IoT And LPWAN, Explained: The Networks Built To Whisper

Ask ten engineers to define the Internet of Things and you will get ten slightly different answers, most of them correct. The term has been stretched to cover smart speakers, factory robots, soil sensors, pacemakers, parking meters and cargo containers. The label is so broad that it can seem to mean nothing at all.

It does mean something, though, and the definition gets clearer once you look at the least glamorous part of the stack: how a device with a small battery, sitting in a basement or a field, gets its data out. That question gave rise to a family of networks called LPWAN, short for low-power wide-area network. You rarely see them, but they carry a large share of the world’s quiet machine traffic.

This explainer covers both terms: what IoT is once you strip away the marketing, and what LPWAN is, how its main variants differ, and where each one fits.

The Short Answer, Before The Long One

The Internet of Things is the practice of giving physical objects the ability to sense something about the world and report it over a network, so that software can act on it. The “thing” might be a water meter, a shipping pallet, a vibration sensor on a pump, or a tractor.

LPWAN is a class of wireless networks designed for one specific kind of IoT device: one that sends small amounts of data, infrequently, over long distances, while running on a battery for years. It trades speed for range and battery life. That trade is the whole point.

What “Internet Of Things” Actually Means

The phrase is usually credited to Kevin Ashton, who used it in 1999 while working on RFID supply-chain tracking at Procter & Gamble. His idea was simple. Computers knew almost nothing about the physical world except what humans typed into them. If objects could report their own status, the data would be more complete, more timely and less prone to error.

A quarter of a century later, that idea still holds. What changed is the cost. Sensors, radios and microcontrollers have become cheap and power-efficient enough to put into objects worth a few dollars. Cloud platforms made it practical to collect data from millions of devices without building a data centre for each project.

A useful test is to ask whether an object’s main job is something other than computing. A laptop is a computer. A smartphone is a computer that happens to be portable. A connected gas meter is a meter first, and its connectivity exists only to serve that job. Industry analysts usually count IoT this way, leaving phones, tablets and PCs out of their device totals.

The Four Layers Of Every IoT System

Almost every IoT deployment, from a smart building to an oil field, has the same four layers.

The device. A sensor or actuator, a small processor, a power source and a radio. It measures temperature, location, vibration, flow or whatever else matters, and sometimes it acts, for example by closing a valve.

The connectivity. The network that carries data off the device. This might be Wi-Fi, Bluetooth, a cellular network, a wired industrial bus, a satellite link, or an LPWAN.

The platform. Software, usually in the cloud and increasingly at the edge, that receives messages, authenticates devices, stores data and manages firmware updates.

The application. Dashboards, alerts, analytics and integrations with business systems, where the data finally turns into a decision. Our recent profile of Datanomix shows this layer at work, turning raw machine data into decisions on the factory floor.

Most public discussion of IoT focuses on the first and last layers, the gadget and the dashboard. In practice, the connectivity layer is where many projects succeed or fail, because it sets the device’s battery life, its cost, where it can work, and how long it stays usable.

How Big IoT Has Become

The scale is harder to picture than to count. IoT Analytics tracked 18.5 billion connected IoT devices in 2024 and expected the figure to rise about 14 percent to 21.1 billion by the end of 2025. The firm’s latest State of IoT research projects roughly 39 billion devices by 2030.

Most of those devices do not use the same kind of network. Wi-Fi remains the largest single technology, at about 32 percent of IoT connections, which makes sense for homes, offices and retail. Cellular IoT, which includes 2G through 5G plus LTE-M and NB-IoT, accounts for about 22 percent of connections. Bluetooth covers a large share of wearables and short-range sensors.

LPWAN fills a gap none of those were designed for.

The Connectivity Problem Nobody Talks About

Picture a water utility that wants to read 500,000 meters remotely. The meters sit in underground pits, in basements and behind concrete walls. Most have no mains power. The utility wants each one to report once or twice a day and to run for a decade without a technician visiting to change a battery.

Wi-Fi does not work because there is no Wi-Fi in a meter pit and the range is measured in tens of metres. Bluetooth has the same range problem. Classic cellular, the 4G that powers your phone, would drain a small battery in months and costs too much per module and per subscription for a device that sends a few dozen bytes a day.

This is the gap LPWAN was built for: huge numbers of simple devices, spread over wide areas, often in hard-to-reach places, each sending tiny messages rarely.

