.webp)
.webp)
AWS has outlined a three-part agenda for the telecommunications industry: modernising legacy systems, using agentic AI to automate networks, and linking cloud infrastructure with emerging connectivity models, including satellites.
Ishwar Parulkar, AWS Chief Technologist for Telecommunications, detailed the themes during a roundtable ahead of the Mobile World Congress trade show. The discussion covered everything from migrating long-standing IT estates to how operators might prepare for 6G and address digital sovereignty concerns.
Many operators still run core business and network software on ageing platforms in privately managed data centres. Parulkar said AWS sees a significant opportunity to move these environments to cloud-based deployment models. "We see telcos having a lot of legacy estate in terms of mainframe VMware applications and old legacy applications like Sun Solaris. That, to us, is another area where we can help pretty much every telco in the world that has that type of estate," he said.
AWS cited its work with AT&T as an example. The programme covers more than 1,500 workloads, including more than 1,000 VMware applications. The target environment includes AWS Outposts, which places AWS infrastructure at customer sites and other locations outside AWS regions.
Network modernisation
AWS also tied IT modernisation to changes in how network functions and connectivity services are delivered. It pointed to an arrangement with Nokia to offer a 5G core in a software-as-a-service model, and to work with Lumen on programmable last-mile connectivity linking enterprise branch locations directly to AWS regions.
These approaches aim to shift telecom infrastructure towards more software-driven operations. Operators are under pressure to reduce maintenance overhead and speed up service changes. Cloud platforms have increasingly become part of that strategy, particularly for new digital services and for network components that fit a virtualised model.
Agentic automation
A second theme focused on network automation through agentic AI. The session described a shift from reactive operations to more autonomous approaches for root cause analysis, handling service impairments, and traffic steering. Parulkar described this as a turning point for a long-held industry ambition. "Telcos have been talking about [self-optimisation and self-healing] for a long time, but these were very aspirational. With the agent it all becomes real. I think agent is perfectly suited and capable of solving these problems," he said.
AWS highlighted Ericsson's R Apps as an example of AI-driven optimisation for the radio access network. The applications will be made available as a service on the AWS Marketplace. The pitch is that operators can query network behaviour and request optimisation through natural-language interfaces.
Parulkar also addressed concerns about supplier dependence as operators rely more on hyperscalers for infrastructure and software tooling. He said AWS agent primitives support third-party frameworks such as LangChain and LlamaIndex, as well as external models from OpenAI and Gemini.
Sovereignty questions
Data governance and deployment control remain central to telecom cloud adoption. AWS referenced Dedicated Local Zones as one deployment model and cited use by the Singapore government. It also pointed to the European Sovereign Zone, which it said operates disconnected from other AWS infrastructure.
Parulkar argued that operational caution and organisational habits are often bigger blockers than the underlying technology. He said the industry needs a mindset shift on risk and complexity when moving legacy environments, and linked recent progress in AI tooling to safer, more manageable migrations.
6G direction
On 6G, Parulkar described an evolution that goes beyond cloud-native design. "We see 6G as not only being cloud native but being AI native. Not only will AI be more integrated into the network to optimize as well as manage the network, but the network will be built with AI applications in mind, applications running on top of the network," he said.
He also described 6G as combining digital twins with sensors, robots, and autonomous vehicles, and as merging terrestrial and non-terrestrial connectivity. On satellite links, he said space-based networks will not replace terrestrial infrastructure because of bandwidth limits and physics constraints. Instead, he said, satellites will complement ground networks and integrate more tightly as standards evolve.
AWS tied that view to its own connectivity footprint. Operators such as AT&T and Verizon are building high-capacity fibre connections into AWS data centres. It also pointed to Amazon's Project Kuiper as a non-terrestrial option for reaching areas where fibre build-outs are uneconomic.
Telco revenues
The roundtable also addressed a persistent challenge for the sector: generating growth beyond basic connectivity. Parulkar argued that agentic AI could give operators deeper visibility into customer experience and enable more personalised services. He cited SK Telecom's Aster personal assistant and T-Mobile's live network translation as examples of differentiation.
He also pointed to the commercial value of data assets held by operators.
"The other big piece that telcos have, which is underutilised today, is data. Telcos have a lot of data assets, and I think there's an opportunity for telcos to create data products that agentic AI applications can use to create new sources of revenue," said Parulkar.
AWS also cited recent customer activity in the Asia-Pacific region. Vocus in Australia reported more than a 50% improvement in coding efficiency and time-to-code after using Amazon AI-powered tools. KT in South Korea built a business-to-business customer service assistant using Amazon Bedrock and the Amazon Nova Pro models for its communications platform.
In South Korea, the YKCS Open RAN Collaboration Centre, a government-backed research consortium, is using AWS for experimentation and evaluation linked to 6G AI RAN. AWS also said it is joining the AI Network Alliance in Korea to collaborate across the AI RAN ecosystem.