From Moving Bits to Moving Intelligence: Turning Telcos into AI Opcos with Intelligent Infrastructure

At MWC26 Barcelona, where Snowflake served as the Intelligent Infrastructure theme sponsor, one idea came through clearly in conversations across the industry: telecom operators are entering a new phase of transformation.

For two decades, telecom operators engineered networks to do one thing exceptionally well: move bits with extreme reliability. That operating model was built for scarcity — scarce spectrum, scarce interoperability, scarce coverage — and it worked.

But the economics shifted. Network traffic continues to rise, while pricing power remains constrained in many mature markets. Ericsson reports total monthly global mobile network data traffic reached 188 exabytes in Q3 2025, growing 20% year-over-year (Q3 2024 to Q3 2025), and expects video to represent 76% of mobile data traffic by the end of 2025. Connect Europe (ETNO) notes EU telecom retail revenues grew only 0.7% in 2021 and 2.1% in 2022 — increases more than offset by inflation in those years.

That divergence is the strategic wedge: demand keeps compounding, but value capture keeps moving up the stack.

At MWC 2026, the central theme of many conversations was intelligent infrastructure. I joined an executive roundtable on “Connected Intelligence: Building the Intelligent Infrastructure for the AI Era,” because the next telecom wave is not primarily about moving more bits faster. It is about moving — and operationalizing — intelligence.

Why intelligence is becoming infrastructure

In a world of agentic AI, robotics, and always-on inference, intelligence behaves less like a software add-on and more like infrastructure. It must execute under latency and determinism constraints. It must respect local regulation and data residency. It must be anchored in identity, trust, and policy. And it must integrate with the physical world — devices, vehicles, sensors, and industrial systems.

That combination makes intelligence distribution look like a telecom problem, except the payload is not voice packets or internet traffic — it is decisions.

This is the conceptual leap behind an AI Opco: a network operator engineered to run intelligence as a governed, metered, and reliable operational workload, rather than as a feature bolted onto connectivity.

What telcos often underestimate is that the hardest part is not “having data.” It is combining telco data and network capabilities to create governed, consumable, and billable products that developers and enterprises can adopt at scale. In my view, that is the real purpose of intelligent infrastructure.

What intelligent infrastructure really means

If you strip away the buzzwords, intelligent infrastructure is a shift from network operations to network intelligence. It is what happens when cloud, edge, and network architectures become one coherent fabric capable of supporting AI at scale. It is the ability to orchestrate near real-time analytics and automation across hybrid environments, so the network can move from reactive operations to predictive — and then increasingly autonomous — decisioning.

In practice, it shows up as measurable operational outcomes: predictive maintenance, anomaly detection, AI-assisted capacity planning, automated provisioning, self-optimizing systems, and experience-aware service personalization. It also shows up as ecosystem convergence, with telecom’s role extending beyond connectivity into powering IoT, automotive, and enterprise ecosystems built on shared data intelligence.

A quick proof that foundations beat AI bolt-ons

One instructive parallel comes from Capita’s work modernizing contact center operations. Rather than “adding AI” on top of fragmented systems, the focus was on unifying operational data, establishing a governed source of truth, enabling near real-time analytics, and standardizing a repeatable pattern that could be deployed quickly and consistently. Ultimately, this turned what previously took weeks or months into deployments that could be stood up in days.

Why Snowflake is foundational to the AI Opco journey

Snowflake is the AI Data Cloud. For telecom leaders, this means Snowflake can be the control plane that connects network, cloud, and ecosystem data, applies AI where the data lives, and makes it practical to productize trusted network signals as intelligent infrastructure.

Here are the Snowflake capabilities that matter most for turning telcos into AI Opcos:

1. Governed data products. AI Opcos need network truth as data products with contracts — schema, lineage, access controls, privacy policies, and service levels — across RAN and core telemetry, OSS alarms, BSS entitlements and charging, device identity signals, and consent. Snowflake supports open architectures like Apache Iceberg tables, combining AI Data Cloud performance capabilities with external cloud storage you manage.
2. Consumption economics that match how intelligence is bought. Monetization shifts from gigabytes to capability consumption, including API calls, inference, dataset subscriptions, quality-on-demand reservations, and policy checks. Snowflake warehouses are billed only while running, with per-second billing after a one-minute minimum.
3. Privacy-first collaboration without data sprawl. With Secure Data Sharing, no actual data is copied or transferred between accounts. Snowflake Data Clean Rooms enable privacy-preserving collaboration where providers control what analyses are allowed. Data providers can also use privacy-enhancing techniques such as differential privacy. This is the foundation for cross-industry monetization that respects privacy without revealing personally identifiable information.
4. AI inside the governance boundary. Snowflake’s AI features are designed to operate within its security and governance perimeter, with role-based access controls and explicit customer-data handling principles. That matters in telecom, where trust and compliance are built into the product and not added on as an afterthought.
5. Distribution rails for platform scale. A platform business does not scale on bespoke integrations. Snowflake Marketplace and Native Apps allow operators to package data products, analytics applications, and decision services, and distribute them to partners and customers with far less friction than traditional project-based delivery.

