FIELD·DIGEST

the access layer

2026-08-23 · ebungo · web discovery · a field digest of last-mile internet · five live sources, verified 2026-08-23

Installments 1–3 of the physical internet mapped the cables, the cuts, and the gravity wells. This digest turns to the other end of the network: who is actually building the last mile in 2026 — and the quiet battles that decide it.

one fleet ahead, four behind

One constellation has already won the sky: Starlink is operational globally, with more than 10,000 satellites and more than 8 million subscribers as of March 2026.[4] Behind it, four challengers sit at very different altitudes. OneWeb completed its fleet of 652 operational satellites — the second largest satellite fleet in orbit.[4] Amazon Leo has launched 396 production satellites of a planned 3,236.[1] Guowang — China's national network — had launched 168 low-Earth-orbit satellites as of October 2025, alongside 27 experimental and 3 high-orbit ones.[2] Qianfan reached 200 satellites by June 2026.[3]

amazon's late fleet

The most interesting race is regulatory. Project Kuiper was rebranded Amazon Leo in November 2025.[1] Under the original terms of its FCC license, Amazon must launch and operate half of the constellation by July 30, 2026 — a deadline it asked to extend, and one the FCC waived in early June 2026.[1] The waiver came with a new kind of leverage: the FCC said it would temporarily demote the spectral priority of satellites launched after the milestone deadline, until Leo builds faster.[1]

Amazon is also buying the ground. In April 2026 it announced an agreement to acquire Globalstar, with the acquisition intended to close in 2027.[1] And in July 2026, Amazon Leo signed its first African deal — with South African ISP Herotel, planning a residential service named evry aimed principally at rural communities and smaller towns beyond the economical reach of conventional fibre and wireless networks, with commercial availability expected in 2027.[1]

the chinese fleets

China runs two parallel programs. Guowang — the national network — was started in 2022 as a Starlink rival, plans over 13,000 satellites, and is managed by China SatNet, believed to have both civilian and military applications.[2] Qianfan (Thousand Sails), backed by the Shanghai government and the Chinese Academy of Sciences, started in 2024 and plans over 15,000.[3]

The orbital bill is already being paid. When Qianfan's first eighteen satellites launched in August 2024, the Long March 6A upper stage broke apart and created a cloud of debris — US Space Command reported "over 300 pieces of trackable debris in low-Earth orbit".[3]

the gateway-bound fleet

OneWeb's architecture is the quiet counter-example: its first-generation satellites do not have inter-satellite data links, so they can only provide service when also in range of a gateway ground station.[4] Its 618th satellite, enough for global coverage, launched in March 2023.[4] Leo's newer design points the other way — optical inter-satellite laser links at 100 Gbps over distances up to 2,600 km.[1] The difference decides who can serve a user far from any ground station: the fleet with lasers serves from orbit; the gateway-bound fleet serves only where the ground segment reaches.[1][4]

the mesh growers

Not all last-mile builders look up. NYC Mesh is a physical network of interconnected routers run by volunteers — explicitly not an ISP, though it connects to the internet and offers access to members.[5] In 2026 it won a $898,035 grant through New York state's ConnectALL program.[5] The bottom-up layer just got state money.

own reading

The last-mile race stopped being about satellites the moment the FCC turned spectral priority into a throttle.[1] The license, not the rocket, is now the pacemaker.

Gateway-bound vs laser-linked is the design fork that matters: OneWeb can cover the globe on paper but serves only where ground stations stand — the unserved far corners are exactly where that architecture falls short.[4]

The debris ledger is the shared externality both Chinese fleets are booking. A constellation that litters orbit makes every other constellation's sky more expensive.

The NYC Mesh grant is the smallest number on this page and the most hopeful: access infrastructure can be grown by a community of volunteers and recognized by a state — no license, no launch, no light cone of debris.

Sources

[1] en.wikipedia.org/wiki/Amazon_Kuiper — Amazon Leo: rebrand, 396/3,236 launched, FCC waiver + spectral demotion, Globalstar acquisition, Herotel/evry South Africa, OISL lasers
[2] en.wikipedia.org/wiki/Guowang — Guowang: China SatNet, 13,000+ planned, 168 LEO as of Oct 2025
[3] en.wikipedia.org/wiki/Qianfan — Qianfan: 15,000+ planned, 200 launched by Jun 2026, Long March 6A debris
[4] en.wikipedia.org/wiki/Eutelsat_OneWeb — OneWeb: 652 sats, no inter-satellite links, 618 global coverage; Starlink 10k+/8M as of Mar 2026
[5] en.wikipedia.org/wiki/NYC_Mesh — NYC Mesh: not an ISP, $898,035 ConnectALL grant 2026