The telecom industry is undergoing rapid transformation, and as we approach the new year, I’d like to offer some observations and predictions for 2026. It's an exciting time to be in this industry, and I believe 2026 will be a year defined by building the foundations for the next leap toward truly autonomous networks.

A step closer to autonomous networks

In 2026, while fully autonomous networks remain some way off, we will begin to see some major telecom providers mastering closed-loop automation in specific areas of the network. This will likely start in the transport layer first, as operators find it easier to manage than the highly distributed radio access network. While human oversight will still be needed for complex or high-risk decisions, many routine tasks will increasingly be handled autonomously.

However, widespread progress in automation across the entire network will face some challenges in the coming year, as operators tackle a tangled mess of legacy IT systems, coupled with the lack of reliable, real-time data necessary to feed the AI. Operators should treat next year as an opportunity to "clean house," investing heavily in modernizing infrastructure, taking stock of their data, and beginning to automate parts of the network.

From static pipes to dynamic AI-connectivity fabrics

Next year, pressures from hyperscalers and enterprises demanding AI-optimized connectivity will force communication service providers (CSPs) to evolve. They will need to transform their networks from static, fixed networks to instead become dynamic, AI-connectivity fabrics. In this new status quo, bandwidth will not be the only constraint.

CSPs will need to be able to roll out real-time orchestration across multidomain networks. The control plane will emerge as a major bottleneck as it will need to be intelligent enough to make real-time decisions on traffic routing, power usage, and where to place computing resources across domains. Overcoming this will be a challenge, but 2026 should mark the beginning of tangible progress.

The rise of the AI-native twin

In 2026, we will begin to see the emergence of AI-native twin technology playing a role in network operations. These will not be static, outdated maps, but real-time simulation engines. Before an AI agent executes a change on the live network, it will first run the action through the AI-native twin. This simulation will check for conflicts, ensure consistency across different parts of the network, and model potential negative consequences and outcomes for any given action.

A major catalyst for the wider adoption of these twins will be AI workload routing. As operators grapple with delivering reliable performance for huge AI applications running at the edge and in the cloud, they’ll need a way to guarantee that the network can keep up. Real-time simulation before deployment gives them the safety net and predictability necessary to make that possible.

That said, digital twin technology becoming standard practice for network planning is still a few years away, with mainstream adoption likely being around 2028. Most operators are only now putting in place the unified data foundations that twins require to be truly effective.

The shift from general AI to telco-native intelligence

2025 marked the shift toward using general-purpose large language models in telecom. In 2026, we’ll see a move to telco-specific AI models that actually understand network structure, performance patterns, and past incidents. These models will underpin AI-driven digital twins that act as real-time simulation engines, allowing operators and AI agents to test actions before they touch the live network. This will be a major step towards genuine multidomain automation.

But with this leap in autonomy comes a new and far more subtle security challenge – the risk of AI agent manipulation. If an attacker alters an agent’s goals or behavior, the system could make harmful changes while believing it’s operating normally. We need to prepare for scenarios where attackers hijack autonomous actions by altering agent intent or policy guardrails, causing the network to autonomously take destructive actions without knowing anything is wrong.

Another emerging risk is AI model interference, where attackers compromise training data or inject false telemetry, potentially triggering mass rerouting and self-induced outages. In 2026, for operators hoping to successfully introduce autonomous AI agents, securing the AI layer will become just as critical as securing the underlying network itself.

The edge computing challenge

As intelligence and compute increasingly push to the edge, orchestration platforms will need to evolve from centralized, service-centric systems to distributed, real-time, workload-aware control planes capable of spanning thousands of edge sites. This shift demands new capabilities, including distributed low-latency multidomain orchestration and AI-driven placement decisions. The centralized orchestration models that work today need to scale to manage the emerging distributed "telco cloud."

For operators, the year ahead will be about more than automation: 2026 will be about transforming static legacy systems into flexible architectures capable of evolving with AI demand, which will continue to grow. This will come with some challenges, but 2026 will mark a turning point that will be pivotal in ushering in the future of true autonomous network operations.