At about 20 kilometers up, above the cruising altitude of commercial jets and below the orbit of imaging satellites, Sceye wants to park a fleet of solar-powered airships over Japan.
The high-altitude platform company signed a five-year master services agreement with SoftBank Corp. on Nov. 3. Under the terms, Sceye will build, launch and operate vehicles at roughly 20 kilometers and supply wide-area coverage across the country. Both companies describe high-altitude platform stations, or HAPS, as a supplement to ground networks rather than a replacement, with remote regions and disaster response as the first targets.
The structure puts the capital risk on Sceye and the customer relationship on SoftBank. Sceye owns the aircraft, the launch operations and the ground control; SoftBank buys coverage as a service. For an operator that has spent decades building towers it must maintain, power and repair, paying for capacity that arrives from above is a different kind of contract, and a simpler one to abandon at renewal.
The vehicles sit closer to airships than to drones. A fabric hull filled with a lifting gas carries a payload of communications equipment, and solar panels on the surface charge batteries that keep the craft aloft through the night. Station-keeping is the hard part. The air at that altitude is thin enough that control surfaces accomplish little, so the vehicle maneuvers by climbing or descending to find a favorable current, or by running propellers when the wind will not cooperate.
The appeal is geometry. A tower serves whatever its antenna can see, which in mountain valleys or across water is often very little. A platform at 20 kilometers has a line of sight to a far larger footprint, and one vehicle can stand in for a scattering of base stations that would each need a road, a power connection and a maintenance visit.
That arithmetic matters most in a country like Japan, where a substantial share of the population lives in coastal towns and mountain villages that shrink every year while the cost of keeping their networks running does not. Operators have been closing marginal sites for a decade. A platform that can be repositioned where demand appears, rather than where the fiber already runs, changes the calculation for the least profitable parts of the map.
Disaster response is the second use case, and in Japan it is not hypothetical. Earthquakes and typhoons take out transmission lines, cut power and isolate communities for days. A vehicle that can be flown to the affected area and switched on does not need a road, a generator or a permit to dig.
The idea has a long and mostly unhappy history. Google’s Loon spent years flying balloons over rural regions and shut down in 2021 after failing to bring costs down against terrestrial alternatives. Several smaller ventures folded. The problem was never that the technology could not fly; it was that keeping a fleet aloft, replacing failed units and selling capacity to operators who already owned spectrum produced a cost per subscriber that no carrier would pay.
What has changed is the shape of the contract. Loon tried to build a global business on wholesale capacity and advertising partners. A five-year master services agreement with a single incumbent operator, priced against the cost of a rural tower, is a narrower proposition and an easier one to underwrite. Sceye also flies fewer, larger and longer-lived vehicles than the balloon approach allowed, which shifts spending from launch operations toward manufacturing.
Satellites are the other comparison, and the one investors reach for first. Low-earth-orbit constellations have driven the cost of delivering a megabit to unserved ground down sharply over the past five years, and they have the advantage of global reach. But capacity in orbit is shared across a footprint hundreds of kilometers wide, and the satellites pass overhead rather than hovering. A platform that stays over one region can concentrate capacity where the customers are, and its round-trip delay is measured in milliseconds, which matters for voice and for anything interactive.
None of this happens without spectrum. A HAPS payload uses licensed mobile frequencies, which means coordination with the operator’s existing allocations, with neighboring countries whose networks share the band, and with aviation authorities that must clear the airspace and the flight profiles. Those processes are slow, and the agreement does not remove them.
The engineering risks are unglamorous. A vehicle that loses altitude control comes down somewhere, and the recovery and replacement costs sit with the operator of the fleet. High winds at altitude can force a descent or an early landing. Power budgets are tight: every kilogram of payload has to be paid for in solar cells and battery mass, and the telecom equipment competes with the flight systems for both.
SoftBank has not said how many vehicles the first phase covers, what it will pay, or when commercial service begins. The agreement sets a framework; the deployments will be announced one at a time.
For now the bet is narrow and testable. If a single vehicle can hold station over a stretch of coastline for months at a time and carry enough traffic to keep a village connected, the case for the next twenty is arithmetic rather than vision. If it cannot, the contract will be remembered as another experiment conducted at 20 kilometers, where the view is clear and the customers are far away.
