ZTE and Ucell prove AI cuts mobile network OpEx

AI is driving telecom energy savings as Ucell and ZTE cut mobile network OpEx through dynamic power management.

Power consumption consistently dominates the operating expenses of national mobile operators and enterprise private network owners. Running communications hardware at peak capacity around the clock incurs massive utility bills that drain capital from other engineering priorities. 

Ucell, the state-owned mobile operator in Uzbekistan, recently completed an AI deployment alongside Chinese vendor ZTE to reduce these baseline costs. The companies reported a 10.6 percent reduction in overall network energy usage across the deployment footprint.

Radio Access Network (RAN) equipment draws heavy wattage even when idle. By applying predictive algorithms to user demand, operators can force underutilised base station components into low-power sleep modes during off-peak periods. The algorithm then wakes the hardware milliseconds before regular traffic patterns resume, maintaining the expected quality of service without illuminating the entire grid unnecessarily.

Translating energy efficiency into P&L improvements

For a national carrier, cutting electricity consumption by a double-digit margin removes millions of dollars from annual operating expenses. This directly improves margins without requiring consumer price increases or aggressive subscriber acquisition campaigns.

Provisioning a private cellular system for absolute peak capacity guarantees wasted capital. Building algorithmic intelligence into the power management layer allows the infrastructure to scale its consumption directly alongside operational intensity. IT directors evaluating vendor proposals increasingly scrutinise total cost of ownership over the expected hardware lifecycle. 

Demonstrable reductions in power overhead offer a highly effective business case compared to raw throughput metrics. An autonomous system that manages its own utility footprint changes the financial profile of industrial connectivity investments.

To deploy machine learning models within live radio environments, operators must integrate predictive software overlays atop base station equipment that was originally designed for always-on operations. The hardware components undergo thermal and mechanical stress when constantly powered up and down. Network planners must calculate the mean time between failures against the exact value of electricity saved to ensure hardware replacement costs do not consume the utility savings.

The models directing these sleep cycles rely on continuous ingestion of telemetry data. Training these time-series algorithms requires massive historical datasets to ensure the software does not incorrectly predict a low-traffic window during an unmapped local event.

False negatives – where the system assumes low demand but actual users attempt to connect en masse – result in dropped packets, high latency, and breached Service Level Agreements. Trusting an algorithm to control core network availability requires highly reliable fail-safe mechanisms.

The computing overhead required to process this telemetry must also be factored into the final efficiency calculation. Running continuous inference models on regional servers consumes power. The net savings reported by Ucell accounts for the energy required to run the predictive engine itself, but enterprise IT teams building smaller private networks must verify that local compute costs do not outweigh the base station savings.

Of course, handing autonomous control to an algorithm requires entirely new oversight tooling. Network Operations Centre engineers must transition from manual capacity provisioning to monitoring algorithmic behaviour. This requires distinct dashboard interfaces that highlight the reasoning behind automated decisions rather than just displaying raw hardware status alerts.

Global vendor dynamics and enterprise ecosystems

The telecoms supply chain features intense competition regarding the total cost of ownership. ZTE and domestic competitor Huawei use operational efficiency metrics to maintain market share against European alternatives like Ericsson and Nokia.

Demonstrating a live, large-scale deployment that actively reduces utility bills gives vendors tangible data for enterprise procurement discussions. When IT directors assess partners for industrial upgrades, live proof points of automated expense reduction carry heavy weight in the final vendor selection process.

This methodology aligns with broader IT industry efforts to apply predictive computation to infrastructure management. Cloud hyperscalers use similar machine learning techniques to manage data centre cooling and server allocation based on global traffic loads.

As enterprise IT directors converge their cloud architecture and private networking teams, the expectation for automated resource scaling crosses over from software applications to radio frequency hardware. Network infrastructure must adapt autonomously to the physical environment it serves, ensuring that massive digital transformation projects remain economically-viable over the long term.

See also: KDDI targets AI infrastructure and 6G readiness in new 3-year plan

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the IoT Tech Expo and Cyber Security & Cloud Expo. Click here for more information.

Telecoms is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.