SoftBank Tests Stratospheric Edge Computing as HAPS Platform Processes Smartphone Traffic at 68ms
By Tech Desk
SoftBank and US aerospace company Sceye just wrapped up a trial that puts edge computing in a place most engineers wouldn't expect: 16.5 kilometers up, aboard a lighter-than-air platform drifting through the stratosphere.
The test wasn't just about connecting smartphones to a base station in the sky. SoftBank installed mobile core network functions and a web server directly on the High Altitude Platform Station (HAPS), processing requests in the stratosphere rather than routing them through ground networks to distant cloud services. The result? An average round-trip response time of 68 milliseconds — more than 40 percent faster than the internet-based cloud service SoftBank used for comparison.
What Happens When You Move the Edge to 16 Kilometers
The trial ran from August 9 to late August 2026, with Sceye's platform traveling more than 15,000km across the Pacific before entering Japanese airspace on August 23. It operated around Muroto City in Kochi Prefecture, maintaining position within a 5km radius during part of the operation.
SoftBank tested voice calls, text messaging, video calls, and video streaming between the HAPS and smartphones on the ground. The companies also ran earthquake and tsunami warning messages and completed a call to Japan's 118 maritime emergency number with the Japan Coast Guard. A 4G LTE-equivalent base station aboard the platform connected to a ground gateway, which linked to SoftBank's mobile core network.
But the headline isn't the connectivity. It's where the processing happened. By running mobile core functions and a web server onboard, SoftBank showed that edge computing doesn't need to be tethered to ground-based infrastructure. According to the company, this was the first time both a mobile core network and web server operated entirely aboard a HAPS, processing smartphone communications without relying on terrestrial routing.
Why Stratospheric Processing Matters
Cloud providers have spent years building regional edge nodes to cut latency, but physical distance between cell towers and data centers remains a bottleneck. Latency-sensitive applications — autonomous vehicles, remote surgery, real-time industrial control — can't tolerate the round-trip delays that come with traditional cloud architecture.
SoftBank's approach flips the problem. Instead of pushing cloud infrastructure closer to users through ground-based edge nodes, the company placed the compute at altitude. The 68ms latency figure matters because it demonstrates that stratospheric platforms can deliver performance comparable to terrestrial edge deployments, with the added benefit of coverage over terrain that's difficult or expensive to wire.
The trial also extended connectivity to a drone operating in the test area, using the HAPS connection for remote flight control, location-data transmission, and video from the aircraft. SoftBank has separately tested spectrum-sharing techniques to reduce radio interference between aerial and terrestrial base stations, though the company didn't confirm whether those methods were used in the latest trial.
From 2017 Development to 2027 Commercial Launch
SoftBank has been working on HAPS technology since 2017, covering aircraft platforms, telecommunications payloads, antennas, batteries, motors, and related network technologies. In 2025, the company tested a 5G HAPS radio architecture using a light aircraft at about 3,000 metres, achieving average downlink speeds of roughly 33Mbps.
The operator expanded its HAPS program in June 2025 by investing in Sceye and securing exclusive rights to provide services in Japan using the company's lighter-than-air platform. SoftBank said it planned to begin pre-commercial HAPS services in 2026, targeting communications recovery during large-scale disasters and connectivity for mountainous regions and remote islands.
SoftBank and Sceye plan to use data from the trial to refine operating procedures and communications performance ahead of commercial services starting in 2027. The International Telecommunication Union's World Radiocommunication Conference in 2023 identified spectrum in the 2GHz and 2.6GHz bands for HAPS operating as IMT base stations, providing a regulatory framework for commercial deployment.
SoftBank's wider program includes work with Japan's National Institute of Information and Communications Technology, ArkEdge Space, and Kiyohara Optics on optical links between HAPS and low-Earth orbit satellites. The organizations plan to attempt bidirectional optical communications between a HAPS and a LEO satellite in 2027, with equipment designed for bidirectional transmission at 10Gbps. NICT said the planned HAPS-to-LEO test could cover distances up to approximately 2,000km.
Technical Details Behind the Latency Improvement
The 68ms latency figure represents a substantial improvement over traditional cloud routing. When SoftBank compared processing on the HAPS versus routing through its selected internet-based cloud service, the stratospheric approach eliminated multiple network hops and reduced the physical distance that packets needed to travel.
According to networking principles, each hop in a terrestrial network adds processing delay, queuing delay, and propagation delay. By placing compute functions closer to the radio access layer — in this case, aboard the HAPS itself — SoftBank reduced the need for backhaul to centralized facilities. This approach aligns with multi-access edge computing (MEC) principles standardized by ETSI and 3GPP, which advocate for compute resources at the network edge to reduce latency for IoT, AR/VR, and industrial applications.
The trial also provided data on equipment performance in the stratosphere. Operating at approximately 16.5km altitude exposes hardware to extreme temperatures, low atmospheric pressure, and increased radiation exposure. SoftBank did not disclose specific thermal management or radiation hardening techniques used in the trial, but the successful operation of both radio equipment and compute servers suggests adequate environmental controls.
For telecom engineers, the results validate concepts discussed in recent whitepapers about non-terrestrial networks (NTN) as extensions of the terrestrial edge. The 3GPP Release 17 and 18 specifications already include provisions for integrating satellite and HAPS platforms into 5G networks, and trials like this one provide real-world performance data to inform future standards development.
What This Means for Distributed Computing
For practitioners building cloud and edge infrastructure, the trial offers a proof point that edge processing can work well beyond traditional ground-based architecture. Whether stratospheric edge nodes become standard telecom infrastructure or remain specialized for disaster response and remote connectivity will depend on cost, regulatory approval, and competition from satellite-based backhaul.
The 68ms latency figure is the number to watch. If SoftBank can maintain that performance at scale — and if Sceye's platforms can stay aloft for weeks or months at a time — stratospheric edge computing stops being a curiosity and starts being a viable alternative to fiber and terrestrial wireless for certain use cases.
SoftBank President and CEO Junichi Miyakawa said the company aims to build "next-generation communications infrastructure that seamlessly connects the ground, the sky and space." The Japan trial suggests that vision isn't just marketing. It's becoming testable reality.
For more on Cloud & Edge Computing, read our latest coverage on distributed infrastructure and edge deployment strategies.
For deeper reading on HAPS technology and stratospheric platforms, see the GSMA whitepaper on non-terrestrial networks which details how platforms like Sceye's integrate with 5G core networks.