Satellite Internet: How Starlink, Amazon Leo, OneWeb and Direct-to-Cell Are Building a Space-Based Internet

 The Internet Is Moving Into Orbit

Starlink made satellite broadband visible to millions of people. But the bigger story is no longer one company or one dish on a roof. A second layer of the Internet is taking shape above Earth — thousands of moving routers, linked by radio and increasingly by laser, with some already capable of talking directly to ordinary phones.

Satellite internet is no longer a niche connection for remote places. Thousands of spacecraft are becoming part of the infrastructure that connects homes, aircraft, ships and entire regions to the global internet.

Orbit	Typical altitude	Main advantage	Where it fits
GEO	~35,786 km	Very large coverage per satellite	Highest latency; classic satellite broadband and broadcast
MEO	~2,000–35,786 km	Middle ground between coverage and latency	Enterprise, government, mobility; SES O3b mPOWER
LEO	~300–2,000 km	Low latency; many satellites required	Starlink, OneWeb, Amazon Leo, Lightspeed, direct-to-device

Starlink did not invent satellite internet. It changed what people expected from it.

For decades, “satellite internet” usually meant a dish pointed at a spacecraft that seemed almost motionless in the sky. The satellite sat in geostationary orbit, about 35,786 kilometers above Earth, and covered an enormous region. That architecture was brilliant for television and invaluable where cables never arrived. It also came with a limit no software update could remove: distance.

A signal traveling from a home to a geostationary satellite and back to Earth covers roughly 72,000 kilometers before we count the return journey, terrestrial routing, processing or queueing. Even at the speed of light, those distances add noticeable delay. For web browsing it can be acceptable. For cloud gaming, video calls, remote control or other interactive work, it is hard to hide.

Low-Earth-orbit broadband flips the design. Instead of relying on a few very distant spacecraft, it uses hundreds or thousands of satellites only a few hundred to a couple of thousand kilometers up. The shorter path cuts propagation delay dramatically, but creates a harder networking problem: the satellites are constantly racing across the sky. A terminal cannot stay attached to one of them. It has to hand traffic from satellite to satellite while the network keeps recalculating where each packet should go.

The result is less like a giant radio tower in the sky and more like a global network whose routers never stop moving. Starlink is the largest example, but it is only one version of an architecture now being pursued by commercial operators and governments around the world.

First, what actually happens when you open a website through a satellite?

At first glance, the path looks simple. Your terminal sends a radio signal upward, a satellite receives it, and the data comes back down through a gateway connected to the terrestrial Internet. The reply follows the same general route in reverse. Modern networks add several layers of complexity to that basic trip.

The user terminal, for example, is not simply a passive dish. Modern flat terminals use phased-array antennas: many small antenna elements whose signals are combined electronically so the beam can be steered without physically turning the antenna. The terminal follows satellites moving overhead at several kilometers per second, then hands the connection to the next one before the current satellite disappears toward the horizon.

In orbit, satellites divide the ground into radio cells with steerable beams, shifting capacity toward areas where demand is highest. On the ground, gateway stations connect the constellation to fiber networks and Internet exchange points. Increasingly, though, a packet does not need to return to Earth at the first opportunity. Optical inter-satellite links — laser connections between spacecraft — can carry it across several satellites before it descends near its destination.

That opens an interesting possibility. Fiber is extremely fast, but light travels more slowly through glass than through a vacuum, and real terrestrial routes rarely follow perfect great-circle paths. On some very long routes, a well-routed optical satellite mesh can therefore compete surprisingly well on latency. That does not make space automatically faster: congestion, handovers, gateway placement, weather and routing decisions can matter more than the theoretical speed advantage.

Satellite internet can operate from very different altitudes. GEO satellites provide enormous coverage from nearly 36,000 kilometers above Earth, MEO systems occupy the middle ground, while LEO networks trade wide individual coverage for much lower latency and thousands of fast-moving satellites.

Why LEO feels so different from old satellite broadband

Latency is the difference users notice first. Early independent measurements of Starlink found minimum round-trip times around 20 milliseconds to nearby destinations — a completely different experience from traditional geostationary broadband. Larger studies later complicated the picture: real-time applications such as video calls and streaming can work very well, but performance is less uniform than fiber because the network underneath the user is constantly changing.

