satellites from about 550 to 480 kilometres during 2026.⁵⁶ About 16,900 satellites were active in September 2026, by one count.³

The sky astronomers lost

Astronomers’ concern was practical. Wide-field survey telescopes take long exposures of large patches of sky, and satellites reflecting sunlight leave streaks across them. The Vera C. Rubin Observatory estimated that about a tenth of its images would contain at least one trail if 40,000 satellites were in low orbit; the International Astronomical Union’s centre for protecting the dark and quiet sky, founded in 2022, put the share at 30 to 40 percent.⁵⁷ SpaceX tried a dark coating, then sunshades, then a mirror film on its second-generation satellites that it says cuts brightness about tenfold, which it offers to other operators at cost. Radio astronomers using the LOFAR telescope found that most Starlink satellites they observed leaked radio emission in protected bands, and that the second generation leaked more. No regulation covers such unintended emissions.⁵⁷ The IAU recommends that satellites be no brighter than magnitude 7; a 2026 preprint found that almost all satellites in the large constellations exceed it.⁵⁸

The atmosphere is a newer worry. In 2023 NOAA researchers flying through the stratosphere found metals from re-entering spacecraft in about a tenth of sulphuric acid particles. A 2024 model estimated that re-entering satellites had already raised atmospheric aluminium above natural levels and that alumina from megaconstellations could deplete ozone. The effects have not been measured.⁵⁹

Who pays for the last house

In December 2020 Starlink won $885.5 million in the FCC’s Rural Digital Opportunity Fund auction. In August 2022 FCC staff rejected the award, finding that SpaceX had not shown it could deliver the promised speeds; the commission upheld the rejection 3–2 in December 2023, with Commissioners Carr and Simington dissenting that SpaceX was being held to targets years early.⁶⁰ In June 2025 the federal agency running the $42.45 billion Broadband Equity, Access and Deployment programme removed its preference for fibre and made the programme technology-neutral. When the last state plans were approved in August 2026, satellites were to serve 21.9 percent of the locations for about 4 percent of the deployment money, and the agency later asked states to trim their satellite awards.⁶¹

Phones that talk to satellites

Satellites now reach ordinary phones directly. Apple paid $450 million, mostly to Globalstar, for emergency texting on the iPhone 14 from November 2022.⁶² T-Mobile and SpaceX announced a partnership in 2022, and on 23 July 2025 T-Mobile’s satellite texting service went live on 657 Starlink satellites, adding data for some apps in October.⁶³ In September 2025 SpaceX agreed to buy spectrum from EchoStar for about $17 billion, half in cash and half in SpaceX stock, to build a faster next generation of the service.⁶⁴ By late 2026 Starlink said its mobile service reached 22 million users a month. AST SpaceMobile, which builds very large antennas in orbit, had launched a few satellites but not yet started commercial service.²⁶⁵

The end of the dish on the roof

The geostationary consumer business could not match low latency at similar prices. HughesNet’s subscribers fell to 622,000 by mid-2026, down almost a quarter in a year. On 3 August 2026 Hughes filed for bankruptcy to reorganise around businesses and the military, and its restructuring officer blamed a structural shift to low-orbit services.⁶⁶ Viasat, which reported record revenue for the year to March 2026, had already turned toward aircraft, ships and governments.⁶⁷

A public company in orbit

SpaceX priced its initial public offering at $135 a share on 11 June 2026 and raised $75 billion, the largest such sale ever. In its first full quarter as a public company it reported Starlink revenue of $5.25 billion, up 69 percent on a year earlier, and its connectivity business provided almost all of SpaceX’s profit.⁴² Average revenue per subscriber was falling over the year as growth moved to cheaper markets and plans, and prices in the United States were restructured, cut on promotion and then raised in June.⁶⁸ On 28 September a Starship rocket reached orbit and released the first of a third generation of Starlink satellites.² On 24 July 2025 the whole network had gone down for about two and a half hours because of a failure in the software that runs its core network, a reminder that a system spread across thousands of satellites can still have a single point of failure on the ground.⁶⁹

What satellite internet is today

Satellite internet is now mainly a low-orbit service. One operator, Starlink, has most of the satellites, most of the subscribers and the only fully deployed consumer constellation. It sells to homes in rural and remote areas, to airlines and ships, to governments and armed forces, and through mobile operators to phones. Eutelsat OneWeb, with European states as its largest shareholders, sells to governments and businesses. Amazon Leo is about to begin service. China’s constellations are being launched at speed. The geostationary operators that served rural homes for two decades are retreating to specialised markets.

The technology is a mesh of satellites a few hundred kilometres up, linked by lasers, talking to dishes that cost a few hundred dollars and increasingly to ordinary phones. The economics is vertical integration: whoever owns a reusable rocket can replace satellites every few years at low cost. The governance is national licensing of a global system, with the ITU coordinating spectrum. No institution governs the shells themselves.

How satellite internet changed reaching the world

From a fixed point to a moving swarm

For half a century a communications satellite was a single object in a fixed place, toward which a dish could be pointed and left. Now it is a swarm, no satellite overhead for more than a few minutes, and the connection is the network handing users from one to the next.

From the last resort to the first choice in remote places

Rural satellite internet was a service of last resort with long delays and data caps. Low orbit turned it into a broadband service with latency of a few tens of milliseconds, and for ships, aircraft and many remote communities the best available. That changed the economics of wiring the countryside: a government asked whether to pay for fibre to a remote house now has a cheaper option, with different long-term costs.

From infrastructure states built to infrastructure a company runs

Telephone networks, satellite systems and the early internet were built or tightly governed by states and their agencies. The most important new global network is owned by a single private company, licensed by one government and used by many others, including armed forces. Its operator’s policies on where it works matter in war and in peace.

From empty orbits to a crowded commons

Low orbit was a place where a dead satellite could linger for decades without much risk. It is now occupied by many thousands of working satellites making hundreds of thousands of avoidance manoeuvres a year. Its use has changed from occasional to continuous, and a shared resource that nobody owns has acquired dominant users.

From a dark sky to a sky with moving lights

For most of history the night sky beyond city lights was the same for everyone. It now contains thousands of moving satellites, recorded as streaks by wide-field telescopes. The change is small for most people and large for some, and nobody was asked.

What the history suggests

Satellite internet repeats several patterns found across this site. Geostationary satellites are a case of a defining strength becoming the constraint: the orbit that let one satellite cover an ocean imposed a delay that no improvement could remove, and Viasat and Hughes improved everything except that. Teledesic is a case of a grand design losing to a small working thing, and Starlink’s dish sold to rural customers who had nothing better is the small working thing. SpaceX’s integration of rocket, satellite and service shows how making something cheap moves the cost to someone else: cheap broadband from low orbit shifted some of its costs to astronomers, to the atmosphere and to every other operator of a satellite in a crowded shell. And the FCC’s five-year rule replaced a permissive default that had been borrowed time.

