Taiwan's bid to challenge global telecom incumbents with 5G Open RAN (O-RAN) is faltering, as network outages in automated factories undermine trust and local vendors struggle to monetize investments. Once championed as a bridge for Taiwan's IT sector to break into communications technology and loosen the grip of Nokia, Ericsson, and Siemens, O-RAN's commercial promise is now in doubt. Industry leaders and analysts see the emerging concept of AI-RAN as the sector's potential reset.
From O-RAN promise to real-world setbacks
Taiwan's Ministry of Economic Affairs (MOEA) initially pushed a four-track strategy spanning 5G mobile chips, small cells, O-RAN, and private 5G networks, supported by NCC's release of the n79 band. In 2022, the Industrial Development Administration (IDA) funded 10 domestic vendors to deploy 13 O-RAN vertical solutions in Kaohsiung's Asia New Bay Area, aimed at smart factories, hospitals, and ports, with further R&D subsidies rolled out in 2023.
Yet field results tell a different story. A leading OEM/ODM disclosed that an O-RAN-based smart factory suffered a two-hour outage caused by core network failure, forcing production to a standstill. Even with Wi-Fi backup for automated guided vehicles (AGVs) and AOI defect inspection, unstable handovers frequently halted AGVs mid-operation. Such reliability lapses have fueled skepticism over whether O-RAN can support mission-critical industrial automation.
Economics and spectrum constraints compound the problem. Beyond upfront hardware, enterprises face opaque maintenance costs and unclear vendor requirements, leaving ROI calculations elusive. In Southeast Asia, the absence of dedicated private 5G spectrum often forces firms to lease bands like n40 from telecom operators, erasing cost savings and throttling performance compared to n78 or n79 frequencies.
These risks are steering buyers back to established suppliers. At Kaohsiung Port's Terminal 7, Evergreen Marine bypassed domestic O-RAN and instead tapped Chunghwa Telecom to install Nokia gear, prioritizing operational stability over experimentation.
Monetization challenges and industry fractures
O-RAN's commercial woes rival its technical failures. With cash flow absent, many Taiwanese firms have already quit the space. The grand plan, delivering end-to-end, one-stop O-RAN systems covering radio units (RU), distributed units (DU), centralized units (CU), and core networks tailored for verticals, has unraveled without tangible returns.
Ting Pang-an, Vice President and General Director of ITRI's Information and Communications Research Laboratories, flagged two critical challenges: "How can O-RAN move beyond private networks to monetize in public networks? And as AI drives heavier 5G traffic, how can O-RAN sustain performance?" He urged Taiwan's sector to build organizational alliances, avoid destructive price wars, and strengthen hardware-software partnerships.
The O-RAN Alliance, founded in 2018 by 32 operators and 269 vendors, and research institutes, symbolized this ambition, staging summits to champion open interfaces and white-box gear after 4G WiMAX's collapse. Taiwan viewed O-RAN as a pathway back into global telecom infrastructure, but the gulf between aspiration and commercial reality is widening.
The AI-RAN pivot
SoftBank's January 2025 white paper has refocused attention on AI-RAN, categorizing it as AI for RAN, AI and RAN (shared infrastructure), and AI on RAN (using RAN for AI applications). Nvidia is spearheading this shift by embedding GPUs into 5G base stations, turning them into edge-computing hubs capable of high-speed, reliable AI workloads.
Observers argue that O-RAN's open interfaces could underpin AI-RAN, if certified internationally, broadening operator choices beyond Nokia, Ericsson, and Huawei while spurring academic innovation. Ting underscored that O-RAN's founding vision — standardized hardware, open interfaces, open-source software, and app-based network management — still holds, and AI may be the lever to unlock its long-delayed value.
Ting concluded that Taiwan's years of O-RAN experience provide a springboard to "pivot elegantly" into AI-RAN, where artificial intelligence could finally deliver the profitability and resilience that O-RAN never realized.
Article edited by Jack Wu