A 2024 ACM study drawing on millions of crowdsourced measurements, together with controlled experiments in several countries, traced that variability to more than the radio link itself. Handovers, ground-station placement and periodic network reconfiguration all leave fingerprints on performance. Global modeling work points in the same direction: geography and atmospheric conditions matter. There is no single “Starlink speed.” What a user experiences depends on location, weather, local demand and how the constellation is routing traffic at that moment.

That explains an apparent contradiction. In a remote area, LEO broadband can feel remarkably close to a good terrestrial connection. In a dense city, fiber is still the stronger architecture. The first comparison is with whatever slow or nonexistent connection was available before; the second is with a cable capable of carrying enormous and comparatively stable capacity into a neighborhood.

Starlink in 2026: the first megaconstellation at true Internet scale

Starlink has reached a scale where it makes more sense to think of it as infrastructure than as a fleet of spacecraft. By mid-2026, SpaceX said the service had about 12 million subscribers and was available across more than 160 countries, territories and markets. By September, Reuters reported more than 11,000 Starlink satellites in orbit. The exact total changes continually as spacecraft are launched, maneuvered, retired and deorbited.

Those numbers matter, but not for the obvious reason. A broadband constellation is not useful simply because satellites are overhead. It needs enough spacecraft in the right orbital planes, enough spectrum, enough gateways, enough terminals and enough terrestrial backhaul to carry real traffic. SpaceX has an unusual structural advantage here: it manufactures the satellites, operates the network and owns the rockets that launch most of them. Reusable Falcon 9 boosters turned replenishment from an occasional space mission into something closer to industrial logistics.

Starlink V3 pushes that logic further. SpaceX says each V3 satellite is designed for up to 1 terabit per second of downlink capacity and 160 gigabits per second of uplink, with far more steerable beams than the previous generation. On September 28, 2026, Starship reached orbit and deployed 26 V3 satellites on its first revenue-generating orbital mission. For Starlink, that was more significant than another routine launch: if Starship can eventually carry large batches of these heavier, more capable satellites on a regular cadence, SpaceX can add capacity much faster than Falcon 9 allows.

Even then, satellite throughput is only one layer of the machine. Starlink also depends on gateway sites, terrestrial backbone links, laser crosslinks and software that has to route packets through a topology changing minute by minute. The engineering achievement is not simply sending hundreds of megabits to a dish. It is making thousands of moving spacecraft behave coherently enough that the person using the connection rarely has to think about any of them.

A Starlink-style connection does not always travel straight from a satellite back to the nearest ground station. Optical inter-satellite links can move data across several spacecraft before it descends through a gateway connected to terrestrial fiber networks.

The competitors are not simply building “another Starlink”

Constellation size is an easy scoreboard, and a misleading one. These networks are being built for different customers and different economics. Some are chasing millions of households; others are aimed at airlines, ships, mobile operators, governments or remote industrial sites. Some live in LEO, some in MEO, some in GEO, and a growing number of future services are likely to combine more than one orbit.

Amazon Leo: the most obvious mass-market challenger

Amazon Leo — formerly Project Kuiper — is the clearest attempt to build a mass-market LEO broadband rival backed by a company with global consumer infrastructure. Amazon plans more than 3,000 satellites connected by optical links, supported by gateway stations and compact user antennas. By mid-2026 it had launched roughly 400 spacecraft: nowhere near Starlink’s scale, but enough to move the project out of the prototype phase and toward initial service.

Its hardest problem is not only building satellites; it is getting them into orbit fast enough. Amazon has booked launches on Ariane 6, Atlas V, Vulcan, New Glenn and even Falcon 9. That spreads risk across several providers, but also exposes the schedule to delays in several rocket programs. The 2026 agreement to acquire Globalstar adds a second strategic piece: spectrum and infrastructure for direct-to-device services, giving Amazon a path toward phones as well as dedicated broadband terminals.

Eutelsat OneWeb: fewer satellites, different customers

OneWeb is already operational with more than 600 LEO satellites, but it is not trying to put a dish on every roof. Eutelsat focuses the network on telecom operators, governments, enterprises, aviation and other professional customers. OneWeb satellites orbit higher than Starlink’s, allowing broad coverage with fewer spacecraft, while accepting different latency and capacity trade-offs.

In September 2026, Eutelsat ordered 229 additional satellites from Airbus on top of 440 already procured for renewal and expansion. The strategic significance is at least as important as the technical one. OneWeb remains the largest operational LEO broadband network outside Starlink, giving European governments an existing platform they can fold into wider plans for communications resilience and sovereignty.