What is particular to satellite internet is how much depends on distance and on things nobody owns. The difference between a half-second delay and a twentieth of one is the difference between 36,000 kilometres and 550, and the history turned on who could afford to fill the nearer orbit. That orbit belongs to no one under the treaty that governs space, while one company occupies most of it, and the service reaches people no other network reaches through an operator whose decisions bind governments. The philosophy has to start there.

Understanding the philosophy

What follows takes eight ideas that run beneath satellite internet and examines each in turn: in plain terms, in the tradition that has thought hardest about it, in the history above, against the best objection to it, and for what it suggests to people deciding now. Two are psychological. The links drawn to philosophers, lawyers, geographers, economists and psychologists are the author’s interpretation. The engineers and executives in the history were solving practical problems and were not, for the most part, answering these thinkers.

1. A commons in orbit

The idea in plain terms

Low orbit is a resource everyone can use and no one owns. Each satellite placed there takes a little of something shared: room to move without collision, a share of the safe paths through a shell. If every operator acts in its own interest, the space can fill until it becomes dangerous for all of them. Whether that outcome is inevitable, and what prevents it, is one of the oldest questions in political economy.

The tradition

In 1968 the ecologist Garrett Hardin published “The Tragedy of the Commons” in Science. His example was a pasture open to all herders. Each gains the full benefit of adding an animal and bears only a fraction of the cost of overgrazing, so each keeps adding until the pasture is ruined. Hardin’s real worry was population, and his remedy was what he called mutual coercion, mutually agreed upon: rules, taxes or private property. He later said he should have written about the unregulated commons.⁷⁰

The political economist Elinor Ostrom spent decades studying communities that had in fact governed shared fisheries, forests and irrigation systems over long periods without either privatisation or a distant state. Governing the Commons (1990) set out the conditions she found in long-lasting cases: clear boundaries around who may use the resource, rules suited to local conditions and made partly by the users, monitoring by people accountable to the users, graduated sanctions, cheap ways to resolve conflict, and recognition by outside authorities of the users’ right to organise. A 2010 review of 91 studies found these principles well supported. Ostrom shared the 2009 Nobel memorial prize in economics, the first woman to receive it.⁷¹

Where it shows up

The Outer Space Treaty of 1967 states that outer space “is not subject to national appropriation by claim of sovereignty, by means of use or occupation, or by any other means”. No one owns a shell at 550 kilometres. A constellation of thousands of satellites occupies it all the same, and its operator’s avoidance manoeuvres, over 200,000 in six months, are the cost of sharing a crowded commons. Kessler and Cour-Palais supplied the mechanism by which the tragedy could happen. Economists have proposed Hardin’s tax: a 2020 paper in PNAS by Rao, Burgess and Kaffine estimated that a fee per satellite-year, rising over time, could more than quadruple the long-run value of the space industry by reducing congestion. Researchers at MIT have estimated how many satellites low orbit can safely hold and found, in 2025, that greenhouse gases cooling the upper atmosphere could reduce that capacity by half or more by 2100.⁷² The commons is finite, and its size is changing.

The strongest objection

Kessler himself is the best witness against the alarm. In a 2010 paper he and colleagues noted that the term “Kessler syndrome” was coined by someone else, John Gabbard of NORAD, and that popular use had exaggerated and distorted the 1978 paper. Cascading collisions are usually slow and would take decades to transform the environment, though rapid, poorly planned growth could speed them up.⁷³ Low shells also clean themselves: drag pulls failed satellites down in a few years. And unlike Hardin’s herders, the largest user of the shell has a strong private interest in keeping it safe, since it owns most of what a collision would destroy.

That last point cuts both ways. A commons dominated by one user is less exposed to careless strangers and more likely to be run on the dominant user’s rules. Ostrom’s conditions, users making the rules together and monitoring each other, are hard to meet when one participant has two-thirds of the satellites and others are national rivals with whom it does not coordinate. Scholars who apply her work to orbit admit their proposals are untested.⁷¹

What it means today

The site’s idea page Governing What Everyone Shares sets out Ostrom’s principles at length. For orbit they raise plain questions. Who counts as a user of a shell, and how does a newcomer get a say? Does anyone other than the operator check its reports of manoeuvres? Would a fee for occupancy change behaviour more than a five-year disposal rule?

2. First come, first served

The idea in plain terms

When a scarce resource is allocated to whoever claims it first, those who are ready earliest take the best of it. In orbit, the countries and companies with rockets file first, coordinate first and launch first. Latecomers inherit what is left. Is that fair, and by what standard?

The tradition

John Locke, in his Second Treatise of Government, argued that a person who mixes their labour with something in common makes it their property, but only, he wrote, “where there is enough, and as good, left in common for others”. Robert Nozick called this the Lockean proviso in Anarchy, State, and Utopia (1974) and read it as a test of whether an appropriation leaves others worse off.⁷⁴

John Rawls’s A Theory of Justice (1971) offered another test. Imagine choosing the rules of a society from behind a veil of ignorance, not knowing whether you will be rich or poor, early or late. Rawls argued that people in that position would accept inequalities only if they benefit the least advantaged. Critics including Nozick replied that a distribution arising from free and fair exchange is just however unequal it turns out, and that redistributing it to fit a pattern violates liberty.⁷⁵

Where it shows up

The ITU’s constitution adopts both principles at once. Member states must use orbits and frequencies “rationally, efficiently and economically”, and so that countries may have “equitable access” to them, “taking into account the special needs of the developing countries”. In practice coordination priority goes to whoever files first, and speculative filings, sometimes called paper satellites, can block others.¹² The Bogotá Declaration of 1976 was a Lockean complaint in the language of sovereignty: equatorial states argued that the orbit above them was a natural resource that richer countries were taking. The 1977 plan for broadcasting satellites, reserving positions for countries that had not launched, was closer to Rawls. In low orbit there is no such plan. The shells SpaceX occupies were licensed by the FCC, and the best of them were taken first by the company able to launch most.

The strongest objection

Rawls’s own test may favour the incumbent. Starlink’s early occupation of low orbit produced a service that reaches remote villages, ships and war zones no one else served. If the least advantaged include a family in a remote valley with no other connection, they may be better off because one company moved first and fast. Waiting for an equitable plan, as the ITU did for broadcasting, might have produced fairness and no service.

The reply is that the test applies to the rules, not to one outcome. A rule that lets the first mover occupy the shells, keep the spectrum and set the terms of use for decades is a rule people behind the veil would be wise to doubt, because they might be the country with no launcher or the rival that arrives in 2030. Locke’s proviso asks a factual question that no one has answered: whether enough and as good is left for others in low orbit.

What it means today

Regulators allocating a scarce position, an orbital shell or a band of spectrum, can ask Locke’s question directly and require evidence on it. Companies that move first can ask Rawls’s: would the rule that favours us look fair to someone who did not know which side of it they would be on?

