The Telco Edge: Will it be different this time?

From 5G Hype to Real Demand – Forces that could finally make the Telco Edge matter

By Shane McClelland, Vice President, Emerging RAN Products, Ericsson

Note: This blog was produced under WIA’s Innovation and Technology Council (ITC). The ITC is the forum for forecasting the future of the wireless industry. Participants explore developments in the wider wireless industry, from 5G network monetization trends and streamlining infrastructure deployment to future spectrum needs and cell site power issues. These views are not a WIA endorsement of a particular company, product, policy or technology.

For years, the telecom industry has been promising the Telco Edge was coming. Mobile Edge Computing (MEC) was going to revolutionize how applications were built and deployed. 5G was going to unlock a wave of ultra-low latency use cases that would drive enterprises and developers to abandon centralized cloud architectures and move compute closer to where data was generated.

It didn’t really happen. Hyperscalers got faster and expanded their offerings. Developers stayed on AWS. And the “killer app” for the telco edge never quite materialized.

But, a confluence of new technologies – AI, Integrated Sensing and Communication (ISAC), and 6G – may finally create the pull that 5G MEC alone never could. To understand why, it helps to first understand the landscape of where compute actually lives today.

The Compute Continuum: Five Tiers of Processing

Modern computing doesn’t happen in one place. It exists across a spectrum – a Compute Continuum – that stretches from massive hyperscale data centers down to the device in your hand. Understanding this spectrum is essential to understanding where the telco edge fits, and why it matters.

Tier 1: Hyperscale Cloud

This is the world of AWS, Microsoft Azure, and Google Cloud. Operating at a multi-state or national scale, these platforms serve the broadest range of workloads – AI training, SaaS applications, large-scale analytics, media streaming, and government systems. Mainly workloads that don’t require real-time responsiveness. The market here is enormous and it represents the gravitational center of enterprise IT today.

Tier 2: Metro / Regional Edge

As requirements tighten, compute moved closer to population centers. City and metro-scale deployments – from AWS Local Zones to Azure Edge Zones to Google Distributed Cloud. This tier serves applications like gaming platforms, financial trading systems, media production, and regional enterprise applications. The market is smaller, but the use cases are more latency-sensitive and increasingly important. Players like Equinix have built significant businesses here.

Tier 3: Carrier / Telco Edge (AKA MEC)

This is where the telecom industry has long staked its claim. It’s tied directly to telecom network physical locations – central offices, aggregation hubs, cable headends, and cell sites. The use cases here are the ones that 5G promised to enable: Vehicle-to-Everything (V2X) communication, real-time video analytics, 5G IoT, and Smart City infrastructure. This is the tier that has been discussed, debated, piloted, and – if we’re honest – largely under-deployed for the better part of a decade.

Tier 4: On-Premise / Campus Edge

Moving inside enterprise and campus environments, Tier 4 brings compute directly into the operational environment. This is where agentic AI in enterprise operations is beginning to take hold – AI that is no longer just doing back-office analytics but is embedded in real-time operational systems. Manufacturing floors, hospitals, logistics hubs, stadiums, and airports are all candidates. The shift here is significant: AI is moving from training and analysis to real-time inference, embedded directly in the environments where decisions need to be made in milliseconds.

Tier 5: Far Edge / Device Edge

At the furthest point of the continuum sits the device itself – sensors, cameras, robots, vehicles, and endpoints of every kind. Processing here is highly distributed, highly constrained, and increasingly capable as AI silicon and software matures. This tier is less about hosting workloads and more about enabling the final few milliseconds of action.

So … Why Didn’t the Telco Edge Take Off the First Time?

From my perspective, Tier 1 and Tier 2 compute capabilities are good enough for most things, and the applications that genuinely required Tier 3 latency – true sub-10ms, network-integrated compute – were niche at best. Enterprises experimenting with 5G private networks often found that on-premise solutions (Tier 4) solved their problems without needing to rely on carrier infrastructure. Developers didn’t want to be locked into individual carrier footprints when they could deploy globally on a hyperscaler.

