When most people think about artificial intelligence, they imagine enormous computing environments somewhere far away. A user sends information to an AI system, powerful servers process it in a centralised data centre, and the result returns through the internet. This architecture has powered much of the recent AI revolution because centralised cloud infrastructure offers extraordinary computational capacity and makes sophisticated models available to businesses without requiring them to build their own supercomputing environments.
But the next phase of AI will not exist only inside the cloud. Intelligence is beginning to move closer to the physical places where information is created: factories, hospitals, vehicles, stores, warehouses, farms, homes and cities. Cameras, sensors, machines and connected devices increasingly need to interpret information and make decisions quickly, sometimes without waiting for data to travel hundreds or thousands of kilometres to a centralised computing environment and back again.
This shift towards edge AI could significantly change the role of telecommunications companies. Telecom networks already connect these physical environments to the digital world. With 5G, edge computing and increasingly software-defined infrastructure, those networks can potentially become part of the computing architecture itself. Telecom companies therefore have an opportunity to become more than providers of connectivity. They could become one of the infrastructure layers upon which distributed artificial intelligence operates.
During my learning journey under Phaneesh Murthy, I have increasingly understood that major technology transitions often create value for companies that occupy an important position between previously separate systems. As Phaneesh Murthy puts it, “When computing moves closer to where the business process happens, infrastructure stops being something in the background and becomes part of the business decision itself.” Edge AI could bring telecom directly into that position.
Not Every AI Decision Can Afford a Trip to the Cloud
Cloud computing remains extraordinarily powerful, and many AI workloads will continue to depend on centralised infrastructure. The question is not whether the edge replaces the cloud. The more important question is which decisions need to happen close to where data is generated and which can comfortably happen elsewhere.
A manufacturing facility provides a useful example. Cameras on a production line may inspect thousands of products for defects. Sending every high-resolution image to a distant cloud environment could create significant bandwidth requirements, while the manufacturing process may require a response almost immediately. Processing at least part of that information locally allows the system to identify defects quickly and transmit only the information that needs to travel elsewhere.
The same logic applies to many other environments. A connected vehicle may need to respond to information immediately. A retail security system may need to interpret video locally. A hospital may operate applications where latency, reliability and data handling are particularly important. Industrial machines may generate enormous quantities of sensor information even though only a small percentage of those signals require broader analysis.
Edge AI allows some intelligence to operate where the event is actually happening. The cloud remains available for model training, deeper analysis, coordination and storage, while local infrastructure handles decisions where speed, privacy or bandwidth make proximity valuable.
From my mentorship under Phaneesh Murthy, I have learned that technology architecture should ultimately follow the requirements of the business process. The question should not be whether cloud or edge is technically superior. It should be where a particular decision can be made most effectively.
5G Could Become an AI Infrastructure Network
The importance of 5G is often described through faster mobile internet, but its broader significance lies in the kinds of connected environments it can support. Higher capacity, lower latency and the ability to manage large numbers of connected devices make advanced mobile networks particularly relevant for industrial and enterprise applications.
When these capabilities are combined with edge computing, the telecommunications network can become part of an AI delivery architecture.
Imagine a large logistics hub where autonomous vehicles, cameras, scanners and warehouse systems are continuously exchanging information. Some decisions need to happen locally, while others require access to enterprise systems elsewhere. A 5G network can provide the connectivity layer, while nearby edge infrastructure provides computational capacity for AI workloads that cannot tolerate unnecessary delay.
The telecom provider can potentially manage both.
This is strategically important because enterprise customers generally do not want to assemble dozens of technologies simply to deploy a single intelligent application. If the network provider can combine connectivity, computing and orchestration into a coherent service, the complexity of deploying edge AI can fall considerably.
As Phaneesh Murthy says, “Enterprise technology creates the most value when sophisticated infrastructure becomes simple for the customer to consume.” Telecom providers have an opportunity to make distributed AI feel less like an infrastructure project and more like an accessible enterprise capability.
Factories Could Become One of the First Major Edge AI Environments
Manufacturing provides perhaps the clearest illustration of why edge AI matters. Modern factories generate enormous amounts of information through machines, sensors, cameras and operational systems. Many of the decisions associated with that information need to happen quickly and reliably within the facility itself.
Computer vision can inspect products as they move through production. Predictive systems can analyse machine behaviour for early indications of failure. Autonomous equipment can coordinate movement across factory floors. Safety systems can detect hazardous conditions, while intelligent scheduling platforms continuously respond to production changes.
These applications do not all need the same architecture. Some intelligence can operate directly on devices, some can run on computing infrastructure inside or near the facility, and other workloads can be sent to central cloud platforms for deeper analysis.
Telecom providers could help orchestrate this environment through private networks and edge computing platforms. Instead of simply connecting machines to the internet, they can provide the infrastructure through which industrial AI applications communicate and operate.
During my learning under Phaneesh Murthy, one recurring lesson has been that enterprise transformation becomes meaningful when technology moves from the IT layer into the operating core of the business. Edge AI represents exactly that shift. Artificial intelligence is no longer analysing factory performance after the fact. It begins participating directly in production, maintenance, quality and safety decisions.