3. Who decides the exception

The idea in plain terms

Most of the time a network simply works. In an emergency someone must decide whether it works here, for these people, for this purpose. Whoever makes that decision holds a kind of power that political theorists have long associated with the state. When the decider is a company, the theory has to stretch.

The tradition

In a lecture published in 1919, the sociologist Max Weber defined the state as a human community that successfully claims the “monopoly of the legitimate use of physical force” within a territory.⁷⁶ Three years later the German jurist Carl Schmitt opened Political Theology (1922) with the claim, in one standard rendering, that the sovereign is the one who decides on the state of exception: whoever can suspend the normal rules and act outside them in a crisis.⁷⁷ Schmitt is a troubling source. He sided with the Nazis after 1933 and became known as the regime’s “Crown Jurist”, and critics have traced continuities between his legal theory and his politics. He is used here as a diagnostic, for his question of who in fact decides in an emergency, and not as a guide to what should happen.

More recent work moved the question to networks. In 2019 the political scientists Henry Farrell and Abraham Newman argued in “Weaponized Interdependence” that global networks have a few hubs, and that states with jurisdiction over them gain two powers: a panopticon effect, the ability to see what flows through, and a chokepoint effect, the ability to cut others off. Their cases were SWIFT and the internet. The design theorist Benjamin Bratton, in The Stack (2015), described planetary computation as a layered structure that creates forms of sovereignty alongside those of states.⁷⁸

Where it shows up

The history of Starlink in Ukraine is a sequence of exceptional decisions. The network was switched on by a reply to a tweet. Coverage was not extended to Crimea when Ukraine asked, under a geofence the company had set. Drone use was limited. In 2026 Russian terminals were shut out through a whitelist, and later a speed cap was imposed amid dispute over who had ordered it. The same powers appear in Iran, where Starlink was switched on without a licence and the state answered with jamming, and in Myanmar, where SpaceX disabled 2,500 terminals near scam compounds. In each case a private operator decided where the network worked. Its home government could also act through it, as Farrell and Newman’s framework predicts: the United States sent terminals to Iran and now pays SpaceX to build a military network it will own. Starlink is a hub in their sense, and a company holds the switch.

The strongest objection

Every infrastructure provider chooses where its service runs, and no telephone company or airline is called a sovereign for it. SpaceX’s decisions in Ukraine were mostly in Ukraine’s favour; it complied with American sanctions and law; and its refusal to extend coverage to Crimea followed a policy meant to avoid escalation, a reasonable caution for a private company with no mandate to wage war. Calling this sovereignty flatters a contractor.

Yet the comparison fails at the point that matters. A telephone company’s choices are made in advance, under licences and regulations, and seldom during a battle. The Starlink decisions were made during a war, quickly, often in public posts, about a network that one side’s army depended on, with no agreed procedure and no appeal. Poland’s foreign minister learned the terms of his country’s dependence in an exchange on social media.

What it means today

For governments, the lesson is to settle in contracts, before any crisis, who decides where and how a critical network works, and to keep an alternative, as European states are trying to do with OneWeb and IRIS². For companies whose products become infrastructure, it is to write down their rules for exceptional use before they are tested, and to say who may change them. The site’s idea page Whoever Controls the Gate follows the same problem through other platforms.

4. Universal service

The idea in plain terms

Some services matter so much to ordinary life that a society may decide everyone should have them, wherever they live and whatever it costs to reach them. Electricity, the post, the telephone and now broadband have all been argued about in these terms. The question is how such service gets built: by a monopoly obliged to serve everyone, by public money, or by competition.

The tradition

The phrase “universal service” comes from Theodore Vail, president of AT&T, whose slogan around 1907 was “One Policy, One System, Universal Service”. He meant a single interconnected Bell network, rather than a telephone in every home. In Universal Service (1997) the communications scholar Milton Mueller argued that what actually spread the telephone across America was competition between rival, unconnected networks in the decades before Bell’s monopoly, and that the social meaning of the phrase, service to everyone, was largely a later construction used to defend that monopoly. The Communications Act of 1934 set a goal of making communication service available “so far as possible, to all the people of the United States”. Rural electrification shows what public money could do: federal loans under the Rural Electrification Act of 1936 helped take electricity from about one American farm in ten in 1935 to about nine in ten by the early 1950s. The historian Richard John’s Network Nation (2010) argued that American telecommunications were shaped above all by political choices.⁷⁹

Where it shows up

Satellite internet has always been the answer to the hardest part of universal service, the last remote house. For two decades it was a poor answer, a half-second delay and a data cap, and the FCC did not count it when measuring who was served. Low orbit changed that. The history then shows each model of universal service tried in turn. The FCC’s rural subsidy auction awarded Starlink $885.5 million and then took it back on doubts about capacity. The BEAD programme, rewritten in 2025 to be technology-neutral, gave satellites about a fifth of the remaining locations for a small share of the money. And the private market, Mueller’s engine, reached millions of remote households before any subsidy arrived, because SpaceX chose to sell to them first.

The strongest objection

A satellite subscription is not the same as infrastructure. Fibre laid to a village is a durable public asset; a subscription to a private constellation is a service that can be repriced, rationed or withdrawn, and capacity in each beam is shared, so the quality depends on how many neighbours sign up. Paying a satellite company to serve rural homes may be cheaper now and leave those homes dependent on one supplier for good, where the rural electrification of the 1930s left cooperatives owned by the people they served. Mueller’s competition story also needs competitors. In low orbit there is, so far, one.

Fibre to every remote house costs more than many governments will pay, and while they argued about it millions of people had no service.

What it means today

Public bodies spending on universal service can ask what they are buying: a connection for a household now, or an asset that a community will own and control later. They can write terms on price, capacity and continuity into satellite awards, as NTIA tried to do. And they can recognise that a universal service delivered by a single private network is universal only on that network’s terms.

5. The shrinking of distance

The idea in plain terms

Every communications technology changes what distance means. A message that once took weeks takes seconds, and a place that was remote becomes reachable. But the shrinking is not shared equally. Some people control the speed; others are on the receiving end of it.

The tradition

The Canadian economist Harold Innis, in Empire and Communications (1950), divided media into those biased toward time, heavy and durable like stone and parchment, which favour tradition and continuity, and those biased toward space, light and portable like paper, which favour administration over distance and the growth of empires. The French theorist Paul Virilio, in Speed and Politics (1977), argued that the history of power is a history of speed and of who commands it; he called its study dromology. The geographer David Harvey, in The Condition of Postmodernity (1989), described a “time-space compression” in which falling transport costs and satellite communication so changed the experience of space and time that people had to change how they represented the world, and tied it to capitalism’s turn after 1973 toward flexible production.⁸⁰

The geographer Doreen Massey answered Harvey in 1991. Compression, she argued, is experienced differently by different groups. Some people initiate movement and communication and control it; others are moved, connected or bypassed by it. She called this a power-geometry.⁸¹

Where it shows up

The central fact of this history is distance. Clarke’s relays sat 36,000 kilometres up, and the delay that distance imposed shaped satellite internet for half a century. Moving the satellites to 550 kilometres reduced that delay to a few tens of milliseconds and made a remote farm, a ship or a front line as close to a server as a city flat. In Innis’s terms, Starlink is a medium biased toward space, and among its most important institutional users are armies and governments. Massey’s power-geometry describes the war in Ukraine precisely: the compression of distance that made Ukraine’s units connected was controlled by a company and a government elsewhere, and could be extended, limited or capped from there.

