The Base Station That Crossed the Pacific

At 7:11 a.m. Japan time on August 23, an unmanned aircraft that had spent roughly 13 days crossing the Pacific entered Japanese airspace. It had left New Mexico on August 9, stayed in the stratosphere for more than 15,000 kilometers, and was headed for a test area off Cape Muroto in Kochi Prefecture. Waiting below were not satellite phones or custom terminals. They were ordinary mobile devices tied into SoftBank’s cellular network.

A cell site arrived from another continent

The platform came from Sceye, the New Mexico aerospace and materials-science company in which SoftBank invested in 2025. Sceye builds a lighter-than-air, or LTA, high-altitude platform: a helium-lifted airship designed to operate in the stratosphere. SoftBank’s technical page lists the Sceye vehicle at 65 meters long and describes the intended endurance in terms of months, although the Muroto mission itself did not demonstrate months of station-keeping over Japan.

After reaching Japan, the aircraft operated around Muroto City, off Cape Muroto. SoftBank says operators accounted for stratospheric wind direction and speed and kept the platform within a minimum five-kilometer radius of its designated area for an extended period. That figure is easy to misread: it is a station-keeping measure, not the radius of the communications footprint.

The aircraft then began another Pacific crossing back toward the United States. As of SoftBank’s September 2 announcement, that return flight was still underway. The mission therefore tested more than a radio link. It combined intercontinental repositioning, arrival in Japanese airspace, target-area station keeping, cellular service, and return operations in a single program.

What actually worked on the ground

The airborne payload acted as a 4G LTE-equivalent base station. A ground gateway linked it to SoftBank’s core network. In the Muroto area, SoftBank and Sceye report stable two-way voice calls, text messaging, social-media applications, video calls and video streaming between the HAPS and smartphones on the ground.

The disaster-response pieces were more concrete than a generic “connectivity” demonstration. The companies tested the emergency alert messaging system used for earthquake early warnings and tsunami warnings. With cooperation from the Japan Coast Guard, they also confirmed a voice call to 118, Japan’s maritime emergency number. SoftBank says the trial achieved terrestrial-network-equivalent communications performance while reducing interference with ground base stations.

That is substantial evidence of integration, but not a commercial-load test. The release does not disclose how many handsets were active simultaneously, the exact commercial capacity per platform, service pricing, fleet size or a service-level guarantee. Nor does a limited trial answer how the network behaves when thousands of users try to connect at once after a real disaster.

Why HAPS is not just “another satellite”

High-altitude platform stations occupy an unusual layer of communications infrastructure. The ITU describes HAPS as platforms operating high in the atmosphere, typically around the stratospheric band used for persistent communications. SoftBank generally frames its systems around roughly 20 kilometers altitude—far above conventional aviation and far below low-Earth-orbit satellites.

That physical position changes the network design. A HAPS can potentially use terrestrial mobile spectrum and communicate directly with ordinary handsets, while covering a much larger area than a ground tower. It is also much closer than a satellite, which can improve latency and link efficiency. SoftBank says its HAPS architecture can be designed for coverage on the order of up to roughly 200 kilometers in diameter, though the Muroto trial did not demonstrate that maximum commercial footprint.

The network still needs a route back to the operator. A service link connects the HAPS to phones and other devices. A feeder link connects the airborne platform to a gateway and the core network. During the ordinary communications test in Kochi, the ground gateway was a critical part of that chain.

That creates a disaster-engineering question. A base station can survive in the sky while its gateway, power supply or terrestrial backhaul fails below. A resilient commercial design may therefore need portable gateways, multiple gateway sites, satellite backhaul or other redundant paths. HAPS moves a major piece of infrastructure away from the disaster zone; it does not magically eliminate every terrestrial dependency.

The 68-millisecond experiment may matter as much as the phone call

The trial’s most forward-looking result involved computing, not coverage. SoftBank and Sceye installed a mobile core and a web server aboard the HAPS and processed a smartphone request entirely on the airborne platform rather than sending it through a terrestrial network to a distant cloud.

SoftBank measured an average round-trip processing time of 68 milliseconds and said latency was more than 40 percent lower than in its comparison path using cloud processing over the internet. The company calls the architecture a world first based on its review of public information.

The point is larger than shaving milliseconds from a web request. Remote-controlled drones, live video analytics, autonomous machines and what SoftBank calls physical AI increasingly depend on decisions being made close to where data is generated. If the communications node and the computing node can occupy the same stratospheric platform, an operator could theoretically bring both connectivity and local processing to a disaster zone, offshore area or remote region without first constructing a full ground edge-computing site.

From phones on the ground to drones in the air

SoftBank also used the HAPS link to control an automated drone flight. The test carried flight-control traffic, position information and video. The company points to disaster reconnaissance, delivery of supplies, offshore monitoring, mountain logistics, power-line and road inspection, and forest surveillance as possible applications.

That fits SoftBank’s broader idea of a “three-dimensional communications network.” Cellular networks were built mainly for people, buildings and vehicles on the surface. A 6G-era network may have to serve drones and other aerial mobility at the same time. HAPS becomes one layer between terrestrial radio access and satellites, part of the wider non-terrestrial-network, or NTN, architecture.

The path to Muroto started with disaster communications

SoftBank’s HAPS program is often presented as a futuristic 6G project, but its Japanese history is rooted in something more immediate: what happens when ground networks fail. Company accounts of the program trace part of the motivation to network-restoration work after the 2011 Great East Japan Earthquake.

