SoftBank Demonstrates Stratospheric Edge Computing to Cut Round-Trip Latency by 40%

During field trials over Japan, SoftBank and US aerospace firm Sceye validated direct-to-device mobile connectivity alongside onboard edge processing using a High Altitude Platform Station (HAPS). Operating an uncrewed lighter-than-air platform in the stratosphere at an altitude of approximately 20 kilometers, SoftBank deployed mobile core network functions and a web server directly onto the airborne platform. The setup handled smartphone communications and automated drone control without backhauling requests to centralized ground-based cloud infrastructure. SoftBank reported an average round-trip latency of 68 milliseconds during the test—more than 40% lower than routing traffic through conventional internet cloud services. Deploying autonomous compute and core telecom functions directly in the stratosphere challenges the traditional boundary between connectivity infrastructure and edge processing. In typical remote, maritime, or disaster-recovery scenarios, edge devices either suffer from the latency overhead of multi-hop backhaul to regional datacenters or fail entirely when terrestrial fiber is severed. Placing computing power directly on a wide-area airborne platform—covering up to 200 kilometers in diameter—enables deterministic, low-latency execution for critical workloads such as drone fleet coordination, emergency response systems, and local telemetry filtering independent of ground network availability. This milestone marks a convergence of non-terrestrial networks (NTN), AI-RAN, and distributed edge computing. While Low Earth Orbit (LEO) satellite constellations have expanded ubiquitous global coverage, they introduce propagation delays and variable routing topologies when processing data across distant ground stations. HAPS positions infrastructure within the stratosphere, bridging the gap between terrestrial micro-edge datacenters and orbital satellites. As telecom operators prepare for 6G three-dimensional networking architectures, hyperscalers and carriers are racing to push containerized microservices and lightweight inference models into increasingly distributed, physical compute form factors. For infrastructure engineers and edge system designers, non-terrestrial edge computing opens new deployment paradigms for mission-critical operations. Distributed systems deployed in remote or disconnected environments can begin adopting hybrid topologies where stratospheric nodes act as intermediate processing gateways and local fallback clusters. However, engineering teams must account for stringent size, weight, and power (SWaP) constraints on airborne payloads, localized storage capacity limitations, and intermittent uplink synchronization when the platform eventually syncs back with core data centers. Practitioners operating in industrial IoT, logistics, and disaster management should track HAPS-based edge integration as carriers prepare commercial offerings for 2027.