The strongest objection

Theories of speed can make every new network sound like an instrument of empire, which overlooks who uses it. Most Starlink subscribers are households and small businesses choosing a better connection, and the Arctic community gains more than any distant controller gains from it. Innis’s own biographer warned against reading his “bias” as determinism.

Massey would accept that. Her point was never that compression harms the remote, only that they should ask who controls it. A household with one supplier of connection can be well served and still powerless over the terms.

What it means today

Designers of networks that collapse distance can ask Massey’s question of their own product: who controls the flows, and who receives them? For organisations that depend on such a network, the practical answer is knowing how they would work if it were capped, repriced or switched off by someone else.

6. The starry heavens

The idea in plain terms

The night sky is the one view shared by every human generation and culture. A few thousand points of light have guided navigation, calendars, religion and science. Placing tens of thousands of bright moving objects in it changes something held in common across time, for the benefit of people now. Do we owe anything to those who will look up later?

The tradition

Immanuel Kant ended his Critique of Practical Reason (1788) with a sentence that, in Thomas Kingsmill Abbott’s translation, reads: “Two things fill the mind with ever new and increasing admiration and awe, the oftener and the more steadily we reflect on them: the starry heavens above and the moral law within.”⁸² The sky, for Kant, was a reminder of the scale of the world and of the human place in it.

The philosopher Hans Jonas, in The Imperative of Responsibility (1979), argued that modern technology had given people the power to affect the distant future, and that ethics, built for neighbours and the present, needed a new imperative: to act in ways compatible with the continued existence of genuine human life. He proposed a “heuristics of fear”, taking the worst plausible outcome seriously as a guide. His work is often treated as a foundation of the precautionary principle. In 2020 Venkatesan, Lowenthal, Prem and Vidaurri argued in Nature Astronomy that space is an ancestral global commons, including the sky traditions of Indigenous peoples, and that occupying it as quickly as the constellations do risks repeating the patterns of colonisation.⁸³

Where it shows up

The history above records the sky’s change in technical terms: trails across a tenth of Rubin’s images, or a third, depending on the estimate; radio leakage in bands reserved for astronomy; almost every constellation satellite brighter than the IAU’s voluntary limit; metals from re-entering spacecraft in the stratosphere. SpaceX responded with coatings, shades and mirror films, and the IAU set up a centre to negotiate. No one decided, as a matter of public choice, that the sky should hold these satellites. The decision was a series of licences granted for communications purposes.

The strongest objection

The objection runs that light on the ground already hides much of the sky from people who live in cities, so the satellites’ effect on what most people see with the naked eye is small. The loss falls mainly on professional astronomers, whose work matters but who are a small group with their own interests, and whose telescopes can be scheduled and their images processed to remove trails. Jonas’s heuristics of fear, applied consistently, would block almost any new technology, since every one has a plausible worst case. And the people Venkatesan speaks for include remote communities who use the same satellites to reach doctors and schools.

These points reduce the harm without removing the question of consent: a change to something held in common was made through licences issued for one purpose.

What it means today

Regulators licensing constellations can treat brightness and radio emission as conditions of the licence, not as voluntary goodwill. Companies can measure and publish the effect of their satellites on the sky as they publish coverage maps. And anyone weighing a long-lived technology can try Jonas’s exercise once: describe the worst plausible state of the thing held in common after fifty years of growth, and ask who would have agreed to it.

7. What connection does for the remote

The idea in plain terms

Bringing fast internet to a remote community is usually described as an unmixed good: access to doctors, schools, markets and family. Psychology asks what actually happens to people’s well-being, relationships and attention when connection arrives. The answer turns out to be modestly positive on average and more complicated up close.

The tradition

Richard Ryan and Edward Deci’s self-determination theory holds that people have three basic psychological needs: competence, autonomy and relatedness, the sense of being connected to others. Their satisfaction supports motivation and well-being.⁸⁴ Connection to the wider world can serve all three, and can also crowd out the local relationships that serve relatedness best.

The largest evidence on internet access and well-being comes from Matti Vuorre and Andrew Przybylski at Oxford. In a study published in 2024 they analysed Gallup surveys of about 2.4 million people in 168 countries from 2006 to 2021, running 33,792 different statistical specifications. In 84.9 percent of them, internet access and use were associated with higher well-being. A small share of associations with community well-being were negative, mostly among young women. The design is correlational and cannot show cause. An earlier paper found that the associations between internet adoption and national well-being were small at most.⁸⁵ Studies of internet use and loneliness, mostly among older adults, find small and inconsistent effects.⁸⁶

Where it shows up

The one peer-reviewed study of Starlink in remote Indigenous communities found for this essay, by Rob McMahon and colleagues, followed two communities in Canada’s Northwest Territories. Subscribers found it faster, more reliable and better value than older satellite service. They also reported more exposure to harmful content. After a year in one community, residents were more worried about cyberbullying, scams and risks to young people, and used the connection less for public services, education and health and more for entertainment and socialising. Overall they still judged it positive.⁸⁷

The case of the Marubo people in the Brazilian Amazon shows how the story can be told badly. A New York Times report in June 2024 described elders’ worries after Starlink arrived, including minors watching pornography, alongside its benefits in emergencies. Other outlets turned this into a tribe addicted to pornography. The reporter wrote a follow-up saying that the Marubo were not addicted, and a defamation suit by members of the community against the newspaper and others was dismissed in 2025.⁸⁸

The strongest objection

Research on screens and well-being has been dominated by moral panics, and the cautious reading of the evidence is that connection helps a little on average and that the fears are mostly projected by distant observers. The Marubo episode is a lesson in that. People in remote communities are entitled to the same access, distractions included, that city dwellers take for granted, and studies of a few dozen households in two villages should not be turned into warnings about connecting everyone.

That is a strong argument, and the Canadian study supports part of it: residents still judged connection positive. What the study adds is the shift in use over a year, from services toward entertainment, which no average across 168 countries can show.

What it means today

Programmes that connect remote communities can measure what the connection is used for, not only whether it works, and ask residents what they want from it. Funding the local services that make a connection more than entertainment, health care, schooling and government online, matters as much as the dish. And observers can be careful with stories about other people’s screens.

