Ericsson’s RAN Compute Platform Lands at NTT Docomo: A Practical Step Toward AI-Native Networks

Japanese operator NTT Docomo is taking a concrete step into AI-enhanced radio networks. The company has chosen Ericsson’s next-generation RAN Compute platform to handle its 4G, 5G, and emerging AI-driven workloads across Japan. Deployment began in July 2026, and the move signals how leading operators are investing in hardware that can support smarter, more efficient networks without waiting for every last piece of the AI puzzle to fall into place.

Why This Matters Right Now

Mobile traffic keeps climbing, and operators everywhere face the same squeeze: deliver better performance while controlling costs and energy use. NTT Docomo’s decision highlights a growing reality—AI capabilities are moving from lab demos into production infrastructure. The new platform is purpose-built with Ericsson’s latest silicon and is explicitly optimized for AI and machine learning tasks running inside the RAN itself.

This isn’t just another baseband upgrade. It’s designed to create what the industry calls AI-native networks: systems that can learn from live traffic, predict issues before they hit customers, optimize radio resources on the fly, and eventually support more advanced “agentic” applications that act with greater autonomy.

What the Platform Actually Delivers

Ericsson says the new RAN Compute hardware delivers up to twice the processing capacity of the previous generation while using less than half the energy. That combination is especially valuable in centralized RAN (C-RAN) setups, where multiple sites share pooled compute resources. Lower power draw means cheaper site operations and a smaller environmental footprint—two wins that matter as regulators and customers pay closer attention to sustainability.

The platform supports 4G, 5G, and 5G Advanced today, with a clear path for long-term software evolution. That software-first approach lets operators add new AI features over time without ripping out hardware. NTT Docomo’s own executives noted that the upgrade will help maintain communication quality amid rising traffic and prepare the network for future AI uses.

How It Fits the Bigger AI-RAN Picture

Open RAN and AI-RAN discussions often focus on disaggregation, open interfaces, and GPU acceleration. Yet traditional vendors like Ericsson are also advancing their own silicon and software stacks to embed intelligence directly in the RAN. NTT Docomo’s choice shows that operators aren’t forced into an either/or decision. They can work with proven suppliers on high-performance hardware while exploring multi-vendor Open RAN elements elsewhere in the network.

The platform’s emphasis on AI workloads—traffic optimization, radio performance tuning, fault prediction, and energy reduction—aligns closely with what many Open RAN operators hope to achieve through RAN Intelligent Controllers (RICs) and rApps/xApps. In practice, we’re seeing a convergence: whether the intelligence lives in a proprietary baseband or an open RIC, the goal is the same—networks that respond faster and smarter to real-world conditions.

Energy Efficiency as a First Win

One of the most immediate benefits highlighted is energy savings. RAN equipment is a major power consumer in mobile networks, and every percentage point of improvement adds up across thousands of sites. By achieving double the capacity at half the energy, Ericsson’s silicon offers operators a tangible way to scale capacity without proportional increases in electricity bills or carbon emissions.

NTT Docomo has been an Ericsson customer for years, including a recent commercial rollout of 4.5 GHz Massive MIMO radios. The new RAN Compute deployment builds on that relationship and extends it into the AI era. Both companies have signaled they will continue collaborating on next-generation architectures that anticipate the extra traffic and processing demands created by AI applications.

What Operators Should Watch

For Open RAN enthusiasts, this story offers a useful reality check. While open interfaces and multi-vendor ecosystems remain important long-term goals, performance and efficiency still drive purchasing decisions today. Ericsson’s ASIC-based approach delivers measurable gains in capacity and power that GPU-based alternatives are still proving at scale in many markets.

At the same time, the platform’s software-defined nature means it can evolve. Features like AI-native scheduling, beamforming optimization, and observability tools can be added via software updates. That flexibility mirrors the value proposition of Open RAN’s RIC layer—intelligence that improves over time without constant hardware refreshes.

NTT Docomo’s deployment also underscores the importance of local partnerships and proven track records. Japanese operators have historically favored reliable, high-quality infrastructure, and this move shows they are willing to invest in platforms that support both current 5G needs and future AI capabilities.

Looking Ahead

AI-native RAN won’t arrive overnight as a single revolutionary product. It will emerge through incremental upgrades like this one—new silicon here, smarter software there, better orchestration across the network. NTT Docomo’s choice of Ericsson’s RAN Compute platform is one such step: practical, measurable, and aimed squarely at the challenges operators face right now.

As more networks add AI capabilities at the edge of the radio access network, the lines between traditional RAN, Open RAN, and AI-RAN will continue to blur. The winners will be those operators and vendors who deliver real improvements in performance, cost, and sustainability while keeping the door open for greater openness and innovation down the road.

NTT Docomo and Ericsson’s partnership shows one clear path forward. Others will follow with different combinations of hardware, software, and interfaces. The common thread is clear: the RAN is getting smarter, one deployment at a time.

Sources

• https://www.telecomtv.com/content/5g/ericsson-supplies-ran-compute-platform-for-ntt-docomo-in-japan-55999/
• https://www.rcrwireless.com/20260804/carriers/telco-diary-network-platforms-ai-hill-of-beans
• https://www.ericsson.com/en/press-releases/2026/7/ericsson-supplies-ran-compute-platform-for-ntt-docomo-in-japan