Nokia and Indosat signed a strategic network modernization and AI-RAN rollout agreement that is aimed at moving Nvidia-powered AI-RAN from collaboration to commercial deployment across Indonesia, with field trials scheduled for end-2026 and broad rollout expected to start in 2027.
The deal positions Nokia as the key 5G RAN supplier for Indosat’s nationwide low- and mid-band 5G program, while formally locking in NVIDIA’s GPU-accelerated AI-RAN platform as the compute foundation for AI-enhanced RAN functions and edge AI workloads on Indosat’s network.
What Indosat is buying: AI-RAN compute plus a full radio refresh
The modernization includes Nokia’s next-generation Habrok and Pandion radio models, along with its Levante baseband platform.
It also brings centralized RAN (C-RAN) systems and Nokia network management and automation tools into the scope. The architecture described ties together AI-for-RAN, AI-on-RAN, and AI-and-RAN on a single software-defined infrastructure.
In that model, AI-for-RAN uses machine learning to optimize network operation, including dynamic resource allocation, traffic prediction, and interference management. AI-on-RAN runs edge AI inferencing workloads such as video analytics and generative AI on the same hardware that performs baseband processing. AI-and-RAN treats the hardware as a multi-purpose platform that can host connectivity and third-party AI workloads concurrently.
How the stacks fit: anyRAN/Cloud RAN on Kubernetes with Nvidia Aerial
The cloud-native RAN functions are described as DU and CU running as Nokia anyRAN/Cloud RAN software, orchestrated via Kubernetes on GPU-accelerated servers.
Nvidia supplies the AI Aerial platform, described as the accelerated computing stack and GPUs that host the system. Nokia and Nvidia argue that this approach delivers better spectral efficiency and energy optimization than traditional purpose-built RAN hardware.
Timeline: from MoU and R&D center to end-2026 trials and 2027 deployment
Nokia, Indosat, and Nvidia had been working on AI-RAN for over a year before this agreement, beginning with an MoU at MWC 2025.
In late 2025, an AI-RAN Research Centre was launched in Surabaya and positioned as a long-term R&D hub for AI-native wireless and edge AI in Indonesia. At MWC 2026, the partners demonstrated an AI-powered 5G call linking Barcelona and Jakarta in real time; the article also says this demonstration was cited as the first AI-powered 5G call in Southeast Asia.
According to the rollout plan, field trials on Indosat’s live network are slated for end of 2026. Those end-of-2026 trials are designed to validate spectral efficiency gains, energy savings, and AI service performance in real-world conditions rather than controlled environments, with broad commercial AI-RAN deployment expected to begin in 2027 across Indosat’s footprint.
Commercial pressure points: energy, CAPEX/opex, and real-world reliability
The article frames the rollout as aligned with Indonesia’s “Golden Indonesia 2045” vision, and it also points to the practical challenge of improving connectivity outside Java where coverage remains uneven, while managing dense urban capacity needs without proportionally increasing hardware and energy costs.
At the same time, it notes skepticism that high capex and opex for GPU hardware can be justified in a tight-ARPU market like Indonesia. It also calls energy consumption an open question, saying GPU-accelerated cell sites need to demonstrate efficiency gains that offset their power draw versus traditional RAN systems and that the energy-efficiency case has rested largely on vendor claims.
Finally, it highlights the gap between lab and field: AI algorithms trained under controlled conditions in Surabaya will need to perform reliably across Indonesia’s noisy, heterogeneous real-world environment, described as involving thousands of islands, widely varying traffic patterns, and infrastructure quality ranging from dense urban Jakarta to remote rural sites. The end-of-2026 field trials are said to be designed to answer those reliability questions.