8. Risks too far away to feel

The idea in plain terms

Orbital debris, a brighter sky and metals in the stratosphere are risks most people cannot see, will never encounter directly, and that unfold over decades. Psychology suggests such risks get little weight from the public or from decision-makers, however serious they are. It also suggests caution about how much that explains.

The tradition

In “Perception of Risk”, published in Science in 1987, Paul Slovic summarised a research programme that asked people to rate many hazards. Their judgements clustered along two dimensions: dread, covering lack of control, catastrophic potential and unfair distribution, and unknown risk, covering hazards that are unobservable, new and delayed. Risks high in dread provoke demands for regulation. Lennart Sjöberg later argued that these dimensions explain little of any individual’s judgement.⁸⁹ Yaacov Trope and Nira Liberman’s construal level theory, set out in 2010, holds that the more distant something is in time, space, social distance or likelihood, the more abstractly people represent it.⁹⁰ Applied to climate change, a 2012 British study by Alexa Spence, Wouter Poortinga and Nick Pidgeon found that lower psychological distance was generally linked to concern. But a 2023 review by Anne van Valkengoed, Linda Steg and Goda Perlaviciute found that most experiments failed to show that making climate change feel closer increased action: 25 of 30 did not.⁹¹

Neil Weinstein’s 1980 studies found that people rate their own chances of bad events as lower than other people’s, an effect called unrealistic optimism. Adam Harris and Ulrike Hahn argued in 2011 that much of the effect could arise from statistical artefacts of how it was measured, and a later set of studies controlling for those artefacts found no remaining optimism.⁹²

Where it shows up

Debris in low orbit scores high on Slovic’s unknown dimension and low on dread: invisible, new, delayed, affecting satellites rather than people. Responses came slowly and through specialists. The 25-year disposal guideline survived until 2022, by which time thousands of satellites were being launched each year. Astronomers’ objections were met with engineering improvements rather than limits. The site’s idea page When Small Risks Become Normal describes how organisations come to accept rising risk one step at a time, and the growth in avoidance manoeuvres from year to year has that shape.

The strongest objection

The replication record argues for modesty: psychological distance seems to matter less than its popularity implies, and the optimism research may describe a measurement problem. The slow response to debris has a simpler explanation. The risks were uncertain, Kessler himself said cascades would be slow, and regulators acted once satellite numbers rose. Calling that a bias may only mean disagreeing with the judgement.

The psychological account remains useful at the level where it is strongest. Slovic’s dimensions describe which risks provoke public demands and which are left to experts, and space debris and the night sky are clearly in the second group. That allocation is itself a decision, made without much discussion.

What it means today

Leaders responsible for a slow, invisible risk cannot wait for public pressure, which will come late or not at all. They can set thresholds in advance and publish measurements against them. A useful test for any company is whether a risk it imposes would be acceptable if the people bearing it could see it.

Bringing the threads together

The eight ideas meet at a few points. The commons and first come, first served are two views of the same fact: low orbit is a shared resource without an owner, and those able to launch have taken its best parts first, under rules that call for equitable access and reward early filing. Sovereignty and universal service meet over the last house and the front line, where a private network delivers something societies have long treated as a public responsibility, and its operator decides the exceptions. Innis, Virilio, Harvey and Massey explain why distance, the half-second that geostationary orbit imposed, mattered so much, and who controls its removal. Kant and Jonas, together with the psychology of distant risk, describe costs that fall on people who are not in the market at all: astronomers, future operators, the next generation looking up.

Running through all of them is one change. A set of decisions that had been made by states and international bodies, about orbits, spectrum, rural service and the use of networks in war, came to be made in large part by one company, at its own pace, with governments licensing and paying.

The network overhead

The network overhead does what Clarke promised and more. It reaches a remote farm, a ship in mid-ocean and a trench in eastern Ukraine with a delay small enough for a video call, and it did so faster than any public programme had managed. It also occupies a shared orbit, changes a shared sky and makes one company’s policies matter to governments at war. The history suggests a few working principles.

Distance is a design decision. The half-second delay of geostationary orbit was accepted in 1965 for good reasons and shaped everything built on it for fifty years.

Capital structure decides which visions become real. Teledesic’s plan was right and early; SpaceX built nearly the same thing two decades later because it owned a reusable rocket and could pay for satellites from launch revenue. When an idea fails, it is worth asking whether the idea failed or the money did.

A product that becomes infrastructure acquires duties its makers did not plan for. A dish sold below cost to rural customers became, within two years, part of a war, and the rules for its use were written during the fighting.

Shared resources need rules before they are full. The FCC’s five-year rule came after thousands of satellites, the IAU’s brightness guideline is voluntary, and no institution governs who may occupy a shell. Each was easier to set when the orbit was nearly empty.

And the sky belongs to more people than the market does. The subscribers of satellite internet number in the tens of millions; the people who look up number in the billions, including those not yet born. Licensing it as a communications resource alone leaves most of them out of the decision.