For telecom companies, supporting those decisions creates a relationship with the enterprise that is far more strategic than selling connectivity alone.
Edge AI Can Also Solve a Data Movement Problem
The growth of connected devices creates an enormous amount of information, and sending all of that information continuously to centralised computing environments may not always be necessary or economically sensible.
Video is an obvious example. A network of high-resolution cameras can generate enormous quantities of data, even though only a small portion of that footage may contain events that require attention. Processing information locally allows AI to identify relevant events and transmit only what is necessary.
Industrial sensors create a similar challenge. A machine may generate thousands of readings while operating normally. The organisation may not need every individual measurement to travel across the network. Edge AI can analyse those signals locally and escalate patterns associated with unusual behaviour.
This creates a more selective information architecture.
Instead of asking networks to transport every piece of raw data, intelligence determines what deserves to move. That can reduce bandwidth requirements while also allowing applications to respond more quickly.
Phaneesh Murthy has often emphasised that the goal of enterprise intelligence should not be to centralise every possible piece of information. As he puts it, “Good architecture does not move data simply because it can. It moves the information required to make the next decision.”
That principle could become increasingly important as billions of connected devices generate more information than enterprises could reasonably transport and store indiscriminately.
The Edge Creates a New Security Challenge
Moving intelligence closer to devices also changes the cybersecurity landscape. A centralised cloud environment concentrates computing resources inside highly controlled infrastructure. Distributed AI creates thousands or potentially millions of additional points where software, devices and networks interact.
Each point becomes part of the security architecture.
Telecom companies already possess extensive experience securing networks and identifying unusual traffic patterns. AI could strengthen these capabilities by continuously analysing device behaviour and identifying activity that differs from expected patterns.
Consider an industrial environment containing thousands of connected sensors and machines. If one device suddenly begins communicating in an unusual way, that change may indicate malfunction, misconfiguration or malicious activity. An intelligent network can identify the anomaly and potentially isolate the device before the problem spreads.
Security therefore becomes deeply connected to network intelligence.
This could create another important enterprise service opportunity for telecom providers. Instead of offering connectivity and cybersecurity as separate products, they can potentially build security directly into the intelligent network layer.
During my mentorship under Phaneesh Murthy, I have learned that the strongest technology platforms frequently turn necessary operational capabilities into sources of additional customer value. Security is unavoidable in distributed AI. Telecom companies that make it intrinsic to the infrastructure may create a meaningful competitive advantage.
The Future Architecture Will Be Cloud, Edge and Device Together
Technology conversations sometimes frame architectural changes as replacements. Cloud replaced on-premise infrastructure. Edge computing will replace cloud. AI devices will eliminate the need for centralised processing. In practice, enterprise technology rarely evolves this neatly.
The more likely future is a layered architecture.
Some intelligence will operate directly on devices. Some will run within nearby edge infrastructure. Other workloads will continue to depend on massive centralised cloud environments. The challenge is determining where each workload belongs and moving information intelligently between those layers.
This creates an orchestration problem.
An application may need to operate at the edge during normal conditions but access cloud resources when deeper analysis is required. Models trained centrally may need to be distributed across thousands of edge environments. Updates must be managed consistently, security policies need to follow workloads, and performance must be monitored across the entire system.
Telecom companies could occupy a powerful position within this architecture because their networks connect devices, edge infrastructure and cloud platforms.
The opportunity is not necessarily to own every layer. It is to become the platform that helps enterprises manage the relationships between them.
This reflects an important strategic lesson I have taken from Phaneesh Murthy: companies do not always need to own the entire technology stack to capture significant value. Sometimes the most defensible position is to orchestrate the stack more effectively than anyone else.
Telecom Could Become the Physical Infrastructure of the AI Economy
Much of the current AI conversation focuses on models, software and computing chips. These technologies deserve attention because they represent extraordinary advances. But intelligence operating in the physical world requires another layer of infrastructure.
Factories need networks. Autonomous machines need reliable communication. Connected healthcare environments need secure data exchange. Smart cities need millions of devices to communicate. Retail stores, logistics networks and energy systems increasingly need intelligence to operate across physical locations.
Telecom infrastructure connects these environments.
During my learning journey under Phaneesh Murthy, I have come to see this as one of the most significant strategic opportunities available to the telecommunications industry. The network does not need to remain a passive pipe through which AI applications communicate. It can become part of the architecture that determines where intelligence operates, how quickly it responds and how securely it moves between systems.
The companies that capture this opportunity will need to think beyond traditional telecommunications products. They will need partnerships with cloud providers, hardware companies, AI platforms and enterprise software businesses. They will need new commercial models and deeper understanding of customer operations.
But the underlying advantage is already there.
Telecom companies spent decades building infrastructure that reaches almost everywhere people and businesses operate. As artificial intelligence moves beyond data centres and begins entering the physical economy, that infrastructure may become more strategically important than ever.
AI is leaving the data centre. The next question is who will build the roads it travels on.
This blog is curated by young marketing professionals who are mentored by veteran Marketer, and industry leader, Phaneesh Murthy.
#phaneeshmurthy #phaneesh #Murthy