SoftBank began experimenting with tethered-balloon relay systems in 2012 and published results in 2013, including a configuration that could use satellite backhaul if the normal source base station was unavailable. The company later developed drone relays and higher tethered-balloon systems, gradually accumulating the antenna-control, rapid-deployment and disaster-network experience that fed into HAPS.

In 2017 SoftBank began dedicated HAPS aircraft development and created HAPSMobile later that year. In September 2020 its fixed-wing, solar-powered Sunglider reached roughly 19 kilometers over New Mexico and spent 5 hours 38 minutes in the stratosphere. HAPSMobile and Alphabet’s Loon also demonstrated LTE communications during the flight. In 2021 SoftBank acquired roughly 200 HAPS-related patents from Loon after that venture shut down.

In 2023, a HAPS test aircraft carrying SoftBank’s 5G payload operated over Rwanda, providing 5G communications for about 73 minutes at an altitude reaching 16.9 kilometers. That same period also brought major regulatory progress: the 2023 World Radiocommunication Conference expanded international options for using HAPS as IMT base stations, while SoftBank continued work on interference suppression, beamforming and spectrum sharing with terrestrial networks.

Why Sceye, when SoftBank already had Sunglider?

The partnership is not simply a replacement of one aircraft by another. SoftBank has spent years on Sunglider, a heavier-than-air, fixed-wing HAPS that stays aloft through aerodynamic lift and solar power. Sceye’s platform is lighter than air, using helium buoyancy and an airship form factor. The two approaches solve the endurance and payload problem differently.

SoftBank’s current HAPS technical page lists Sceye’s airship at 65 meters long and Sunglider at a 78-meter wingspan. The company describes both concepts as targeting endurance measured in months. What Sceye offered in the near term was a platform with a growing record of long-duration stratospheric operations. In a 2026 endurance mission before the Japan flight, Sceye reported more than 12 days and 6,400 miles of stratospheric travel, including more than 88 hours holding over an area of operation.

SoftBank’s 2025 investment in Sceye came with an agreement giving it exclusive rights for HAPS services in Japan using Sceye’s platform. The dual-track strategy lets SoftBank continue developing its own fixed-wing technologies while using an LTA system to pull the commercial timetable forward.

Spectrum is as important as lift

A HAPS cannot simply transmit on whatever frequency is convenient. Wide-area radio from the sky can interfere with terrestrial cellular networks using the same spectrum. SoftBank has therefore spent years on cylindrical phased-array antennas, beamforming and nullforming—techniques that steer energy toward intended users and suppress it toward protected ground cells.

WRC-23 was a turning point in the international framework. SoftBank’s Japanese release on the conference said additional bands in the 700–900 MHz range, 1.7 GHz range and 2.5 GHz range were formally added to the HAPS-as-IMT-base-station framework, increasing the options available to national regulators. The ITU likewise presents HAPS/HIBS as a tool for rural broadband and disaster recovery with relatively little ground infrastructure.

But global spectrum rules do not equal a Japanese commercial license. Aviation rules, radio authorization, interference management, cross-border flight planning, cybersecurity, maintenance and fleet operations all have to work together. HAPS is a telecom system that must also behave like an aviation system.

Does HAPS compete with Starlink?

The rise of low-Earth-orbit broadband makes the question unavoidable. LEO satellites can cover huge areas and cross oceans without station-keeping over every region. They are well suited to global reach. HAPS, by contrast, can concentrate capacity over a chosen area, operate physically much closer to users, and potentially integrate with an operator’s existing licensed mobile spectrum and handsets.

The tradeoff is persistence and operational complexity. An aircraft has to be launched, positioned, maintained, replaced and routed around atmospheric conditions and airspace rules. A single HAPS covers only a region, not the globe. Building continuous national coverage would require fleet orchestration, ground support and economics that have not yet been demonstrated publicly at scale.

The more useful model is not HAPS versus satellite. It is a layered network in which terrestrial base stations carry dense everyday traffic, HAPS adds targeted wide-area capacity and resilience, and LEO or GEO satellites provide broader backhaul and reach. The business competition will be over which layer handles which traffic most efficiently.

Commercialization is now an operations problem

SoftBank and Sceye say they will use data from the mission to improve operating methods and communications quality with the aim of commercializing HAPS in Japan from 2027 onward. That wording matters. The announcement does not give a launch month, retail price, initial coverage map, number of aircraft, simultaneous-user capacity or guaranteed availability.

Even so, Muroto marks a line in the program’s history. The 2020 Sunglider flight showed that an autonomous fixed-wing HAPS could reach the stratosphere and deliver LTE. The 2023 Rwanda demonstration showed a 5G payload operating from the stratosphere. The 2026 Japan mission joined flight operations, SoftBank’s core network, ordinary handsets, emergency functions, drones and onboard computing into one service-test scenario.

The next questions are less cinematic and more important. Can a fleet operate through Japan’s seasons? Can it restore communications when the terrestrial disaster is real rather than simulated? Can operators rotate aircraft without users noticing? Can the economics beat the cost of extending ground networks or buying satellite capacity?

For years, “flying base station” sounded like a metaphor for a distant network future. Off Cape Muroto, it became something closer to an operating model. The hard part now is not getting a base station into the stratosphere. It is making connectivity from the stratosphere ordinary.