Summary

  • NTT DOCOMO and Abside Networks signed an MOU on 13 October 2026 to study potential future HAPS collaboration, including AI-based network control and use of stratospheric operating data.
  • The announcement defines a research agenda; it reports no joint deployment, service date, equipment order or AI-control result.

A radio link from the stratosphere is only one part of a HAPS network. Someone must decide where a beam points, how capacity is assigned, which spectrum is usable and how interference changes those choices. That control problem—not a new service launch—is the centre of the agreement NTT DOCOMO and Abside Networks announced on 14 October.

The companies say they signed a memorandum of understanding on 13 October 2026. Its stated purpose is to explore future communications infrastructure using high-altitude platform stations, or HAPS, alongside AI-enabled next-generation networks. The planned work is a joint study toward possible future collaboration. The release does not announce a commercial service, a deployment, a procurement commitment, a schedule or a result from a DOCOMO–Abside test. That status matters: a list of technical work areas is not evidence that a control system is operating.

The scope is unusually specific about the decisions that such a system might have to coordinate. The parties name system architecture for wide-area, high-capacity and massive-connectivity communications. They also list AI applications to demand forecasting and to control of beams, capacity, spectrum and interference in a stratospheric radio access network. A separate work area concerns collection and use of data about communications quality, network operations and sensing; another looks toward future integration with external applications and service platforms. These are the subjects of study stated in the DOCOMO announcement, not a published implementation plan.

The pieces are coupled. A demand forecast can inform capacity allocation, but the value of that forecast depends on what the system observes and how quickly conditions change. Beam steering changes where radio energy is concentrated; spectrum and interference decisions constrain which beams or capacity allocations are viable. Quality and operations data could help evaluate decisions after the fact, while sensing data may describe the environment that the network is trying to serve.

The release names those inputs and control targets but does not specify their interfaces, update cycles, objective function, safety limits or responsibility when a decision degrades service. Those details may be worked out later; they are not established by this announcement.

The physical setting helps explain why the control question is difficult. DOCOMO says HAPS equipment and RAN design must account for stratospheric conditions that differ from those on the ground. Its release defines HAPS as unmanned platforms operating around 20 kilometres above Earth. A platform can potentially extend coverage across areas difficult for terrestrial networks, but that reach does not make the stratospheric radio layer equivalent to a tower. The equipment, geometry and relationship with ground networks have to be adapted.

The two companies describe complementary capabilities. DOCOMO has been studying and demonstrating HAPS communications and says its future network vision would combine terrestrial networks with non-terrestrial networks, including the stratosphere. It says realization requires RAN and orchestration technologies that can control terrestrial and non-terrestrial networks together. Abside says it is developing radio equipment for terrestrial and non-terrestrial RANs using many antenna elements for precise, high-resolution beam control, as well as software for integrated control across ground, stratosphere and space.

Those descriptions explain why the companies might study the subject together. They remain company statements about experience, development and vision, rather than independent proof of joint capability.

There is real HAPS evidence in DOCOMO’s earlier work, but it answers a different question. In 2025, DOCOMO and Space Compass reported a test in Kenya using AALTO’s Zephyr aircraft at about 20 kilometres. A non-regenerative relay connected an LTE base station and ground gateway to 4G devices, with the release reporting forward-link throughput above 4.66 Mbps. That demonstrates a particular link in a particular test. It does not establish Abside’s participation, AI-based beam or spectrum control, or a commercial HAPS service.

A separate 2024 announcement about a US$100 million consortium investment in AALTO set out an earlier service-entry target for 2026; that investment and roadmap involved different parties and should not be folded into the October MOU as if it were its implementation.

The useful next evidence would show the study turning into a defined control loop: what the system measures, which actor can change a beam or capacity allocation, how spectrum constraints enter the decision, what performance objective is used, and how operators can inspect or override the result. A test would also need to distinguish prediction quality from service outcomes. Better demand forecasts do not automatically mean higher availability or lower interference; those are separate measurements.

Heng Lu Note 65, “Running-Code Primacy,” offers an editorial lens for this distinction: architecture matters when its behavior can be observed in operating systems. It is not evidence about either company. Here the announced architecture is prospective. The MOU makes the agenda legible; the next milestone is evidence that its proposed sensing, orchestration and radio controls work together under stated conditions.

Sources