Telesat Lightspeed: Canada is betting on enterprise-grade LEO

Telesat Lightspeed is being built less like a consumer utility and more like a high-performance global network for telecom, enterprise and government users. In 2026, Telesat expanded the planned initial constellation to 225 satellites while continuing work on landing stations and network infrastructure. The first production satellites were targeted for late 2026, with global service planned for 2028. Optical crosslinks, digital beamforming and onboard processing are central to the design, allowing traffic to be routed through the constellation rather than simply bounced back to the nearest gateway.

SES O3b mPOWER: the case for MEO

Low latency does not automatically require thousands of satellites. SES uses medium Earth orbit for O3b mPOWER — much higher than LEO, but far below geostationary orbit. In September 2026, SES launched the final three spacecraft in the 13-satellite second-generation constellation. MEO adds more propagation delay than a low LEO shell, but each satellite can see a much larger part of Earth, letting SES serve high-value enterprise, government, maritime and cloud customers with a far smaller fleet.

That middle layer is a useful reminder that the future will probably not be a simple contest between LEO and fiber. A ship, airliner, mine or government network may value guaranteed capacity, redundancy and security more than a few milliseconds of latency. Multi-orbit systems can trade among those priorities instead of forcing every kind of traffic through the same architecture.

GEO is not dead: Viasat and high-throughput satellites

Geostationary satellites still possess one enormous advantage: coverage. A small number of spacecraft can serve vast regions without constant handovers. Viasat’s ViaSat-3 generation shows how far GEO has moved beyond the consumer satellite systems of the 2000s. In 2026, the company brought additional ViaSat-3 spacecraft into service across Asia-Pacific and the Americas, completing a global high-throughput platform. GEO cannot beat LEO on propagation delay, but it remains useful for broad coverage, mobility, government services and markets where a few extremely capable satellites make more economic sense than thousands of smaller ones.

China, Europe and the rise of sovereign constellations

Satellite Internet is also becoming national infrastructure. China is deploying two large Starlink-style systems, Guowang and Qianfan/SpaceSail. Their combined fleets were still only in the hundreds by mid-2026, far behind Starlink, but the programs sit inside a broader push to build domestic launch capacity, satellite factories and communications networks. SpaceSail has also pursued international agreements, including plans for remote connectivity in Brazil.

Europe is taking a different route with IRIS², a secure multi-orbit system intended for governments, critical infrastructure and commercial users. In August 2026, the European Commission said the reinforced architecture would include 348 satellites — 330 in higher LEO and 18 in MEO — with first launches brought forward to 2029. It is not really a consumer clone of Starlink. Its purpose is to ensure that Europe has communications capacity it can control politically and operationally.

Russia, meanwhile, is pursuing a smaller LEO constellation called Rassvet. Reuters reported that the first 16 satellites launched in March 2026, with a long-term goal of roughly 900. Taken together, these projects show how quickly the framing has changed. Orbital connectivity is starting to be treated like fiber backbones, undersea cables or navigation systems: infrastructure that states may not want to depend on a single foreign provider to supply.

There is no single model for the future of satellite internet. Starlink, Amazon Leo, OneWeb, Lightspeed and IRIS² are developing large LEO networks, while systems such as O3b mPOWER use higher MEO orbits and traditional operators continue to improve GEO satellites.

The biggest shift may not be the dish. It may be the phone in your pocket.

Traditional satellite phones made their compromises obvious: special hardware, prominent antennas and expensive service. Direct-to-device networks are trying to make the distinction disappear. The engineering goal sounds deceptively simple — make a satellite look enough like a cellular base station that an ordinary phone can connect without a dedicated satellite antenna.

Starlink’s Direct to Cell satellites already support commercial service through mobile partners. In the United States, T-Mobile’s T-Satellite service has expanded beyond emergency-style messaging toward selected data applications on ordinary recent smartphones. The satellite transmits in cellular spectrum, so from the phone’s point of view the “tower” just happens to be hundreds of kilometers overhead.

AST SpaceMobile is approaching the same problem with a different spacecraft design. Its BlueBird satellites carry extremely large phased arrays intended to create cellular coverage from orbit, and the company says its next generation is designed for 4G/5G broadband directly to standard phones. Several next-generation BlueBirds were launched through 2026, with more planned as AST tries to move from intermittent passes to continuous coverage.