Selected bibliography

• Locke, John. Second Treatise of Government.
• Kant, Immanuel. Critique of Practical Reason. 1788. Translated by Thomas Kingsmill Abbott.
• Weber, Max. “Politik als Beruf.” Lecture, Munich, 1919. English translation in From Max Weber: Essays in Sociology, edited by H. H. Gerth and C. Wright Mills. New York: Oxford University Press, 1946, 77–128.
• Schmitt, Carl. Politische Theologie. 1922. English translation: Political Theology, translated by George Schwab. Cambridge, MA: MIT Press, 1985.
• Clarke, Arthur C. “Extra-Terrestrial Relays: Can Rocket Stations Give World-wide Radio Coverage?” Wireless World, October 1945, 305–308.
• Innis, Harold A. Empire and Communications. Oxford: Clarendon Press, 1950.
• Hardin, Garrett. “The Tragedy of the Commons.” Science 162, no. 3859 (1968): 1243–1248.
• Rawls, John. A Theory of Justice. 1971.
• Nozick, Robert. Anarchy, State, and Utopia. 1974.
• Virilio, Paul. Vitesse et politique. Paris: Galilée, 1977. English translation: Speed and Politics, translated by Mark Polizzotti. New York: Semiotext(e), 1986.
• Kessler, Donald J., and Burton G. Cour-Palais. “Collision Frequency of Artificial Satellites: The Creation of a Debris Belt.” Journal of Geophysical Research 83, no. A6 (1978): 2637–2646.
• Jonas, Hans. Das Prinzip Verantwortung. 1979. English translation: The Imperative of Responsibility. Chicago: University of Chicago Press, 1984.
• Weinstein, Neil D. “Unrealistic Optimism about Future Life Events.” Journal of Personality and Social Psychology (1980).
• Slovic, Paul. “Perception of Risk.” Science 236, no. 4799 (1987): 280–285.
• Harvey, David. The Condition of Postmodernity. Oxford: Blackwell, 1989.
• Ostrom, Elinor. Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge: Cambridge University Press, 1990.
• Massey, Doreen. “A Global Sense of Place.” Marxism Today 38 (1991): 24–29.
• Mueller, Milton L. Universal Service: Competition, Interconnection, and Monopoly in the Making of the American Telephone System. Cambridge, MA: MIT Press; Washington: AEI Press, 1997.
• Ryan, Richard M., and Edward L. Deci. “Self-Determination Theory and the Facilitation of Intrinsic Motivation, Social Development, and Well-Being.” American Psychologist 55, no. 1 (2000): 68–78.
• Chan, Sharon Pian. “The Birth and Demise of an Idea: Teledesic’s Internet in the Sky.” Seattle Times, 7 October 2002.
• Trope, Yaacov, and Nira Liberman. “Construal-Level Theory of Psychological Distance.” Psychological Review 117, no. 2 (2010): 440–463.
• Cox, Arnold and Villamayor Tomás. “A Review of Design Principles for Community-Based Natural Resource Management.” Ecology and Society 15, no. 4 (2010): 38.
• John, Richard R. Network Nation: Inventing American Telecommunications. Cambridge, MA: Belknap Press of Harvard University Press, 2010.
• Kessler, Johnson, Liou and Matney. “The Kessler Syndrome: Implications to Future Space Operations.” AAS 10-016, 2010.
• Spence, Alexa, Wouter Poortinga, and Nick Pidgeon. “The Psychological Distance of Climate Change.” Risk Analysis 32, no. 6 (2012): 957–972.
• Bratton, Benjamin H. The Stack: On Software and Sovereignty. Cambridge, MA: MIT Press, 2015.
• Farrell, Henry, and Abraham L. Newman. “Weaponized Interdependence: How Global Economic Networks Shape State Coercion.” International Security 44, no. 1 (2019): 42–79.
• Rao, Burgess and Kaffine. “Orbital-Use Fees Could More Than Quadruple the Value of the Space Industry.” Proceedings of the National Academy of Sciences 117 (2020): 12756–12762.
• Venkatesan, Lowenthal, Prem and Vidaurri. “The Impact of Satellite Constellations on Space as an Ancestral Global Commons.” Nature Astronomy, published online 6 November 2020.
• Isaacson, Walter. Elon Musk. New York: Simon & Schuster, 2023.
• Murphy et al. “Metals from Spacecraft Reentry in Stratospheric Aerosol Particles.” Proceedings of the National Academy of Sciences 120, no. 43 (2023).
• van Valkengoed, Anne M., Linda Steg, and Goda Perlaviciute. “The Psychological Distance of Climate Change Is Overestimated.” One Earth (2023).
• Vuorre, Matti, and Andrew K. Przybylski. “A Multiverse Analysis of the Associations Between Internet Use and Well-Being.” Technology, Mind, and Behavior (2024).
• Ferreira, Huang, Nomura and Wang. “Potential Ozone Depletion from Satellite Demise during Atmospheric Reentry in the Era of Mega-Constellations.” Geophysical Research Letters 51, no. 11 (2024).
• McMahon, Rob, Murat Akçayır, Brenda Norris, and Lyle Fabian. “Assessing the Impacts of Low-Earth Orbital Satellite Systems in Remote Indigenous Communities.” Telecommunications Policy 49, no. 2 (2025).
• Parker, Brown and Linares. “Greenhouse Gases Reduce the Satellite Carrying Capacity of Low Earth Orbit.” Nature Sustainability (2025).

Research limitations

Much of the recent history rests on company statements relayed by the press. Starlink’s subscriber milestones before 2026 come from company announcements reported by trade and fan sites; its 2026 figures come from SpaceX’s filings and results as reported by CNBC, and some details of the third-quarter results, such as the 22 million users of its mobile service, come from a call transcript. Counts of Starlink satellites differ by tracker and definition: about 11,200 in orbit by one count in November 2026 and 9,929 operational by another. Avoidance manoeuvre totals are SpaceX’s own reports, not independently audited. SpaceX has never published its internal launch cost, and no estimate is used here. The early cost of the user terminal is reported as $1,300 and as $1,500 for the same remarks.

Several figures conflict between sources. Accounts of the Bogotá Declaration give seven or eight signatories, depending on whether Brazil is counted. Walmart’s first satellite network is dated 1983 and 1985. The number of Ukrainian terminals bought by USAID is given as 1,333, about 1,500 and 1,508, and the inspector general’s report is internally inconsistent; SpaceX’s stated costs for Ukraine in 2022 vary between outlets, and so do reports of what Poland pays each year, for which this essay uses Sikorski’s own figure. Estimates of the share of Rubin Observatory images affected by trails range from about 10 percent to 30 to 40 percent under different assumptions. The LOFAR finding that second-generation satellites leak more radio emission is reported at up to 32 times in secondary coverage. HughesNet’s subscribers are reported at 622,000 and 641,000 for mid-2026. Reports of the stakes taken by the UK and Bharti in OneWeb in 2020 differ. The accounts of Musk and Wyler discussing cooperation in 2015, SoftBank’s decision not to invest further in OneWeb, the Reuters report on Starlink in the 2025 minerals talks (denied by Musk and Rubio), the Wall Street Journal’s reports of terminals sent to Iran and of the October 2026 speed cap, and the reported FCC approval of the EchoStar spectrum purchase each rest on a single source or anonymous sourcing. Who ordered the speed cap on military use in Ukraine in October 2026, and why, is unresolved. Iran’s claims to have disabled Starlink terminals have not been independently verified.

The account of Starlink and Crimea in September 2022 follows Walter Isaacson’s correction of his biography and Musk’s statement; the version in which Starlink was switched off has been retracted. The atmospheric effects of re-entering satellites are model projections, apart from the measurement of metals in stratospheric particles; no effect on ozone has been measured. The 2026 brightness study cited is a preprint. Pending at publication: whether Amazon Leo begins commercial service in 2026, the closing of the EchoStar spectrum transfer and of Amazon’s purchase of Globalstar, Hughes’s bankruptcy, Starlink’s commercial launch in India and its licensing in South Africa, and states’ revisions of their BEAD satellite awards. Exact wording of Hardin’s best-known line, Vail’s annual report and the Schwab translation of Schmitt was not verified, and those works are paraphrased.

The philosophical connections to Locke, Kant, Weber, Schmitt, Innis, Hardin, Rawls, Nozick, Virilio, Jonas, Weinstein, Slovic, Harvey, Ostrom, Massey, Mueller, Ryan and Deci, Trope and Liberman, John, Spence and colleagues, Bratton, Farrell and Newman, Venkatesan and colleagues, van Valkengoed and colleagues, Vuorre and Przybylski, and McMahon and colleagues are the author’s interpretation; their works are paraphrased apart from short, attributed phrases. The decisions in the decisions box are judged by the author, separating what could be known at the time (process) from what happened (call).

