Top Telecom Trends for 2025: the mid-year review

As we hit the halfway mark of 2025, we revisit our trends predictions from the start of the year – as voted for by our social media followers – to see how each trend is progressing.

Six months on since we first looked forward to what could be this year’s top telecoms trends, it’s time to review those predictions and assess how each is progressing globally.

As 2025 began, we asked our followers on LinkedIn and X to vote on the top telecom trends for the year. The community picked six priorities:

• Artificial Intelligence
• Satellite connectivity
• Open Digital Architecture (ODA)
• Digital engagement
• Customer centricity
• Cybersecurity

… reflecting the industry’s push toward smarter networks, broader connectivity and more customer-focused services. Now, with half the year behind us, it’s time to measure up which of these themes are actually driving change.

Artificial Intelligence

A recent Nvidia industry survey reported that 44% of telco professionals are investing in AI for customer experience optimisation, and 40% are using AI in network planning and RAN operations.

Earlier this year, Telstra formed a joint venture with Accenture to scale its AI roadmap. Telstra’s CEO noted that the company already has “hundreds of value-driving AI use-cases” from customer service to plans for “self-healing” networks and other agentic AI applications.

The shift toward agentic AI is well underway, with operators in every region launching initiatives. Telenor and Ericsson unveiled a proof-of-concept agentic system in early March: trained on Telenor’s network data, the AI autonomously optimises 5G RAN capacity and power.

A study from Stanford shows workers want agents to replace repetitive tasks and free them up for higher value, cognitive work: “Moving beyond a simple automate-or-not dichotomy AI agent integration may reshape the core human competencies, shifting from information-focused skills to interpersonal ones.”

Meanwhile, SK Telecom continues to enhance its AI-powered assistant, adot, adding Google’s Gemini 2.0 Flash model and other LLMs to improve user experiences, while KT Corp says deploying Microsoft Copilot and AI agents has cut task completion times by 50% and boosted infrastructure efficiency by 20%.

Yet despite these stats, the real picture is more complicated.

Half of the companies that set out to replace customer service with AI are changing their minds, according to Gartner; chief among them is buy-now-pay-later firm Klarna, whose CEO made a big deal out of sacking hundreds of customer support staff and replacing them with AI… before quietly rehiring them.

There have been a number of wins for telcos implementing AI to improve networks, automate operations and personalise service – but amidst all the successes, there are more than a few pretenders seeking to ride the wave of hype. Builder.ai – and its $1.5 billion valuation – collapsed earlier this year after it was revealed that its “AI-powered” agent platform was in reality hundreds of programmers manually coding projects.

And the biggest name to throw cold water on AI? One of the biggest names in tech.

After the muted response to Apple Intelligence, Apple released a research paper that argues that Large Language Models are architecturally incapable of reasoning or understanding. Apple’s paper reportedly demonstrates that when confronted with tasks beyond the scope of their training data, the ability to solve these problems collapsed. Strangely, even when presented with the solution to a problem, the model still failed.

However, not only is GenAI’s own performance hitting the wall – this might also extend to its users in the long term; based on brain scans of users who rely on GenAI to complete tasks, “LLM users consistently underperformed at neural, linguistic, and behavioural levels,” according to a study published by MIT.

Though this news is unlikely to put off many users, in what could be the biggest challenge to GenAI so far, Disney and Universal have sued Midjourney for copyright infringement, citing numerous instances of Midjourney and DALL-E generating copyrighted characters and movie scenes almost verbatim. Should their legal action bear fruit, it could kick off a wave of further proceedings against AI firms.

Source

Speaking at Nvidia GTC2025, Meta’s Chief AI Scientist Yann LeCun declared that he thinks LLMs are reaching their limits, and that anyone looking to get into AI should “work on next-gen AI systems that lift the limitations of LLMs” and reflect how cognition actually works, with modular systems, symbolic processing and world models, rather than just adding more tokens.

Google CEO Sundar Pichai has reiterated these sentiments, stating that when it comes to LLM development, “the low-hanging fruit is gone… the hill is steeper.”

Whilst telcos are still optimistic about AI having value, they’re yet to demonstrate that the value is truly transformational. Instead, should they consider AI as normal technology, that isn’t always the right solution for all circumstances, and challenge its limitations until the risks are overcome – if they can be?

Satellite connectivity

Satellite-based networks have, in the last half-year alone, made significant strides as a complementary telecom asset. LEO (Low-Earth Orbit) technology is rapidly advancing and is on pace to solve coverage problems in many rural and underserved regions.

The last few months have seen several landmark launches and partnerships that expand coverage via LEO constellations.

T-Mobile launched a beta of its nationwide satellite-to-cellular service using SpaceX’s Starlink, with the full rollout of “T-Satellite” scheduled for October, whilst in March, Vodafone partnered with AST SpaceMobile to bring direct-to-device (D2D) services to Europe. Vodafone demonstrated this with the world’s first in-space mobile video call using AST’s satellite.

Not to be left behind, Amazon’s Project Kuiper finally got its first 27 satellites off the ground in April, with aims to put 1,663 satellites into service by mid-2026. However, the most recent launch has been delayed by rocket booster issues.

The EU’s own IRIS² project is also moving forwards: in early 2025, SES won a €1.8 billion contract to build 18 medium-Earth orbit (MEO) satellites for the IRIS² secure-communications network, though full service is not due until 2030.

Meanwhile, Reuters reports that China’s SpaceSail is in talks with Brazil and other countries to launch a total of 15,000 satellites by 2030.