Amazon’s Globalstar deal points toward the same convergence. Apple has already normalized the idea that a phone can fall back to a satellite for messaging. The industry now wants to push further — from emergency contact to routine text, messaging apps, voice and eventually useful data wherever there is open sky.

Terrestrial towers are not about to become obsolete. A satellite has to spread finite radio capacity across a huge footprint, while a city can reuse the same spectrum again and again across thousands of small cells. Direct-to-device is therefore most compelling where towers do not exist at all: mountains, deserts, oceans, disaster zones, rural roads and the dead zones between networks. Its breakthrough is not “better 5G from space.” It is the possibility that far fewer places will leave a normal phone completely disconnected.

The next step in satellite connectivity may remove the satellite dish entirely. Direct-to-device systems are being designed to let ordinary smartphones connect to satellites when terrestrial mobile coverage disappears.

Lasers are turning constellations into networks in space

The simplest broadband satellite acts as a “bent pipe”: traffic goes up from the user and is quickly sent back down to a nearby gateway. That design works only when a suitable gateway is visible and connected to the wider Internet. Optical inter-satellite links change the geometry. A packet can stay in space for hundreds or thousands of kilometers, passing through several satellites before it finally descends.

Laser links can carry large amounts of data without using scarce radio spectrum, and their narrow beams are less prone to accidental interference. They also make routing much harder. Satellites move, lines of sight appear and disappear, each optical terminal can track only a limited set of neighbors, and the best path may change within minutes. Networking researchers increasingly treat LEO constellations as a distinctive routing problem: a data network whose topology is predictable, but never stationary.

That is where satellite Internet begins to overlap with 6G and so-called non-terrestrial networks. The long-term vision is not a separate “space Internet” that people consciously select. It is a network in which phones, vehicles, ships, aircraft, sensors and terrestrial base stations can move traffic between ground and orbit according to availability, price and performance.

Can satellite Internet replace fiber? Mostly no — and it does not need to.

A good way to see the limits of satellite broadband is to ask where the next customer lives. In a city apartment block, one fiber build can serve hundreds of homes with enormous aggregate capacity for decades. A satellite beam covering the same neighborhood has to divide finite radio capacity among its users, with every additional bit competing for spectrum and power.

Move the same customer to a farm 40 kilometers from the nearest fiber route and the economics reverse. Digging a trench for one house may make no sense at all. A LEO constellation already passing overhead can add that user with little local infrastructure beyond the terminal. The same calculation favors satellites on ships and aircraft, at remote construction sites, in polar regions and in temporary emergency networks.

The strongest future is therefore complementary. Fiber handles dense and predictable demand. Cellular networks serve mobile users where population density justifies towers. Satellites fill the gaps, provide backup routes, connect moving platforms and restore service where terrestrial infrastructure is too expensive, damaged or unavailable. Space matters most where geography has traditionally been the limiting factor.

The hard limits: physics, spectrum, weather and congestion

Coverage and capacity are easy to confuse. A satellite may be able to see an enormous area, but it cannot give every user beneath it unlimited bandwidth. Spectrum is finite, each beam has a capacity budget, and customers in the same cell share it. As adoption grows, operators have to add satellites, acquire or reuse spectrum more efficiently, improve beamforming, expand gateways and route traffic more intelligently. That is why systems such as Starlink V3 emphasize capacity density rather than only headline speed.

Weather adds another constraint, particularly at Ku- and Ka-band frequencies where heavy rain can weaken a signal. Modern networks compensate with power control, coding, beam management and alternative paths, but severe conditions can still reduce performance. In everyday use, trees, buildings and terrain are often the more obvious enemy. A LEO terminal needs a sufficiently open view of the sky because its target is moving; one obstruction can cause the same brief dropout every time a satellite passes behind it.

Then there is power. A fiber modem can be very efficient. A phased-array satellite terminal has to steer radio beams electronically and may consume tens of watts or more. That is trivial for a house on mains power and much harder for a backpack, remote sensor or battery-powered field system. Direct-to-device shifts much of that complexity into orbit, but the trade is demanding: the spacecraft must hear and serve tiny phone antennas from hundreds of kilometers away.

Satellite Internet is resilient — but not invulnerable

Physical diversity is one of the strongest arguments for space-based networks. Floods, fires, earthquakes and wars can cut fiber and destroy cell towers, while a constellation remains overhead. Portable terminals can restore a communications link before anyone has rebuilt the local network.

