AI can enhance 5G network performance by 30%: Ericsson

Mumbai: Telecom operators can improve the performance of their existing 5G networks by 25-30% using artificial intelligence (AI), without necessarily investing in additional hardware, according to Swedish telecom equipment maker Ericsson. This offers a potentially cost-effective upgrade path for operators such as Reliance Jio and Bharti Airtel as they prepare for the transition to 6G.

A large proportion of the equipment Ericsson supplied in 2025 and 2026 is already capable of supporting AI features, Ericsson chief technology officer Erik Ekudden told ET.

"Customers do not necessarily incur an additional hardware cost because this capability is part of the Ericsson products they buy. They do not need to wait for GPUs or 6G to benefit from AI. Software can already deliver gains of 25-30%, and performance will continue to improve as new generations of AI models and software run on existing hardware," he said.


Indian telecom operators, which have invested billions of dollars in rolling out nationwide 5G networks since 2022, are under pressure to monetise those investments while preparing for the next generation of mobile technology.

Ekudden said AI capabilities could be introduced progressively into existing networks through software, although the extent of upgrades would depend on the age of the installed equipment and the features operators wanted to deploy.

"Looking at what Ericsson delivered in 2025 and what we have today in 2026, a large part of our equipment is already upgradeable for AI features and can be continuously enhanced through software," he said.

However, software upgrades would still need to be procured and deployed, while newer hardware would eventually be required to support additional frequency bands and capabilities associated with 6G, he said.

"There is no need to wait for 6G to introduce AI in the RAN or begin moving towards autonomous networks. These capabilities can be developed over time through AI software, with new features, functions and models running on existing hardware," Ekudden said.

Unlike the large AI models developed by companies such as OpenAI and Anthropic, telecom networks require smaller, specialised models designed to perform specific tasks reliably, he said.

However, the growing integration of AI into critical telecom infrastructure raises concerns about cybersecurity and the potential misuse of advanced AI models.

Ekudden said Ericsson was stress-testing its systems against emerging threats and deploying safeguards to prevent AI models from behaving unpredictably. "The AI industry needs a degree of self-control to prevent rogue agents from being unleashed on society," he said.

On India's transition to 6G, Ekudden said the country was well positioned to become an early adopter, potentially as early as 2029 or 2030, provided the broader device ecosystem matures.