The mega-constellation thirst trap

A mega-constellation is what you call a satellite network too large to keep operating with traditional ground-based mission control. The threshold is fuzzy — most analysts pin it at five hundred operational spacecraft — but the cultural meaning is clear. A mega-constellation is the kind of system where you stop thinking about individual satellites and start thinking about the network as a single distributed machine. There are precisely two of those running in 2026, and one of them isn't yet generating positive cash flow.

Starlink, with roughly 7,200 operational satellites at the time of writing, is the only fully operational, profitable, growing mega-constellation in history. Project Kuiper, with about 600 satellites in orbit and several thousand more in production, is operational and serving customers but burning Amazon money on a scale that would worry any standalone business. Everything else announced in the past decade is either much smaller, indefinitely paused, or in the regulatory holding pattern where most ambitious filings go to die.

What's actually flying versus what was promised

The full list of formally announced LEO mega-constellations runs to about eighteen entries. The discrepancy between announcement and reality is the entire story of this market.

Starlink (SpaceX). Announced 12,000 satellites; later filed for an additional 30,000. Currently operational at ~7,200. Generating an estimated $9-10 billion in annualised revenue as of late 2025. The reference case for the whole sector.

Project Kuiper (Amazon). Announced 3,236 satellites in 2019. Currently has ~600 in orbit. Targeting first commercial service in late 2025. Amazon has not disclosed how much it's spent so far; analysts estimate north of $10 billion in capital commitments through 2027. Pricing strategy not yet visible.

OneWeb. Announced ~650 satellites; operational at 634. The company went bankrupt in 2020, was rescued by the British government and Bharti Global, and merged with Eutelsat in 2023. Now flying as Eutelsat OneWeb. Profitable revenue, but at a much smaller scale than Starlink and with a much more conservative satellite-replacement cadence.

SpaceMobile (AST SpaceMobile). Announced 168 BlueBird satellites for direct-to-cell service. Five operational. Has signed agreements with AT&T, Verizon, Vodafone and Rakuten. The technology is genuinely novel; the deployment timeline keeps slipping right.

Lynk Global. Announced ~5,000 satellites for direct-to-phone. Has 5 operational. Has commercial roaming partnerships in 50+ countries but a tiny revenue base. Listed via SPAC merger in early 2024.

Chinese state-backed constellations. Guowang (G60 Qianfan) is targeting roughly 14,000 satellites and has launched approximately 90. SpaceSail (formerly Honghu) targeting 1,296 satellites with around 100 launched. These are the most credible non-Western mega-constellations in active deployment.

The pause-and-quietly-disappear list. O3b mPOWER (operational at the much smaller MEO scale, ~9 satellites). Telesat Lightspeed (announced 1,671 satellites in 2017, downsized to 198 in 2023, first launches now scheduled for 2026). Boeing's V-band constellation (announced 147 satellites in 2017, never launched). LeoSat (announced 108, declared bankruptcy 2019). LaserLight (announced 12, never launched). SpinLaunch SuperBatch (announced ~250, indefinitely paused). RBC Signals' Chorus, GHGSat (a real EO operator, not a mega), and roughly half a dozen others that have either downsized to single-digit demonstrator missions or quietly dropped the announcement off their corporate websites.

Of the ~85,000 satellites collectively announced for LEO mega-constellations since 2017, approximately 8,500 are operational. About 90 percent of the announced capacity has not been built.

Why the math doesn't close

A mega-constellation is a strange financial animal. The capex is enormous and almost entirely front-loaded — you have to build the manufacturing line, get the first ~500 satellites into orbit, and stand up the ground network before you can sign your first paying customer. The revenue then arrives slowly as you onboard subscribers. The window from "first dollar of capex" to "positive operating cash flow" is at least four years for the most aggressive operators and seven to ten for the rest. During that window the operator is bleeding cash at an industrial scale and is entirely dependent on continuing capital raises. SpaceX could survive that window because Falcon 9 was already a profitable launch business funding the constellation; Amazon can survive it because Amazon. Nobody else has either luxury.

The other reason the math is brutal is unit economics on the customer side. Starlink's residential service is roughly $80 a month in the US. To recover the capex on a 12,000-satellite constellation at that price you need on the order of fifteen million paying subscribers. Starlink is at approximately five million as of late 2025 and growing fast — and that's the success case. The constellations chasing enterprise and government customers can charge much more per seat but the addressable market is narrower. Telesat Lightspeed's pivot from broadband-for-everyone to government-and-enterprise-only is the canonical example of a constellation rationalising its way down to a payload-sized business.

Direct-to-cell adds yet another layer of difficulty. AST SpaceMobile and Lynk both have to convince mobile network operators to share spectrum that the operators paid billions for at auction; the spectrum-sharing economics still aren't worked out at any meaningful scale, and the regulatory approvals are coming in country by country. The path to mass-market revenue runs through dozens of national regulators. The first $10 of revenue is a long way from the first $1 billion.

The regulatory cliff edge

Most of the announced mega-constellations exist on paper because the FCC and ITU process makes it cheap to file and slow to revoke. A US operator can file an application for a 5,000-satellite constellation for a few hundred thousand dollars in legal fees, get an FCC authorisation contingent on launching half within six years, and then sit on the licence indefinitely while the announcement does its real job — generating press coverage, attracting partners, and propping up a valuation.

The FCC has, since 2024, started enforcing the milestone deadlines more aggressively. Several authorisations have been formally cancelled. The trend continues; over the next two years another ten or fifteen of the announced mega-constellations will probably have their authorisations rescinded for failure to meet deployment milestones. Most of those companies will not announce that this has happened.

The ITU adds an additional layer. There are only so many available LEO orbital shells; spectrum allocations in some bands are now first-come, first-served on a global basis. Amazon, OneWeb and the Chinese constellations have aggressive ITU filings designed to lock in spectrum that subsequent applicants can't access. This is not a market that newcomers can easily enter even if they raise the capital.

What the survivors look like in 2030

Best estimate: by 2030 there will be three operational LEO mega-constellations larger than 1,000 satellites — Starlink, Kuiper, and the Chinese state combine. There will be perhaps two or three smaller direct-to-cell networks (AST SpaceMobile and Lynk both possible), and a small number of single-purpose narrowband or IoT constellations like Iridium NEXT. The remaining announced systems will either have downsized to a few hundred satellites and a niche customer base, or have been quietly cancelled.

Starlink, by then, will have somewhere between 12,000 and 18,000 operational satellites, will be doing $25-30 billion in annualised revenue, and will be the largest commercial space business by some distance. It will also probably be a separately listed company, given the IPO momentum.

The smaller players in this list aren't bad businesses. They're just not mega-constellations. Iridium has been profitable for a decade with 75 satellites. Globalstar has been profitable with 24. The successful long-tail constellation businesses look much more like Iridium than like Starlink, and that's both fine and what the market actually needs. Calling them mega-constellations was always more of a fundraising tactic than an engineering description.

The thirst trap is the gap between the two. Five hundred satellites isn't enough to be Starlink. Fifty isn't enough to be a useful niche player. The companies that announced themselves into the middle of that gap — fifty to five hundred satellites, broadband-class ambitions, no anchor customer — are the ones whose math never closed and whose announcements you'll see disappear over the next five years.

Constellation counts as of April 2026 from public sources and the New Space Tracker satellite-communications directory.