At the moment, that word is AI, and the sector is clearly growing weary of it. AI is undeniably creating real impact, but its relentless overuse has become a little exhausting.
At recent events such as MWC Barcelona, DTW and ITW, AI-themed demos have dominated the show floors. Vendors have frequently promised “self-healing” networks, “cognitive fabrics” and “AI-native RAN.”
Every PowerPoint slide declared a revolution, but peel away the gloss and most deployments look a lot like something telecoms has been doing for decades: automating processes and crunching data.
AI has become a label of convenience, a marketing umbrella covering everything from predictive maintenance to customer chatbots. In many cases it’s little more than a rebrand of analytics tools that pre-date ChatGPT.
And operators are noticing. As one European CTO told Capacity privately: “We don’t have an AI strategy. We have a slide deck.”
The fatigue is real
The problem isn’t that AI has no value. It’s that most of the time its value is marginal, not monumental, despite what several vendors claim. Predictive analytics does help prevent outages. Machine-learning algorithms can fine-tune radio parameters. But none of this justifies the idea that telecoms is entering an “AI age.” After years of hype, executives are asking harder questions: Where’s the ROI?
A 2024 TM Forum survey found that while 71% of telecom leaders have an AI roadmap, only 22% have achieved measurable business impact. The rest remain trapped in the proof-of-concept phase, where pilots multiply but budgets stall. Implementation costs are rising, the skills gap persists, and few vendors can demonstrate performance gains that justify the marketing spend.
This fatigue mirrors what happened with cloud transformation a decade ago, when every operator claimed to be “becoming a platform company,” but most remained access providers with slightly better billing software. The same dynamic is unfolding now. “AI” has become the badge of innovation, not the substance.
Incremental, not transformative
Look closely at the main use cases across global carriers and the pattern is clear. Network optimisation? Automated resource allocation refined by machine learning. Fraud detection? Statistical modelling with better branding. Customer-service chatbots? Natural-language wrappers around call-deflection systems that existed for years. These are improvements, not paradigm shifts.
In wholesale and interconnect, AI’s footprint is even smaller. Some routing engines now use predictive algorithms to balance capacity, and there’s growing interest in AI-assisted traffic forecasting. But this remains marginal compared with the transformative language used at several industry conferences.
Behind closed doors, many engineering teams are quietly rolling their eyes. “AI-driven” is increasingly seen as AI-washed, a way for vendors to justify higher price tags or distract from lack of differentiation.
Structural barriers
Telecoms also suffers from structural inertia. Operators are conservative by design: regulated, asset-heavy and risk-averse. They run on legacy OSS/BSS stacks that can’t easily be rebuilt to integrate new data models.
True AI transformation would require a rewrite of how networks are orchestrated, how services are billed and how decisions are made, and that’s a multi-year, multi-billion-dollar proposition few CFOs will approve.
Meanwhile, the real AI value chain is forming elsewhere: in the cloud hyperscalers. Amazon Web Services, Google Cloud and Microsoft Azure own the compute infrastructure, data pipelines and foundation models that make serious AI work possible. Telcos talk about “AI at the edge,” but they’re often just customers of someone else’s platform. The power balance hasn’t really shifted; it’s consolidated.
The backlash begins
All of this is fuelling a quiet backlash, and executives are tired of vendor decks that promise “autonomous networks” without evidence. Engineers want measurable KPIs, not buzzwords. Even analysts are adjusting their tone: Gartner’s Hype Cycle for Telecom Operations 2024 notes that AI-driven network automation is sliding toward the trough of disillusionment as expectations outpace implementation reality.
That doesn’t mean AI has no future in telecoms. It simply means we’re entering a phase of AI sobriety, where the rhetoric cools, the investment narrows, and only the genuinely useful applications survive. Expect to hear less about “AI-first” and more about “AI-useful.”
Where AI might actually deliver
If AI is to justify the hype, it will need to solve problems that matter such as energy optimisation in power-hungry 5G networks, real-time orchestration across non-terrestrial and terrestrial links, and dynamic capacity management as fibre and edge networks scale. These are areas where learning systems could deliver real efficiency gains.
Until then, the industry might do well to pause the hype cycle and get back to basics. AI isn’t magic, it’s more like math. And no amount of branding can turn incremental progress into a revolution.
RELATED STORIES
AI now lies, denies, and plots: OpenAI’s o1 model caught attempting self-replication
Datacloud Energy 2026
After a standout 2025 edition, we’re back with an even sharper focus on the intersection of data centres, energy, and ESG. As power demand rises and regulations evolve, there’s a growing urgency to rethink how infrastructure is powered, financed, and built for long-term impact.
