DOCOMO and NTT begin joint AI control of base-station workloads and batteries

The Japanese demonstration coordinates where radio processing runs with solar generation and stored energy, including outage and demand-response scenarios.

NTT DOCOMO and NTT have begun a demonstration that treats radio processing and local energy resources as one coordinated system. The [20 October announcement](https://www.docomo.ne.jp/english/info/media_center/pr/2026/1020_00.html) describes AI-generated plans that move communications-processing workloads between sites while controlling solar generation and storage batteries.

The work connects a DOCOMO base station with an NTT research environment simulating virtualised radio processing. It will test ordinary operation, tight electricity supply and outages, with evaluation planned through 31 December 2026. NTT calculates workload plans, while DOCOMO develops operating plans and controls the power equipment.

A network workload has a location

The computing behind a mobile connection consumes electricity at a particular site. If part of that processing can move, its energy demand can move too. A site with surplus solar generation may become a useful destination for work that would otherwise run elsewhere.

This creates a different optimisation problem from turning equipment down during a quiet hour. The operator has to consider communications requirements alongside power availability, battery charge and the capacity of another site to take the work. The plan connects two systems that have traditionally been managed independently.

For a mobile network, the useful result is a feasible operating schedule. A low-energy plan that overloads the receiving site would not achieve the intended purpose. Equally, a plan that conserves battery at one location while creating an unnecessary demand spike elsewhere would miss part of the grid objective.

Three scenarios make the trade-offs visible

During ordinary operation, the question is whether available renewable power can be used more effectively. During demand response, the system can reduce consumption at a participating site through a combination of battery discharge and workload movement.

The third scenario changes the priority. In an outage, conserving energy at the affected location can extend the time that its backup battery supports communications. Workload movement becomes a resilience tool as well as an energy-management technique.

The transition after an event also deserves attention. Batteries eventually need to recharge. If every participating site resumes charging at once, the recovery can create a fresh demand peak. The demonstration explicitly includes coordinated charging after a demand-response period, making recovery part of the operating plan.

From individual equipment to coordinated infrastructure

The project links NTT's reinforcement-learning optimisation with DOCOMO's AI and mathematical planning and its energy-management platform. Its research environment can simulate workload movement in both directions or leave processing in place.

That flexibility is important because there may be periods when moving work offers little value. A coordinated system needs to choose among valid operating modes according to current conditions, rather than move processing simply because movement is possible.

The companies plan to assess the demonstration before considering connections with data centres and a broader set of energy resources. The present milestone is a test of joint control at base stations. It gives telecom operators a concrete example of how AI can connect network computing with energy operations, using the same infrastructure to address renewable integration, grid pressure and service continuity.

Sources

- [DOCOMO and NTT to Begin Demonstration of Coordinated Workload Shifting and Energy Management](https://www.docomo.ne.jp/english/info/media_center/pr/2026/1020_00.html) — NTT DOCOMO and NTT