Whilst Starlink remains the industry leader by a wide margin, the very public falling-out between Donald Trump and Elon Musk is putting this dominance to the test, with former advisor Steve Bannon going so far as to call for SpaceX to be nationalised.

Open Digital Architecture

Open Digital Architecture (ODA), the TM Forum standard for modular, API-driven BSS/OSS, continues to gain traction across the telco sector.

CSPs are recognising the need to dismantle legacy stacks and upgrade to cloud-native BSS/OSS solutions using ODA’s open APIs and microservices to make integrations and the launch of new offerings simpler.

The move to ODA is proving substantive. The fact that dozens of CSPs, with billions of subscribers, are officially Running on ODA shows more and more CSPs worldwide are delivering on the promise of simpler, more flexible operations. Verizon is now Running on ODA, awarded at DTW25 – Ignite, while stc Group became the first Running on ODA-certified operator in the Middle East.

TM Forum has now taken this one step further by launching ODA Component Certification, which enables CSPs to standardise interfaces and behaviours at a component level. This is designed to accelerate deployments, cut costs and foster a truly composable ecosystem that scales with market demands.

In January, Cerillion earned Ready for ODA status, underlining our commitment to open standards, and we are proud to be part of the ODA Component Conformance team which has been working on the new certification programme. We also recently took part in a TM Forum report examining the drivers for open, composable architectures in the face of complex legacy infrastructure.

Digital Engagement

Major telcos have expanded their mobile apps and web portals in 2025 to make their marketing far more personalised; AI-driven systems can now tailor promotions to each user’s profile and suggest plan upgrades and bundles based on user behaviour.

Klarna’s CEO – yes, him again – wants to turn the BNPL platform into a financial superapp that delivers personalised services – and yes, he wants AI to power it. Part of this is the launch of mobile phone plans through a partnership with Gigs, an MVNO start-up provider.

At MWC25, Whale Cloud and Alipay announced a partnership designed to empower enterprises across Africa, Europe, Latin America and the Middle East in building mobile wallet and superapp solutions to equip businesses with scalable and secure mobile payment experiences.

While progress varies by market, there are many examples of digital-transformation wins demonstrating that telcos are investing in richer digital experiences. For example, Verizon Business has launched an industry-first GenAI assistant for small businesses, giving SMEs “the same opportunity to leverage AI as global enterprises to automate and enhance their customer interactions.”

On the enterprise side, Vodafone launched its Managed IoT Connectivity Platform, a cloud portal giving businesses “complete visibility and control” of all their SIMs and devices. The portal offers customisable usage reports and real-time alerts to monitor connected assets.

However, many operators still have work to do in catching up with user expectations, especially when it comes to seamless, personalised engagement across multiple channels.

Customer centricity

CSPs are continuing to roll out new measures to put customer experience at the centre of their businesses.

In the US, AT&T and T-Mobile now allow on-demand plan changes via app, and multiple operators (Verizon included) have introduced bill-shield/price-lock guarantees to avoid surprise costs. EU regulators have spurred this on by mandating clearer roaming and contract terms, and telcos from Telecom Italia to Orange have implemented easy “pause/cancel” options for certain services.

Many companies are also now deploying virtual “assistants” that help customers autonomously resolve common billing or setup issues, upgrade plans or troubleshoot problems without human agents. In January, AT&T launched the AT&T Guarantee, an industry‑first promise to “set a new bar in customer experience” covering both its wireless and fibre services.

In June, T‑Mobile added self‑service SIM management to its T‑Life app: customers can now swap physical SIMs or eSIMs themselves through the app. These changes eliminate support calls and empower users, improving convenience and retention.

And Deutsche Telekom has started offering customers free access to Perplexity Pro through its MeinMagenta app:

Meanwhile loyalty and personalisation programmes are expanding, as CSPs offer tiered rewards, premium content bundles, and hyper-targeted offers aimed at improving customer satisfaction and retention.

As with all things, though, it’s a case of one step forward and two steps back, with the announcement that Meta aims to introduce adverts in WhatsApp. While undoubtedly a boon to advertisers, this marks a significant shift away from the platform’s original promise of a simple, ad-free messaging experience without distractions or commercial interruptions. The arrival of adverts threatens that core experience and could push users to seek out alternative messaging apps that better respect their attention and privacy.

Cybersecurity

In what is now considered the largest data breach in history, over 16 billion login credentials for Apple, Facebook and Google have been exposed.

BT has warned that only three in five SMEs have had cybersecurity training, mainly as a result of budget constraints or no dedicated team. In response, they have launched dedicated cybersecurity training to equip business owners and their teams with practical tools and knowledge to counter digital threats.

Meanwhile, in April, SK Telecom revealed that hackers infiltrated SKT’s Home Subscriber Server and stole the SIM-authentication data of approximately 25 million users. Although financial and personal data were not leaked, SKT offered free SIM replacements and launched enhanced fraud-protection services for all subscribers.

In response to these escalating threats, telecom operators worldwide are rapidly adopting AI-driven cybersecurity. Machine learning models are increasingly deployed to detect anomalies in real time, enabling faster threat identification and response. Simultaneously, there’s a widespread pivot towards zero-trust architectures, where no user or system is automatically trusted, even within the network perimeter.

From a regulatory perspective, Europe has taken a firm stance, with new EU legislation mandating end-to-end encryption, rapid incident disclosure, and adherence to an EU blueprint for cyber crisis management. With the scale and frequency of breaches intensifying, these legal frameworks are forcing major operators to be both proactive and transparent in their security roadmaps.