Footnotes

1. “Elon Musk says SpaceX’s Starlink is active in Ukraine,” Space.com; Fortune, 28 February 2022. ↩

2. “SpaceX Q3 2026 earnings live updates,” CNBC, 20 October 2026; Motley Fool call transcript. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

3. KeepTrack Starlink count, November 2026; Orbital Radar briefing, 1 November 2026; active-satellite totals from azmth. ↩ ↩²

4. “SpaceX raises $75 billion in record-setting IPO,” CNBC, 11 June 2026; NPR via Houston Public Media, 12 June 2026. ↩ ↩²

5. ITU, “Facts and Figures 2025”, 17 November 2025. ↩

6. Arthur C. Clarke, “Extra-Terrestrial Relays,” Wireless World, October 1945, 305–308, reproduction; Lakdiva Clarke archive; on the patent, Really Rocket Science. ↩ ↩²

7. NASA, “Sputnik” and “Communications satellites”; Wikipedia, “Telstar 1”; Wikipedia, “Syncom”. ↩

8. NASA, “Communications satellites”; US Centennial of Flight Commission, “Intelsat”; Wikipedia, “Intelsat I”. ↩

9. Satsig, “Satellite latency”; SpeedGuide. ↩

10. ITU, “Radio regulatory framework for space services”; ITU-R, WRS16 document on the broadcasting-satellite plan; IISL 2013 paper. ↩ ↩²

11. Wikipedia, “Bogota Declaration”; NYU GlobaLex, “Paper satellites and the free use of outer space”; Denver Journal of International Law and Policy. ↩

12. ITU Constitution, Article 44(2), as quoted in NYU GlobaLex. ↩ ↩²

13. Hughes, “Recognized by GVF” and “Defining moments in Hughes’s 50-year history”. ↩

14. Tedium, “The evolution of satellite internet”; PC Magazine, 1999, via tech-insider. ↩

15. Sharon Pian Chan, “The birth and demise of an idea: Teledesic’s internet in the sky”, Seattle Times, 7 October 2002; Wikipedia, “Teledesic”; RCR Wireless, 25 May 1998. ↩ ↩²

16. Wikipedia, “Celestri”; “SkyBridge constellation grows,” Wired News, June 1998; RCR Wireless, May 2000. ↩

17. RCR Wireless, September 1999; “The rise and fall and rise of Iridium,” Air & Space; Loral 8-K on Globalstar; Flightglobal on ORBCOMM. ↩

18. Tim Farrar, “LEO HTS: once again a distraction?”, Via Satellite, 23 March 2015. ↩

19. Harry W. Jones, “The recent large reduction in space launch cost”, NASA, 2018; CSIS Aerospace Security. ↩

20. Viasat satellite history; “ViaSat-1 rides into orbit,” Satellite Today; Satellite Today on Exede; EchoStar XVII release; HughesNet Gen5 release; Viasat: ViaSat-2; EchoStar subscriber figures; on Viasat’s 2020 subscribers, TipRanks; on ViaSat-3, SpaceNews and Space Intel Report. ↩ ↩² ↩³

21. FCC, “2015 Broadband Progress Report”; Benton summary. ↩

22. Wikipedia, “O3b Networks”; “SES exercises option to buy 100% of O3b,” SpaceNews, May 2016. ↩

23. Wikipedia, “Eutelsat OneWeb”; “OneWeb files for bankruptcy,” Space.com; on SoftBank, SpaceNews; on the rescue, CNBC; on the committee, Computer Weekly. ↩ ↩² ↩³

24. “SpaceX satellite projects boosted by $1 billion Google, Fidelity investment,” Via Satellite, 21 January 2015; GeekWire; UPI. ↩

25. NASASpaceflight on SES-10; Wikipedia, “Falcon 9 first-stage landing tests”; The Space Review. ↩ ↩²

26. “SpaceX’s broadband ambitions get green light from FCC,” Via Satellite; TechCrunch, 15 November 2018; Spaceflight Now, 30 April 2019; Gizmodo. ↩

27. “Record-setting Falcon 9 launches first 60 Starlink satellites,” AmericaSpace; Space.com. ↩

28. “SpaceX Starlink service priced at $99 a month,” CNBC, 27 October 2020. ↩

29. “SpaceX spending $1,500 per Starlink dish,” Business Insider via Yahoo; Space Intel Report; “Starlink terminal costs coming down,” Advanced Television, September 2023. ↩

30. “SpaceX starts general rollout of Starlink’s inter-satellite lasers,” ISPreview, September 2021. ↩

31. Subscription Insider; Drive Tesla Canada, 9 million; Drive Tesla Canada, 12 million. ↩

32. “United Airlines to introduce free Starlink wifi to entire fleet,” Skift, 13 September 2024; The Week. ↩

33. UK Department for Business, Energy and Industrial Strategy, “OneWeb: request for ministerial direction” and “ministerial direction”, 26 June 2020; Civil Service World. ↩

34. Eutelsat, “Success of its €670 million rights issue”, 12 December 2025; Airbus, January 2026; DCD on 2025–26 results. ↩ ↩²

35. “FCC says Amazon can proceed with Kuiper satellites,” GeekWire, July 2020; Amazon, “Amazon secures up to 83 launches”, 5 April 2022; ULA, October 2023; Via Satellite on the Amazon Leo rebrand. ↩ ↩²

36. “Amazon asks FCC to extend satellite limit,” CNBC, 30 January 2026; Benton on the FCC waiver; KeepTrack Kuiper count; Benzinga on Amazon’s Q3 call. ↩

37. “Amazon to buy Apple satellite partner Globalstar,” MacRumors, 2 April 2026. ↩

38. Prime Minister of Canada, 13 September 2024; “Telesat’s Lightspeed service launch slips to 2028,” Via Satellite, 17 March 2026. ↩

39. Space.com on Qianfan; Space.com on Guowang; CircleID update. ↩

40. Space.com on USAID’s release; Benton summary of the Washington Post; USAID Office of Inspector General report; Kyiv Independent on CNN’s report; Euromaidan Press on Poland. ↩

41. “Elon Musk biographer moves to clarify details on Ukraine Starlink,” CNBC, 9 September 2023; Snopes; TheWrap. ↩

42. “SpaceX restricted Ukraine’s use of Starlink internet for drones,” Space.com; Defense One, February 2023. ↩

43. “Pentagon confirms SpaceX deal for Ukraine Starlink services,” Defense News, 1 June 2023; Reuters report on Starshield, via Times of Israel, March 2024. ↩

44. “Musk denies US threat to cut Starlink over Ukraine minerals deal,” Kyiv Independent; “‘Be quiet, small man’,” Al Jazeera, 10 March 2025; Notes from Poland. ↩

45. “Starlink blocks Russian use in Ukraine,” CNN, 5 February 2026; Ukraine Ministry of Defence; Euronews fact check. ↩

46. Wikipedia, “Starlink in the Russo-Ukrainian war”; “The untold story of Ukraine’s Starlink crisis,” Asia Times, October 2026. ↩

