Nokia accelerates AI-RAN adoption as global operators embrace AI-native network evolution on NVIDIA platforms

Nokia said AI-RAN trials and proofs of concept using NVIDIA Aerial RAN Computer are expanding across operators in North America, Europe, Asia-Pacific and the Middle East, including A1 Group, Chunghwa Telecom, du, e&, Mobily, stc, TPG Telecom and Zain Saudi. Its AI-RAN platform has delivered more than 20% spectral-efficiency improvement, with additional software-driven gains targeted for 2027 and 2028. The growing operator ecosystem supports Nokia's strategy to make 5G networks software-upgradable to AI-native 6G while improving carrier capacity, network economics and edge-AI service opportunities.

Analysis

The investable implication is less near-term radio-unit demand than a potential shift in RAN economics toward software, accelerated compute refreshes, and vendor lock-in around the AI software stack. NOK can use AI-RAN validation to defend pricing and lower churn in Mobile Networks, but trials do not yet establish purchase commitments; operators will demand measured opex savings and proof that AI capacity gains avoid incremental spectrum or densification spend. The key earnings sensitivity is whether AI-RAN converts into higher software/content per site rather than merely subsidizing NVIDIA compute attached to flat Nokia hardware revenue.

NVDA gains strategic optionality as telecom becomes an incremental edge-compute market, although carrier capex cycles and stringent return hurdles make this immaterial to FY27 estimates absent commercial orders. The more immediate competitive pressure falls on Ericsson (ERIC), whose installed base faces a comparable need to offer cloud/AI-native RAN economics, and on legacy specialized RAN silicon suppliers if general-purpose accelerated compute captures baseband workloads. For TMUS and CHT, efficiency improvements are economically positive only if they translate into lower network capex per incremental GB; competitive wireless pricing could otherwise pass the savings through to consumers.

Consensus is likely to reward the architecture narrative before monetization is visible. Commercial conversion will be gated by latency/reliability results in live networks, power consumption per delivered bit, and whether regulators permit shared edge infrastructure and AI workloads on carrier sites. Over the next 1-3 months, watch for named commercial contracts, site counts, and disclosed capex/opex outcomes; over 6-18 months, software revenue mix and gross-margin progression matter more than trial announcements. A NOK thesis is falsified if Mobile Networks guidance remains flat despite deployments, or if NVIDIA-based configurations increase operator total cost of ownership versus purpose-built RAN silicon.

Key Decisions for Investors

• Maintain a tactical long NOK versus short ERIC pair for 3-6 months, sized modestly: NOK has more narrative and potential software-content upside from commercial AI-RAN conversion, while ERIC is exposed to the same operator budget constraints. Reassess on either company’s next Mobile Networks margin/guidance update; exit if NOK fails to identify paid deployments or if ERIC announces comparable large-scale wins.
• Do not chase NVDA on this development alone. Set an alert for carrier AI-RAN orders with disclosed site volumes or multiyear values; only then consider adding to NVDA exposure, since telecom remains too small to alter near-term data-center revenue estimates and could carry lower margins than core AI infrastructure.
• Use NOK 6-9 month call spreads rather than outright shares only after confirmation of a commercial contract or raised network-business outlook: target a defined-risk structure with approximately 2:1 upside/downside. Trial headlines without order value are insufficient evidence for a durable rerating.
• Monitor TMUS quarterly capex intensity and cost per GB over the next two reporting cycles. A sustained decline without network-quality deterioration would support a 6-12 month long TMUS thesis; rising capex from edge-compute integration or aggressive price competition would negate the efficiency case.