Michael De Nil is the CEO and cofounder at Morse Micro—reinventing Wi-Fi for IoT.
The first wave of connected infrastructure was built around a simple assumption: Devices would collect data, networks would transport it and most of the intelligence would live somewhere else, normally in the cloud.
Edge AI changes that architecture. Cameras, building systems, industrial equipment and energy assets are increasingly becoming intelligent endpoints capable of analyzing data and responding locally. As that happens, the network is no longer simply a transport layer. It becomes part of the system that allows distributed intelligence to function.
That means organizations need to evaluate connectivity differently. The question is no longer simply, “How fast is the network?” The more useful question is whether the network can support intelligence reliably across the environments where it will actually operate. That comes down to five requirements.
1. Range Has To Match The Physical Environment
Intelligent infrastructure is often distributed well beyond traditional indoor coverage. Cameras may sit at property edges. Smart meters and home batteries may be installed in garages or utility rooms. Industrial equipment may operate across large facilities. Energy devices may be located outdoors or behind multiple physical barriers.
In those environments, adding more access points, gateways or cabling can significantly increase cost and complexity. The network therefore has to be designed around where devices actually need to operate, not around where connectivity happens to be easiest to deploy.
2. Power Efficiency Becomes A System Requirement
Edge AI devices already have to budget power for their sensing activities and the local processing. If connectivity consumes too much of that budget, the overall system becomes more difficult to deploy and maintain.
This is especially important for battery-powered cameras, sensors, meters and other devices expected to operate for extended periods without frequent maintenance. The right question is not whether a wireless technology can deliver maximum throughput; it is whether it can deliver the performance the application needs without overwhelming the device’s energy budget. That makes power efficiency part of the architecture, not simply a component specification.
3. The Intelligent Edge Needs Meaningful Bandwidth
At the other end of the spectrum, many low-power networks were designed around simple telemetry: a meter reading, sensor value or periodic status message.
The intelligent edge increasingly moves richer information. A camera may transmit images, event metadata or diagnostics. Industrial equipment may send more detailed operating data. Devices may require software or model updates. Energy systems may need to coordinate multiple distributed assets.
That creates a middle ground between high-throughput conventional networking and narrowband sensor connectivity. Organizations may not need maximum bandwidth everywhere, but they increasingly need meaningful bandwidth, enough to support richer data and evolving workloads without abandoning power efficiency or range.
4. Security Has To Be Built Into The Connectivity Layer
As more intelligence moves into physical infrastructure, the consequences of a compromised endpoint increase. A simple sensor reporting temperature is one thing. A network of cameras, access systems, industrial equipment or energy assets making local decisions is another. Strong authentication, encryption, secure update mechanisms and established network-management practices therefore need to be considered from the beginning.
The same applies to network complexity. Adding proprietary gateways or isolated networking layers may solve a short-term coverage problem but can create additional systems for IT and operations teams to secure, integrate and maintain. The simpler the architecture, the easier it becomes to manage security consistently.
5. Interoperability Determines Whether The System Can Scale
The intelligent edge will not be built around one device type, one vendor or one networking technology. Buildings may combine security cameras, HVAC systems, occupancy sensors and energy equipment. Industrial environments may bring together machines, sensors and monitoring platforms. Utilities may need to connect meters, batteries, solar systems and EV infrastructure. Those systems need to work within broader IP-based architectures rather than becoming isolated connectivity islands.
Standards-based networking becomes increasingly important because it allows intelligent endpoints to integrate more naturally into existing infrastructure and management systems. Newer approaches such as Wi-Fi HaLow, based on the IEEE 802.11ah standard, illustrate how the Wi-Fi ecosystem itself is evolving toward this model by combining sub-GHz reach, lower power operation, IP networking and sufficient bandwidth for richer edge applications.
The broader point is not that one wireless technology will replace every other option. Traditional Wi-Fi will remain critical for high-throughput local networks. Cellular will continue to serve mobility and wide-area applications. Other low-power technologies will remain useful for simple sensing.
The shift is architectural. Organizations building Edge AI systems should stop evaluating connectivity around a single headline specification and instead design around the workload. Where will the device operate? How much power can it consume? What information must it exchange? How will it be secured? And how will thousands of endpoints fit into a larger network?
The intelligent edge is changing how computing is distributed across the physical world. Connectivity needs to evolve with it. For leaders designing the next generation of intelligent infrastructure, the network should no longer be treated as an afterthought. It should be designed as part of the intelligence itself.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?