So What Is LPWAN?

LPWAN is not one technology. It is a category defined by a set of trade-offs:

•           Long range, typically several kilometres in cities and considerably more in open countryside, from a single base station or gateway.

•           Low power, with devices that sleep most of the time and wake only to transmit, allowing multi-year battery life.

•           Low data rates, from a few hundred bits per second to a few hundred kilobits per second, far below what a phone uses.

•           Low cost, for both the radio module and the connectivity subscription.

•           High capacity, so that one gateway or cell can serve thousands of devices.

Most of these gains come from physics. Sending data slowly, over narrow slices of spectrum and often at sub-gigahertz frequencies, lets a receiver pick up very weak signals. That extends range and helps signals get through walls and underground. The price is throughput: you cannot stream video over LPWAN, and you would not want to.

Three families dominate the market today: LoRaWAN, Sigfox, and the two cellular standards, NB-IoT and LTE-M.

LoRaWAN: The Open Network Anyone Can Build

LoRaWAN runs in unlicensed sub-gigahertz spectrum, the same kind of band used by garage door openers and wireless doorbells. Its physical layer, LoRa, uses a chirp spread-spectrum modulation originally developed by Semtech. The network protocol above it, LoRaWAN, is an open specification maintained by the LoRa Alliance.

What sets LoRaWAN apart is who can run it. Because the spectrum is unlicensed, anyone can deploy a network: a telecom operator, a city, a utility, a farm, or a factory that wants coverage only inside its own fence. Devices send messages to gateways, which pass them to a network server, which forwards them to the application.

Adoption has kept climbing. The alliance passed 125 million deployed end devices at the end of 2025, citing a 25 percent compound annual growth rate across its ecosystem. This week, it said deployments had reached 150 million, meaning about 25 million devices were added since that earlier milestone. The mix has also widened beyond smart meters to include tracked pallets, smart padlocks, bicycles and employee safety badges.

That growth shifts the hard problems. Once networks hold millions of devices, onboarding, roaming, interoperability and lifecycle management become as important as radio coverage. The alliance’s recent work reflects this, including new specifications released in August 2026 to automate device setup.

The trade-offs are real. Unlicensed spectrum is shared, so interference is possible, and regional duty-cycle rules limit how often a device can transmit. Someone also has to own the network. A private LoRaWAN deployment gives full control but also full responsibility.

Sigfox: The Pioneer That Stumbled

Sigfox, founded in France in 2010, was the technology that made many people take LPWAN seriously. It uses ultra-narrowband radio in unlicensed spectrum, and it took a different business approach from LoRaWAN. Instead of letting anyone build a network, it ran or franchised one operator per country and sold connectivity as a service.

The design is deliberately minimal. Messages travel at 100 or 600 bits per second depending on the region, and payloads are famously tiny, just 12 bytes per uplink message. For a device that only needs to say “I am here, the temperature is 4°C, battery is fine,” that is enough.

The business struggled. Sigfox entered receivership in early 2022, and Singapore-based UnaBiz acquired the technology and French network that year. UnaBiz now treats Sigfox as a technology brand rather than a company name, marketing it as “0G.” As of December 2024, the network it manages supported more than 14 million active devices worldwide.

The episode taught buyers a lasting lesson. We covered one example in our profile of Sensolus, where DHL’s rolling-cage tracking programme began on Sigfox and later added NB-IoT trackers. A device meant to last ten years may outlive the network it was designed for.

NB-IoT And LTE-M: The Operators’ Answer

Mobile operators did not want to cede massive IoT to unlicensed networks. Through the standards body 3GPP, the industry defined two cellular LPWAN technologies, NB-IoT and LTE-M (also called Cat-M), and published them in 2016. Ericsson notes that both were introduced between 2015 and 2017 to connect very large numbers of low-cost, battery-powered devices over wide areas.

Both run in licensed spectrum on existing 4G networks, usually as a software upgrade to base stations. That gives them managed quality of service, carrier-grade security and the reach of national networks with roaming agreements.

NB-IoT (Narrowband IoT) is the more stripped-down of the two. It uses a very narrow slice of spectrum, offers data rates in the tens of kilobits per second, and is optimised for deep indoor coverage and stationary devices such as meters, alarms and parking sensors.

