Product1 publisher2 min readPublished
Sceye's stratospheric cell served unmodified handsets over SoftBank's core network
The ST1 flight put a mobile core and a web server in the stratosphere and reported 68 milliseconds of round-trip processing onboard. Sceye has not published how many consecutive days one airframe holds station.
The Product Desk · Product desk
What happened
- Sceye launched its Service Test 1 platform from New Mexico on August 9, 2026, and reached Japan 13 days later on the company's first flight to Asia, run in partnership with SoftBank Corp.
- The platform covered nearly 30,000 km, re-entered the continental United States on September 4 and reached airspace near Moriarty on September 5, where Sceye ran further tests and then terminated the flight.
- Over Japan, Sceye and SoftBank served mobile broadband to unmodified devices through SoftBank's core network, and also tested emergency alert messaging and communication with drones.
- A mobile core network and a web server ran onboard the platform, which Sceye and SoftBank describe as the first test of that combination in the stratosphere.
- The payload was SceyeCELL, which Sceye calls a cell tower in the sky built to deliver wide-area mobile broadband directly to standard handsets.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- decision Because the handsets needed no change, the people who now have to judge this are the carrier's core network and spectrum staff, none of whom have ever signed off on an airframe.
- capability With the packet core and the web server both flying, response processing for the tested workload no longer depends on a path through the terrestrial internet.
- exposure Buying coverage in units of one HAPS puts the area Sceye equates to about 500 towers behind a single airframe, so one vehicle failure exposes the whole footprint at once.
Nothing had to be installed on the phone. Sceye says the standard devices it served over Japan reached SoftBank's core network unmodified [3]. For a carrier that removes the two slowest parts of standing up a new access layer, device certification and a firmware campaign. What is left belongs to the core network team and to whoever holds the spectrum grant.
Work the 68 milliseconds backwards. Sceye and SoftBank report an average round-trip processing response of 68 ms onboard the platform and a latency cut of more than 40 percent against internet-based cloud processing [4]. Divide 68 by 0.6 and the baseline they measured against was at least about 113 ms [15]. The account does not say where that cloud endpoint sat or what the workload was, so the comparison is the partners' own [17].
Frandsen's own account is warmer than the flight profile. "Groundbreaking and historic only begin to describe our ST1 flight," said Mikkel Vestergaard Frandsen, Sceye's founder and chief executive [10]. He said the mission "proved not only that our technology, vehicle systems, payload, and operational infrastructure are ready for connectivity services" [11]. The mission was named Service Test 1, and it ended with a planned termination and descent near Moriarty on September 5 [1][6]. The account gives no figure for how many devices were served at once or how much spectrum the test used, and it does not say how many consecutive days the platform held station over Japan [17]. Sceye says its vehicles hold altitude and area of operation through consecutive day and night cycles, operating like geostationary satellites at 1,800 times closer to Earth [12].
"Designed to", "at full scale": that is the wording on the 500-tower coverage figure, and it is a target rather than a measurement taken from ST1 [8]. The follow-up question for anyone writing a resilience plan around it is how fast a second airframe reaches station. ST1 took 13 days to fly from New Mexico to Japan [2].
Two questions decide whether any of this belongs in a network plan next year: whether the gap being filled is coverage or capacity, and whether the need is continuous or episodic. The tested evidence sits in one of those four boxes. Sceye ran emergency calls through an alert messaging system built for large-scale disasters, and tested links to drones [9], which is episodic coverage over a wide area. For continuous coverage of a populated area, the deciding number is consecutive days on station per airframe, and until that is published the honest place for a HAPS in a plan is a standby contract.
What to watch
- A published figure for consecutive days on station per airframe over a single service area.
- Whether SoftBank commits spectrum and a commercial service window in Japan beyond a test flight.
- A second flight to Asia that shows how quickly a replacement airframe can reach station.