The tangible path: how telcos become AI Opcos

If you want an AI Opco outcome, you must build AI Opco infrastructure. Here is the roadmap I recommend – designed to deliver monetizable products early while building toward platform scale.

Step 1: Define the AI Opco product surface that is mapped to tangible business outcomes

Start with capabilities that have clear willingness-to-pay and a defensible telco advantage. These include fraud and identity signals (SIM swap risk, number verification), location verification, quality-on-demand guarantees, network experience indices, and edge inference placement signals.

Step 2: Build a single governed network intelligence data plane and accelerate the time to extract actionable intelligence

The goal is not to create “one big data lake.” Instead, it’s to create one governance and productization plane across telemetry, OSS/BSS, identity, and consent. This includes standardized tiering (raw to curated to productized), SLAs, lineage, and purpose-limited access controls.

Step 3: Build the telecom knowledge plane and establish privacy-by-design monetization primitives

Create a repeatable, reliable and deterministic ontology-based knowledge infrastructure for AI and Agents. Build with secure sharing and clean rooms from day one. Codify anonymization and aggregation rules, allowed analyses, consent handling, and auditability so monetization scales safely.

Step 4: Build the AI factory inside the governance boundary

Operationalize feature engineering, model management, LLM workflows, monitoring, and audit so intelligence is produced continuously, not as one-off experiments.

Step 5: Create data and AI products

Turn internal analytics into external products such as mobility insights, network experience indices, device risk scores, IoT fleet health, edge suitability maps. Package them with documentation, sample queries, and “how to use this” patterns.

Step 6: Expose intelligence through APIs developers can adopt

A platform is defined by exposure. Anchor in APIs accepted widely by the industry. Standardized REST API mechanisms as well as CAMARA APIs with their Fall 2024 meta-release delivering 25 APIs (including SIM Swap, Number Verification, Edge Discovery, and Quality on Demand) could be considered. GSMA Open Gateway positions a global framework of common network APIs designed to provide developers and cloud providers with universal access to operator networks. Use Snowflake as the governed data and AI platform powering these APIs – such as risk scoring, aggregation, contextual decisioning – and run the service layer where it needs to run.

Step 7: Commercialize like a consumption business

Price per call, per event, per dataset, per inference bundle, and per SLA tier (latency class, locality class, trust tier). Align incentives to activation and retention, not one-time integration fees.

Step 8: Scale via ecosystem distribution

Shift from “selling to enterprises” to “powering ecosystems”: banks, retailers, OEMs, mobility platforms, data providers, and governments embedding telco capabilities into their products. Use privacy-safe collaboration to unlock cross-industry value without compromising trust.

The future: telcos that grow because of the intelligence they deliver

PwC projects telecom industry revenues will rise at a CAGR of only 2.9% through 2028. If operators want platform-like growth, it will come from programmable capability: trusted identity signals, quality-on-demand services, edge-aware intelligence, and governed data collaboration.

In the AI era, telcos have a choice. Remain connectivity pipes optimized for cost per bit or become AI Opcos running intelligent infrastructure, where connectivity becomes capability and the network becomes a trusted fabric for data-driven decisions. 

My conviction is that the operators who win will look more like platform companies than utilities. They will monetize “network truth” as products. They will measure growth in adoption and consumption, not only ARPU. And they will grow because of the intelligence they deliver - reliably, locally, and with telecom-grade trust. 

Snowflake is not the whole AI Opco story. Telcos will still need edge compute, orchestration, and network modernization. Snowflake serves as the AI Data Cloud foundation that makes these investments viable – without a fully managed data platform, network signals remain raw data; with it, they become governed, high-value products. It is the only environment capable of bridging privacy-preserving collaboration with secure, in-place AI, providing the essential distribution rails to transition from a traditional carrier to a scalable platform business. 

The question for telecom leaders going into 2026 is not whether AI will reshape networks. It is whether you are building the intelligent infrastructure to capture the value it creates.