MEC, for all its promise, was a solution looking for a sufficiently demanding problem.

That may be about to change.

New Demand Drivers: AI, ISAC, and the Path to 6G

RAN-Native AI: Inference at the Radio

The integration of AI directly into the Radio Access Network is creating entirely new pressures on where compute needs to live. As AI moves from centralized training pipelines into real-time inference embedded in network operations, the latency and data gravity of that AI shifts dramatically. AI workloads that once lived comfortably in Tier 1 are being pulled toward Tier 3 as the intelligence needs to act on network conditions in real time – optimizing spectrum, predicting interference, managing handoffs, and enabling sub-millisecond decisions that simply cannot round-trip to a hyperscale data center.

Physical AI: Bridging the physical world

Physical AI systems represent a compelling driver for edge compute deployment at cell sites and telco locations. Unlike cloud-based AI applications that can tolerate higher latency, humanoid robots and other physical AI systems require near real-time processing to execute precise physical movements, react to dynamic environments, and ensure safe human-robot interaction. The latency demands of these systems make it impractical to route compute requests back to centralized cloud data centers, as even milliseconds of delay can translate into failed grasps, unstable locomotion, or safety-critical errors.

ISAC: Every Radio Becomes a Sensor

Perhaps the most underappreciated development on the horizon is Integrated Sensing and Communication (ISAC). In current networks, radios do one thing: they communicate. ISAC changes that fundamentally – every radio node in the network becomes simultaneously a communication device and a sensing device, generating continuous, real-time data about the physical world.

Think about what that means at scale. Thousands of radio nodes across a city, each sensing motion, presence, environmental conditions, and physical-world dynamics. The data volumes are enormous. The latency requirements can be extreme. And the data cannot realistically be backhauled to a Tier 1 cloud for processing – by the time it gets there, the moment has passed.

ISAC creates a genuine, structural demand for Tier 3 edge compute. It is not a use case that can be served from a centralized data center. It requires network-native AI – processing that lives at or near the radio, that can act on sensing data in real time, and that feeds back into network decisions with sub-millisecond latency. Digital twin applications, smart city infrastructure, autonomous systems, and industrial automation all become dramatically more capable when sensing is continuous and real-time.

The advent of 6G

Sixth-generation networks are still on the horizon, but their architectural direction is becoming clearer. 6G is expected to bring a significant boost in uplink and downlink performance, lower latency, and densification of radio nodes, pushing network infrastructure deeper into indoor and urban environments. 6G embeds compute as a native network function and the telco edge becomes the primary AI inference layer. 6G puts pressure on the Compute Continuum to process that data where it’s generated.

Why the Telco Edge May Finally Come Into Its Own

The narrative around the telco edge has always been about latency. And latency matters – but it wasn’t, by itself, a sufficient forcing function. What’s different now is that the nature of the data itself is changing. The combination of 6G, Sensing, and AI inference creates a technology stack that is fundamentally “edge-first” in a way that 5G never quite was.

ISAC data is continuous, physical-world sensing data from every radio node in the network. It cannot wait for a round-trip to the cloud. RAN AI inference needs to act on network conditions before those conditions change. Agentic AI operating in complex, dense, latency-sensitive environments – manufacturing plants, logistics hubs, smart buildings, transportation networks – needs compute that is physically close to the operational environment.

These emerging use cases aren’t just latency-sensitive in the traditional sense. They are data-gravity-sensitive. The data is generated at the edge, the decisions need to be made at the edge, and the actions need to be taken at the edge. Centralizing that workflow isn’t just slow – it may be architecturally impossible at the scale and speed these applications require.

For telecom operators and tower companies, this represents a genuine strategic opportunity. The physical infrastructure they have already deployed – cell sites, central offices, aggregation hubs – sits exactly where compute will need to be. Physical real estate, cabinets, power systems, thermal management, backhaul – can be repurposed and revamped to host and enable edge AI workloads alongside radio functions. The physical infrastructure is largely already there. The question is whether the use cases and business models can catch up.