News

The company says its new SingleRAN 22.1 wireless software aims to build a multidimensional foundation for the Mobile AI network

As AI models grow more advanced, the use of AI agents continues to proliferate. According to forecasts from IDC, there will be more than 1 billion actively deployed AI agents by 2029, raising to 2.2 billion in 2030. Other estimates are even more bullish, with Huawei’s Intelligent World 2035 report predicting 900 billion AI Agents will be active globally, with 90% operating in mobile scenarios. Indeed, earlier this year NVIDIA CEO Jensen Huang said he envisaged a workforce where each human employee would be supported by over 100 AI Agents. 

The consensus is clear: the ‘Agentverse’ – an interconnected ecosystem of autonomous AI Agents – is coming, if it is not here already.  

This explosion in AI Agents presents a significant challenge to wireless network operators. Unlike traditional legacy traffic, Agent services require continuous multimodal data transmission, real-time interaction, and constant multi-agent collaboration. This places unprecedented demands on mobile networks, requiring massive uplink capacity, deterministic low latency, and guaranteed service levels.  

Against this backdrop, traditional downlink-centric networks focussed on ‘best-effort’ connectivity will soon find themselves obsolete. Wireless networks will need to be systematically redesigned with a focus on capability allocations, dynamic resource orchestration, and novel interaction paradigms.  

This systematic overhaul is exactly what Huawei is targeting with its latest software update, SingleRAN 22.1. 

SingleRAN 22.1: Six features for the Agentic ecosystem  

Huawei’s SingleRAN 22.1, launched in Shanghai earlier this year, is the company’s first commercial software release purpose-built for the Mobile AI era. It includes six key technical advances that, when taken together, form a robust foundation for supporting Agentic AI at scale.

• GigaUplink focusses on overcoming terminal power and spectral efficiency bottlenecks with two new algorithms. The Multi-layer Fusion Uplink Reception algorithm aims to dynamically optimise the network through coordinated power control, non-contiguous resource scheduling, and intelligently combining beams from multiple antennas. The second algorithm, Uplink CoMP Enhancement, works to improve how multiple cells cooperate and suppress interference. 

  When combined, these algorithms have demonstrated average uplink throughput gains of 24% to 33% in live field trials in Macau and Yunnan, expanding 10 Mbps uplink coverage by approximately 2 dB to support uplink-heavy AI streams.  

• GigaBand is designed to optimise existing spectrum across both 4G and 5G. It uses centralised scheduling to dynamically allocate spectrum, time, power, and spatial resources between LTE and NR, improving overall spectrum efficiency by 20%. It also enables more flexible switching between dynamic spectrum sharing and NR-only, delivering 10–30% better NR performance without impacting 4G performance at existing levels. 

  Commercial deployments have achieved 34% higher 5G standalone experience, around 20% uplink improvement, and 5G uplink speeds exceeding 1 Gbps.  

• iBeam 3.0, the latest update to Huawei’s beamforming optimisation solution, is designed to improve 5G downlink performance, particularly under medium-to-heavy network loads. Building on previous iBeam generations, which delivered over 20% gains through interference suppression and user pairing, iBeam 3.0 further improves beam accuracy, user scheduling and interference management. It combines real-time channel conditions with Sounding Reference Signal measurements to more accurately target devices, optimises user pairing, power allocation and modulation, and suppresses interference. 

  In a live Saudi Arabian network, iBeam 3.0 delivered a 15% improvement in downlink rates versus previous iBeam generations through a software-only upgrade. 

• IDEA (Intelligent Differentiated Experience Assurance) is designed to move mobile networks from “best effort” to differentiated, guaranteed experiences. It dynamically allocates network resources according to the needs of different services and users; for example, prioritising high uplink capacity for AI services, low latency for gaming, and stable connectivity for real-time applications. IDEA combines elastic speed control with dynamic resource reservation to provide more predictable performance, achieving over 95% target-rate satisfaction for 2 Mbps uplink and 5 Mbps downlink in heavy-load metro scenarios. 

  A deployment for HKT showed that the proportion of <40 ms latency sessions in low latency gaming scenarios improved by 5%, 5 Mbps uplink rate satisfaction improved by 11.7% for live streaming customers, and VIP user live streaming satisfaction rose from 92.23% to 97.69%. 

• 0 Bit 0 Watt 0 Loss is Huawei’s intelligent energy-saving solution, designed to reduce 5G network power consumption without compromising service performance. It enables network equipment to enter deep sleep during periods of low demand and wake up quickly when traffic increases. In super-deep dormancy mode, residual power consumption falls below 10 W, with a 30-second wake-up, while fast deep dormancy reduces consumption below 50 W, with just a 5-second wake-up and zero service impact. This change means that energy savings are no longer achieved only during periods of predictably low usage, like nighttime, and instead are achieved all-day.  

• RANSpirit is an AI-powered wireless network intelligence agent, designed to automate network optimisation and operations. Instead of relying heavily on engineers to monitor millions of network settings and manually diagnose problems, RANSpirit uses a telecom AI model and digital twin to understand network conditions, identify issues and determine the best actions. It can simultaneously balance user experience, network performance, energy efficiency and operational needs, rather than improving one at the expense of another. By continuously sensing, analysing, deciding and executing, RANSpirit enables minute-level automated optimisation, helping operators move from reactive, manual network management towards proactive, autonomous operations. 

In addition to enhancing the network’s multi-dimensional capabilities in software, SingleRAN 22.1 also provides a foundation for the next generation of baseband processing hardware. This version of the technology is the first to support Huawei’s next-generation UBBPi series baseband boards, combining its advanced algorithms with the latest baseband technology to deliver the diverse network capabilities Mobile AI services will demand.  

A connectivity foundation for an AI-powered future 

For Huawei, these new capabilities represent not only improvements to existing 4G and 5G services but the creation of an entirely new network topology paradigm in anticipation of continued AI growth. These integrated solutions resolve foundational network bottlenecks, allowing operators to systematically deliver the deterministic latency, high uplink capacity, and differentiated performance required to support billions of interconnected AI Agents across the emerging Agentverse.  

“Mobile AI is not a service upgrade — it is a paradigm shift for the industry. As the first commercial version for the Mobile AI era, SingleRAN 22.1 comprehensively reconstructs network capabilities, empowering operators to seize new opportunities in the Token economy, making intelligence omnipotent, omnipresent, and perpetual,” said He Yunru, Vice President of Huawei Wireless Network Solutions. 

Huawei’s SingleRAN 22.1 has already been commercially deployed at 35 operator sites across 23 countries. Over 100 operators globally are expected to have completed deployment by the end of the year. 

Also in the news
Shared Rural Network rollout extends to UK national parks with over 150 4G masts live
Vodafone touts latest attempt to bridge the UK’s digital divide
Virgin Media O2 cuts 5,200 tonnes of carbon dioxide emissions