The telecom industry faces mounting pressure to balance rising data demands with increasing energy costs and sustainability concerns. As 5G and 6G networks evolve, integrating AI into network infrastructure offers performance enhancements but also drives significant power consumption.
AI is already transforming data processing, storage, and security in data centers. However, the high-powered GPUs required for AI workloads consume five to ten times more energy than traditional computing hardware. This challenge has prompted telecom operators to seek innovative solutions.
SoftBank and Red Hat’s AI-Driven Power Optimization
To address these concerns, SoftBank and Red Hat have developed a system to monitor and optimize power consumption in data centers running virtualized Radio Access Networks (vRAN) and AI applications. Their collaboration is part of SoftBank’s AI and Telecom Radio Access System (AITRAS), a converged AI-RAN platform designed to enhance energy efficiency.
Key Highlights of SoftBank and Red Hat’s Partnership:
- AITRAS, developed by SoftBank and Red Hat, integrates AI with mobile networks.
- The solution leverages Kepler, an open-source power monitoring project from Red Hat.
- The initiative enhances sustainability and operational flexibility for service providers.
- SoftBank is deploying AITRAS to GPU-equipped data centers across Japan.
While AI applications improve mobile network performance, running them alongside vRAN demands substantial electrical power. This challenge is exacerbated by growing pressures to adopt renewable energy and reduce carbon emissions from fossil fuels. Additionally, Japan’s urban data center hubs place additional strain on power grids, requiring effective energy load management.
SoftBank’s energy optimization strategy is built on Red Hat’s OpenShift platform, incorporating Kepler to monitor and manage power consumption within Kubernetes environments. Kepler collects real-time power usage data from servers, clusters, and individual applications, feeding this information into the AITRAS Orchestrator for smarter resource allocation.
How AITRAS Enhances Energy Efficiency:
- Dynamic Resource Allocation: AI workloads are assigned to the most power-efficient clusters.
- Power Consumption Limits: Each cluster operates within predefined thresholds.
- Renewable Energy Integration: Workloads are deployed based on carbon intensity and renewable energy availability.
One of Kepler’s standout features is its ability to estimate power consumption based on GPU usage. It utilizes Multi-Instance GPU (MIG) technology, which divides a single GPU into multiple virtual GPUs, ensuring more precise power allocation across workloads. This allows AITRAS to optimize AI applications with real-time energy data, improving overall efficiency.
SoftBank and Red Hat showcased this advanced technology at Mobile World Congress Barcelona 2025, where industry experts discussed its impact on future telecom networks.
Ryuji Wakikawa, VP & Head of the Research Institute of Advanced Technology at SoftBank: “Electricity and telecommunications continue to be critical infrastructure supporting society. By monitoring and predicting power consumption, AITRAS optimizes equipment for energy efficiency while reducing risks through distributed deployment. The integration of telecom and power infrastructure paves the way for AI-driven networks.”
Chris Wright, CTO & SVP of Global Engineering at Red Hat: “Red Hat and SoftBank are shaping the future of 5G and 6G by combining AI with RAN for network orchestration and optimization. With Red Hat OpenShift, AI-RAN introduces a groundbreaking approach to network operations, improving resource efficiency and sustainability.”
SoftBank’s rollout of AITRAS demonstrates how AI can enhance telecom energy management while supporting AI-driven applications. By leveraging Red Hat’s OpenShift and Kepler technologies, telecom providers can move toward more sustainable, AI-powered infrastructure, ensuring the industry remains both energy-efficient and performance-driven in the 5G and 6G era.