47. “Elon Musk activating Starlink in Iran as US eases sanctions,” Gizmodo, September 2022; CNN, 13 January 2026; “Iran targets Starlink as ‘tool of war’,” Al Jazeera, 14 October 2026; on the Wall Street Journal report, Middle East Monitor. ↩

48. “SpaceX disables 2,500 Starlink terminals at scam compounds in Myanmar,” NBC News; The Record, October 2025. ↩

49. MercoPress, 30 August 2024; Reuters, via Stabroek News, 14 September 2024. ↩

50. “What’s holding back Starlink’s Indian foray,” Business Standard, May 2026; BusinessDay on South Africa, 12 December 2025; TechCentral. ↩

51. Donald J. Kessler and Burton G. Cour-Palais, “Collision frequency of artificial satellites: the creation of a debris belt,” Journal of Geophysical Research 83, no. A6 (1978), NASA NTRS. ↩

52. NASA NTRS paper on the 2009 collision; Space.com. ↩

53. “SpaceX satellites’ encounters with Chinese space station were not emergencies, US says,” South China Morning Post; Breaking Defense. ↩

54. “FCC adopts 5-year rule for deorbiting satellites,” Satellite Today, 30 September 2022; Washington Times. ↩

55. “Every SpaceX Starlink satellite has to dodge a collision almost weekly,” Space.com; Kratos summary of SpaceX’s semi-annual report. ↩

56. The Register, December 2025; “Starlink satellite breaks up,” EarthSky, March 2026; Yahoo Tech on lowering orbits. ↩

57. Rubin Observatory on satellite constellations; IAU CPS background; SpaceX, “Gen2 Starlink satellites: brightness mitigation”; The Register on LOFAR, 2023; Bassa et al., Astronomy & Astrophysics, 2024. ↩ ↩²

58. “Satellite constellations are too bright,” Universe Today; Satnews, January 2026; ESO science paper eso2607. ↩

59. NOAA Chemical Sciences Laboratory, “Metals from spacecraft reentry in stratospheric aerosol particles”, 2023; AGU, “Satellite megaconstellations burn and deplete ozone”, 2024. ↩

60. “FCC denies SpaceX bid for nearly $1 billion in broadband subsidies,” CNBC, 10 August 2022; FCC, Order on Review FCC 23-105, dissent; TechCrunch, 12 December 2023. ↩

61. NTIA, “BEAD Restructuring Policy Notice”, 6 June 2025; “With all BEAD plans approved,” Broadband Breakfast, 28 August 2026; StateScoop on the SpaceX rider. ↩

62. “Apple spending $450 million with Globalstar,” CNBC, 10 November 2022. ↩

63. T-Mobile, “Reaching for the stars with T-Satellite”; Business Wire, 1 October 2025. ↩

64. “SpaceX scoops up EchoStar spectrum for $17B,” Fierce Network, September 2025; The Register. ↩

65. AST SpaceMobile, third-quarter 2026 business update, 9 November 2026. ↩

66. “Hughes files for Chapter 11,” Via Satellite, 3 August 2026; Investing.com on EchoStar’s second quarter. ↩

67. Viasat, fiscal 2026 results, 8-K exhibit. ↩

68. Notebookcheck on the $39 offer; KeepTrack cost guide; on falling average revenue, SpaceX 10-Q. ↩

69. ThousandEyes, “Starlink outage analysis”; Engadget, July 2025. ↩

70. Garrett Hardin, “The Tragedy of the Commons,” Science 162 (1968), 1243–1248; Conversable Economist; Wikipedia. ↩

71. Elinor Ostrom, Governing the Commons (1990), summarised at SESMAD; Cox, Arnold and Villamayor Tomás, Ecology and Society 15, no. 4 (2010), DOAJ; on orbit, Morin and Richard, “Astro-environmentalism”, Global Policy 12, no. 4 (2021). ↩ ↩²

72. Outer Space Treaty, Article II, as quoted in NYU GlobaLex; on orbital-use fees, EurekAlert; on carrying capacity, MIT AeroAstro. ↩

73. Kessler, Johnson, Liou and Matney, “The Kessler Syndrome: implications to future space operations”, AAS 10-016, 2010. ↩

74. Wikipedia, “Lockean proviso”; Libertarianism.org. ↩

75. Oxford University Press, key thinker: Rawls; Wikipedia, “A Theory of Justice”. ↩

76. Max Weber, “Politics as a Vocation,” in Gerth and Mills, From Max Weber (1946), text. ↩

77. Lars Vinx, “Carl Schmitt”, Stanford Encyclopedia of Philosophy, revised 2025. ↩

78. Henry Farrell and Abraham Newman, “Weaponized Interdependence”, International Security 44, no. 1 (2019); Wikipedia, “The Stack (philosophy)”. ↩

79. Wikipedia, “Universal service”; Milton Mueller, Universal Service, reviewed in Project MUSE; 47 U.S.C. 151; “Rural Electrification Act (1936)”; review of Network Nation in Enterprise and Society. ↩

80. Wikipedia, “Empire and Communications”; Semiotext(e), Speed and Politics; David Harvey, “Time-space compression and the postmodern condition”. ↩

81. Doreen Massey, “A Global Sense of Place”, Marxism Today 38 (1991), 24–29. ↩

82. Immanuel Kant, Critique of Practical Reason, trans. Thomas Kingsmill Abbott, Conclusion, Project Gutenberg. ↩

83. On Jonas, The Collector; Venkatesan, Lowenthal, Prem and Vidaurri, “The impact of satellite constellations on space as an ancestral global commons”, Nature Astronomy, 2020. ↩

84. Richard M. Ryan and Edward L. Deci, “Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being”, American Psychologist 55, no. 1 (2000). ↩

85. Oxford Internet Institute, “Internet use statistically associated with higher wellbeing”, May 2024; Vuorre and Przybylski, 2024; OII on the 2023 study. ↩

86. SSRC MediaWell summary of a meta-analysis; Shah et al., Journal of Medical Internet Research, 2021. ↩

87. McMahon, Akçayır, Norris and Fabian, Telecommunications Policy 49, no. 2 (2025); Internet Society Pulse, 30 September 2025. ↩

88. The Quint fact-check; AP via ClickOrlando; AP via Turtle Island News, September 2025. ↩

89. Paul Slovic, “Perception of Risk”, Science 236 (1987), 280–285; Lennart Sjöberg, comment in Risk Analysis, 2002. ↩

90. Yaacov Trope and Nira Liberman, “Construal-level theory of psychological distance”, Psychological Review 117, no. 2 (2010). ↩

91. Spence, Poortinga and Pidgeon, “The psychological distance of climate change”, Risk Analysis 32, no. 6 (2012); van Valkengoed, Steg and Perlaviciute, “The psychological distance of climate change is overestimated”, One Earth, 2023. ↩

92. Harris et al., “Unrealistic comparative optimism”, PLOS ONE, 2017. ↩