Orbit, however, is not immunity. Satellite links can be jammed. Gateways and terrestrial backhaul can fail. User terminals can be located or attacked. Software, authentication systems and supply chains can be targeted. Operators can also geofence or restrict service because satellite communications still depend on national spectrum rights, licenses and policy decisions.

Recent conflicts have made that distinction unusually visible. Starlink has kept communications working in places where terrestrial networks were damaged, while adversaries have also tried to jam and disrupt it. The useful lesson is narrower than either extreme: a distributed constellation is difficult to disable physically in one blow, but it still depends on radio spectrum, software, ground infrastructure and the decisions of the company or state operating it.

The uncomfortable side of putting tens of thousands of routers in orbit

Megaconstellations solve one infrastructure problem by creating another in orbit. LEO is vast, but the useful shells around Earth are not empty. Active satellites have to avoid one another as well as debris, and large fleets create enormous numbers of predicted close approaches that operators must screen. Modern constellations use automated collision-avoidance systems and often fly low enough that failed spacecraft will eventually reenter, yet modeling studies consistently show that the operational burden rises as the orbital population grows.

The near-term concern is less cinematic than an instant “Kessler syndrome” that suddenly makes space unusable. It is the accumulation of everyday risk: more conjunction alerts, more avoidance maneuvers, more coordination between operators and more spacecraft crossing occupied shells during launch and deorbit. Space traffic management is becoming an infrastructure layer of its own.

Astronomy: the sky is now part of the network

Astronomers were among the first to experience the external cost of megaconstellations directly. Sunlight reflected from satellites leaves streaks in optical images, especially around twilight. Operators have experimented with darker surfaces, visors, different orientations and better orbital data so observatories can plan around passes. Those measures reduce the problem, but they cannot make thousands of bright, moving objects disappear from sensitive observations.

Radio astronomy faces a different form of interference. Measurements with LOFAR and SKA-Low prototype instruments have detected unintended electromagnetic radiation from Starlink electronics inside frequencies used for scientific work. A 2024 Astronomy & Astrophysics paper reported that some second-generation Starlink satellites emitted substantially more strongly than the earlier generation. A broader 2025 survey found Starlink signals across parts of the SKA-Low band often enough to raise concern for experiments trying to detect extraordinarily faint signals from the early Universe.

None of that makes coexistence impossible. It does mean coexistence has to be designed, measured and regulated. Satellite brightness, unintended electronic emissions, orbital choices and coordination with observatories can no longer be treated as side issues after a constellation is already in the sky.

What happens when thousands of satellites burn up?

LEO satellites are not meant to stay in orbit forever. Operators deliberately deorbit them at the end of their lives so atmospheric drag removes dead spacecraft instead of leaving permanent debris. That is good practice for orbital safety, but it transfers part of the environmental question from orbit into the upper atmosphere.

Researchers have already detected metals from spacecraft reentry in stratospheric aerosol particles. A 2024 Geophysical Research Letters study modeled aluminum oxidation during reentry and estimated that a future with very large constellations could inject hundreds of tonnes of aluminum-oxide material into the upper atmosphere each year. Aluminum oxides can participate in chemical reactions relevant to ozone chemistry.

The uncertainty here matters. Researchers are not claiming that satellite constellations have already produced a new ozone hole, and the long-term chemistry, particle lifetimes and climate effects are still being worked out. What has changed is the scale of the question. When thousands of satellites are routinely replaced and burned up, atmospheric effects are no longer a negligible footnote to spacecraft design.

Why Starlink became so difficult to copy

“SpaceX launches a lot of rockets” is true, but it does not fully explain the gap. A broadband constellation needs inexpensive spacecraft, mass manufacturing, terminals, spectrum, launch capacity, gateways, network software, customer support and enough paying users to finance continuous replacement. Weakness in any one of those layers can slow the whole system.

SpaceX built a reinforcing loop: Falcon launches deploy Starlink; Starlink produces recurring revenue and demand for more launches; that scale supports larger rockets and faster satellite production. A competitor can build excellent spacecraft and still be delayed by its launch provider. It can secure launches and still lose on terminal cost. It can reach orbit and still wait for regulatory approval in the markets where customers actually live.

