Telecom's AI Infrastructure Shift

Mobile World Congress 2026 made it unequivocally clear: artificial intelligence isn't just coming for telecom; it's already here. The industry is no longer debating the 'if' but grappling with the 'who' and 'how' of building the intelligent infrastructure required to harness AI's value. This year's dominant conversation revolved around this shift, moving from simply operating networks to operating intelligence itself.

Across keynotes and executive sessions, three themes converged. First, the 5G and fiber buildout continues, but it's now a foundation for AI. Second, AI is evolving from a feature into an operating layer embedded at the core of network functions. Finally, trust—encompassing fraud prevention, policy, sovereignty, and security—is now paramount in telecom strategy.

Intelligence as the New Infrastructure

For decades, telecom's core business has been reliably moving data. While data volumes have exploded, revenue growth has stagnated, forcing a strategic pivot. The next wave for telecom isn't about moving more bits faster, but about moving, governing, and operationalizing intelligence. This evolution positions companies like Snowflake, which sponsored the Intelligent Infrastructure track at MWC, as key enablers.

Snowflake's role in sponsoring the Intelligent Infrastructure track at MWC highlights their position in helping operators transform data into revenue streams beyond simple growth. They are building the governed data foundations that autonomy demands, embedding intelligence directly into network operations.

When data from networks, customers, and operations is unified and governed, it unlocks predictive optimization, enhanced customer experiences, and new revenue streams like fraud prevention and programmable network capabilities. This transforms telcos into trusted intelligence platforms.

From Network Operations to Network Intelligence

The shift from reactive operations to predictive, autonomous systems was a consistent executive-level discussion. Intelligent infrastructure represents a fundamental move from managing networks to leveraging network intelligence. It’s the convergence of cloud, edge, and network architectures into a cohesive fabric capable of supporting AI at scale.

This requires a unified, governed data foundation across all domains—network, service, customer, and ecosystem. A semantic, or knowledge, layer is crucial for AI agents to understand telecom context across disparate systems and vendors. Crucially, this architecture must span structured and unstructured data across hybrid environments without introducing new data silos.

Fragmentation in data catalogs, tooling, and access is a major bottleneck to AI readiness. Modern capabilities, however, yield tangible benefits: reduced incident resolution times, improved network resilience, and faster time-to-market. They also open new revenue avenues through services like fraud detection, location verification, and network APIs.

Operators often underestimate the challenge of transforming raw telecom data into governed, consumable products for developers and enterprises. Autonomy hinges on building the right data, governance, and architectural foundations first. The sequence is critical: unify data, establish trust and context, then automate.

Connecting Intelligence to Business Outcomes

At MWC, the focus was pragmatic: AI must drive measurable business outcomes. The conversation has moved beyond usage metrics to operational KPIs like time-to-insight, service quality, rollout velocity, and automated action rates. Monetization is shifting from gigabytes to capability consumption—API calls, inference services, data subscriptions, and more.

Telecom estates are inherently complex, with legacy systems and fragmented data. Unifying data across domains reduces friction, accelerates decision-making, and streamlines operations, lowering costs. For many operators, the qualitative impact is profound: a single team reported that with Snowflake, "you enable someone and there is silence after. Everything just works."

The most direct path involves connecting high-value structured data, adding unstructured context, and then layering agentic workflows. This AI-native data engineering approach turns modernization from a lengthy integration project into a scalable operating model. Operational simplicity is key, as removing friction enables step-function improvements.

Proactive monitoring, predictive optimization, and automated remediation demand a unified, governed view of network truth. As networks gain autonomy, the goal is a reliable, deterministic knowledge infrastructure for AI and agents, transforming the network from reactive infrastructure into an intelligent decision fabric.