When AI Meets Telecommunications: Balancing Adoption With Control

Photo Credit: Canva

Nvidia’s recent announcement that it is developing a radio unit chip for future AI-native networks has generated considerable discussions within the telecommunications industry. The bulk of the focus is on AI-RAN, 6G architectures, and it means for conventional network vendors.

Nvidia’s 2026 State of AI in Telecommunications survey reported that 89% of operators plan to increase AI-related spending within the year, while 77% believe AI-native networks will arrive before 6G. The same survey found that operators are already seeing benefits from AI in areas such as customer service, software development, business automation and network operations.

By most measures, telecommunications operators are not resisting AI. In fact, they are actively investing in it. Yet the industry’s enthusiasm becomes noticeably more nuanced when discussions move from business applications towards operationally critical functions, and this distinction may explain why the future of AI in telecommunications is likely to be more complex than many technology narratives suggest.

Telcos Have Never Been Strangers to Automation

Telcos are often portrayed as conservative adopters of new technology. In reality, they're historically amongst the earliest adopters of automation, analytics, virtualisation, cloud-native architectures, and AI. The motivation is more economics than technological curiousity.

Faced with growing traffic volumes, rising customer expectations and persistent pressure on margins, telcos have long sought technologies that improve efficiency, optimise operations and reduce the cost to serve. At the scale at which they operate, even modest productivity improvements can translate into significant financial outcomes.

This explains why many operators have already deployed AI across customer service, software development, workforce productivity, network operations, and cybersecurity.

Equally, automation within the network itself is not new. Technologies such as self-optimising networks, dynamic bandwidth allocation, traffic engineering, predictive maintenance and AI-assisted DDoS mitigation have existed in various forms for years. In many cases, these systems are making decisions at a speed and scale that would be impossible for humans to manage manually.

The issue, therefore, is what happens as that trust reaches its limits.

Critical Infrastructure Operates Under Different Rules

Telecommunications networks operate under a very different set of constraints than most enterprise environments. Decisions made today often remain embedded within operational environments for years, while failures can have consequences that extend far beyond individual customers or business units.

Large-scale infrastructure environments are often more interconnected than what their network diagrams document. Over time, systems become integrated, dependencies accumulate and institutional knowledge becomes fragmented. In some cases, the true importance of a system only becomes apparent when it fails. And this reality directly shapes how operators evaluate technology.

Telcos are responsible for infrastructure that supports Critical Information Infrastructure (CII), essential public services, mission-critical applications and, in some cases, national communications capabilities. Reliability, resilience and security are therefore not simply operational objectives but binding obligations.

A customer service chatbot may be assessed primarily on productivity gains and customer outcomes. An AI-powered vulnerability assessment tool may be evaluated on its ability to improve cyber defence capabilities. A network architecture decision, however, must satisfy requirements around reliability, interoperability, operational complexity, cybersecurity, energy consumption and long-term supportability. The closer AI moves towards operationally critical functions, the higher the burden of proof becomes.

Operators need assurance that they can maintain sufficient visibility, control and accountability while AI becomes increasingly embedded within infrastructure environments.

And the industry is not sitting still with discussions focusing on "telco-grade AI". Earlier this year, GSMA launched the Open Telco AI, an industry initiative aimed at developing AI models, datasets, and evaluation frameworks specifically developed for telecommunications.

Control Requires More Than Human Oversight

One of the most overlooked aspects of telecommunications operations is the extent to which they are built around observability. Telcos have spent decades building environments where network behaviour can be monitored, measured and understood. When something goes wrong, engineers trace events, identify root causes and understand why a particular event occurred. Visibility is fundamental to operating large-scale networks.

As AI becomes more deeply embedded within operational workflows and management processes, maintaining that same level of visibility is important.

The challenge goes beyond human oversight. It extends to ensuring that automation remains observable, governable and accountable as decision-making becomes increasingly distributed and autonomous.

AI Is Becoming Part of Cyber Defence

The discussion becomes even more relevant when viewed through the lens of cybersecurity. Governments and critical infrastructure operators are increasingly exploring how AI can strengthen threat detection, vulnerability assessment, anomaly detection and incident response capabilities. In Singapore, the Cyber Security Agency (CSA) has highlighted both the opportunities AI presents for cyber defence and the importance of securing AI systems themselves.

This reflects a broader reality. Organisations responsible for critical infrastructure must leverage on AI to defend against AI-enabled threats. As AI becomes part of the defence stack, operators must consider not only what AI can do, but also what new dependencies are being introduced in the process.

Why Nvidia’s Investment in Nokia Matters

This context can help explain why Nvidia’s US$1 billion investment in Nokia is particularly significant.

Much of the technology industry tends to frame innovation as a story of disruption, where new entrants replace incumbents, and new platforms displace established ecosystems. Telecommunications has rarely evolved that way. The industry’s complexity, regulatory obligations and operational realities often favour evolution and integration, over disruption.

Viewed through that lens, Nvidia’s investment is interesting from an industry positioning standpoint. At approximately 2.9%, Nvidia is clearly not seeking control of Nokia. Rather, the investment reflects a belief that the future of telecommunications and AI infrastructure will become increasingly intertwined.

More importantly, it may signal a recognition that technical capability alone is not sufficient.

If AI adoption within telecommunications were purely a technology problem, Nvidia could simply continue selling GPUs, accelerated computing platforms, and AI software to operators. Instead, it chose to deepen its relationship with one of the world’s largest telecommunications hardware manufacturer. That decision suggests the next phase of AI adoption is going to be focused on operational integration.

Telecommunications remains an industry built on carrier-grade reliability, interoperability, lifecycle management and operational trust. As AI moves closer to operationally critical functions, operators are likely to place equal importance on how solutions are integrated, governed and operated as they do on the underlying technology itself.

Nokia brings decades of telecommunications expertise, integration capabilities and operational credibility. Nvidia brings leadership in AI infrastructure, accelerated computing and software ecosystems.

One side understands AI. The other understands how to operate critical infrastructure at scale.

As AI becomes increasingly embedded within infrastructure that supports Critical Information Infrastructure, essential public services, mission-critical applications and national communications capabilities, the companies most likely to succeed may be those capable of bridging the gap between innovation and operational trust.

Balancing Adoption With Control

The right question is how operators balance adoption with control. How can they benefit from increasingly powerful AI capabilities while maintaining the observability, governance and accountability required to operate critical infrastructure? How can they ensure that automation strengthens resilience rather than inadvertently undermining it?

These questions go beyond the realm of technology and into the operational, governance and strategic realities of running a telecommunications network.

As AI moves closer to functions that support Critical Information Infrastructure, essential public services, mission-critical applications and national communications capabilities, the conversation is no longer simply about innovation.

It is about ensuring that innovation does not come at the expense of control.

In an industry built on reliability, resilience and trust, that may prove to be the most important balancing act of all.