That is why satellite specifications alone tell only part of the story. The product is the entire industrial system — factories, rockets, spectrum, software, gateways and customers. Starlink’s lead is not simply the number of objects above Earth; it is the speed at which SpaceX can build, launch, replace and turn them into a working service.

What the next phase could look like

The next phase of satellite connectivity may be defined by how little users notice it. Today the technology is obvious: there is a Starlink terminal on a roof or a satellite icon on a phone. The industry is working toward a network that chooses between ground and space quietly in the background.

A phone could use terrestrial 5G in town, fall back to a satellite outside coverage and return to the ground network without the user doing anything. An airliner could combine LEO for low latency with another orbital layer for redundancy. A remote business might keep fiber as its primary connection and satellite as an always-available backup. Ships could shift traffic among constellations while crossing oceans. Governments, meanwhile, are likely to demand that at least some of those paths remain under national or regional control.

The satellites themselves are also becoming more like programmable network equipment. Software-defined payloads can reshape beams and reassign spectrum. Optical links can keep traffic in space across long distances. Onboard processing can reduce unnecessary trips through ground gateways. Over time, communications may be combined with navigation, Earth observation and edge computing until the old category of “communications satellite” starts to look too narrow.

What orbit cannot provide is free bandwidth. Every ambitious service still runs into power budgets, spectrum rules, antenna physics and economics. The strongest systems will not be the ones that pretend to replace the entire terrestrial Internet. They will be the ones that use space where space has a genuine structural advantage.

The bigger story is not Starlink. The Internet now has an orbital layer.

The terrestrial Internet was built by putting routers in data centers, fiber under streets and cables across oceans. Satellite broadband adds a new layer: routers moving at orbital velocity, connected by radio and light, able to reach places where physical infrastructure is difficult, expensive or impossible to build.

Starlink showed that LEO broadband can work at consumer scale. Amazon Leo is trying to turn that lead into a competitive market. OneWeb, Telesat and SES are pursuing enterprise and government demand. AST SpaceMobile and Starlink Direct to Cell are turning ordinary phones into potential satellite terminals. Europe and China are treating constellations as strategic infrastructure, while GEO operators are answering with much more capable satellites of their own.

Fiber is not becoming obsolete. The more interesting change is that the Internet is becoming harder to define by where its infrastructure sits. For most of its history, we built it across the surface of Earth. Now part of that network is moving overhead — one fast-moving node at a time.

FAQ

Is Starlink the same thing as satellite Internet?

No. Starlink is one implementation of satellite Internet, not the category itself. Satellite connectivity also includes geostationary systems such as Viasat, medium-Earth-orbit networks such as SES O3b mPOWER, other LEO systems such as OneWeb and Amazon Leo, and direct-to-device networks designed to reach ordinary phones.

Why is Starlink faster than old satellite Internet?

The biggest reason is altitude. Starlink satellites operate far closer to Earth than traditional geostationary spacecraft, so the radio signal has much less distance to travel. The constellation also combines many satellites, steerable beams, extensive ground infrastructure and optical crosslinks. Together, those choices cut latency dramatically compared with classic GEO consumer broadband.

Can satellite Internet be as fast as fiber?

A speed test can sometimes make the two look similar, but the architectures are very different. Fiber generally offers much more capacity in dense areas and more consistent latency. Satellite Internet is most valuable where fiber or good cellular service is absent, too expensive to build, moving with the user, or needed as a backup.

Will normal smartphones connect directly to satellites?

Yes — for limited services in some markets, this is already happening. Starlink/T-Mobile and other systems are expanding from messaging toward data, while AST SpaceMobile is building satellites specifically for broadband to standard phones. The available capacity will still be far below what dense terrestrial cellular networks can provide, especially in cities.

Are satellite megaconstellations dangerous for astronomy?

They can interfere with astronomy in two main ways: reflected sunlight can contaminate optical images, and unintended radio emissions can affect radio observations. Operators and observatories are developing mitigations, but peer-reviewed studies show that both effects are measurable and become more difficult to manage as constellation size grows.

What is the biggest obstacle to global satellite Internet?

There is no single bottleneck. The challenge is the whole system: launch capacity, spectrum, satellite and terminal manufacturing, ground infrastructure, congestion, national regulation, debris avoidance, astronomy impacts and the cost of replacing spacecraft every few years. Reaching orbit is only the beginning; keeping a very large constellation useful, affordable and sustainable is the harder long-term task.