LTE-M offers more. It supports data rates up to around 1 Mbps, handles movement between cells the way a phone does, and can even carry voice. That makes it better suited to trackers, wearables and anything that travels.

Both rely on power-saving features that let devices sleep deeply for long periods between check-ins. These include extended discontinuous reception (eDRX) and Power Saving Mode.

Cellular IoT as a whole is large and still growing. The June 2026 Ericsson Mobility Report put total cellular IoT connections at around 4.5 billion at the end of 2025 and forecast they would approach 8 billion in 2031. A steady driver is the shutdown of older networks. Operators are retiring 2G and 3G to reuse that spectrum for 4G and 5G, and NB-IoT and LTE-M are positioned as their successors for low-power devices.

How The Options Compare In Practice

These are general characteristics, not guarantees. Real-world range and battery life depend on local terrain, building materials, how often a device reports and how much data it sends.

What LPWAN Is Not Good At

It is worth being clear about the limits, because vendors are not always eager to point them out.

LPWAN is a poor fit for anything that needs high bandwidth, such as video, audio or large file transfers. It is not built for low latency either, so a factory robot that must react within milliseconds belongs on wired industrial Ethernet, Wi-Fi or private 5G. We looked at those options in our guide to industrial IoT and industrial connectivity.

Connected cars are another example of where LPWAN is the wrong tool. Modern vehicles move large volumes of data and need full broadband cellular connections, a shift we traced in how IoT rewired the car.

Downlink, meaning messages sent to the device, is also limited on most LPWAN technologies, especially Sigfox and some LoRaWAN device classes. That makes firmware updates over the air slow or impractical in some deployments. Engineers have to plan for it from the start.

Where LPWAN Shows Up In The Real World

LPWAN earns its place where data is small, devices are many and batteries have to last. Typical uses include:

•           Utility metering: water, gas and heat meters reporting daily readings from underground or indoor locations.

•           Asset tracking: roll cages, pallets, containers, rental equipment and other items with no power source of their own.

•           Smart buildings: occupancy, air quality, temperature, leak detection and desk usage sensors.

•           Agriculture: soil moisture, weather stations, livestock location and irrigation control across large fields.

•           Smart cities: parking bays, waste bin fill levels, streetlight faults and environmental monitoring.

•           Safety: panic buttons, lone-worker alarms and smoke detectors.

What these share is a profile: one message every few minutes to once a day, a payload of a few bytes, and a device nobody wants to touch for years.

The Satellite Layer Arrives

Terrestrial networks, whether LoRaWAN gateways or cell towers, still leave large gaps over oceans, deserts, farmland and remote infrastructure. Satellites are increasingly filling them.

Some approaches pair LPWAN with space. Netmore, for instance, is combining terrestrial LoRaWAN with Kinéis satellite connectivity to reach beyond ground coverage. The 3GPP standards have also been extended so that NB-IoT and LTE-M can work over satellite links, known as non-terrestrial networks.

At the higher-bandwidth end, companies are building constellations that connect directly to ordinary phones and cellular devices. Our profile of AST SpaceMobile looks at one of the most ambitious efforts, a plan to put a cell tower in space. For IoT buyers, the practical result is that “no coverage” is becoming less of a hard limit, though satellite connectivity usually costs more per message and uses more power.

Choosing A Network That Outlives The Device

For organisations planning an IoT rollout, the hardest decision is often not which sensor to buy but which network the sensor will depend on for the next decade. A few questions help narrow the field.

How much data, how often? A few bytes a day points toward LPWAN. Anything involving images, audio or frequent updates points elsewhere.

Does the device move? Stationary devices work well on NB-IoT or LoRaWAN. Moving ones usually do better on LTE-M.

Who should own the network? A private LoRaWAN network gives control and predictable costs inside a defined site. An operator network removes the need to maintain infrastructure but adds a recurring subscription and a dependency on that operator.

What does coverage look like at the actual locations? Coverage maps are a starting point, not proof. Field testing in the real basements, pits and warehouses matters.

What happens if the network changes? The Sigfox experience and the ongoing 2G and 3G shutdowns show that networks can change within a device’s lifetime. Multi-network modules and a clear upgrade path reduce that risk.

A growing number of deployments no longer pick just one. LoRaWAN is increasingly deployed alongside cellular, Wi-Fi, Bluetooth and satellite, with each device assigned the network that suits its power, coverage and mobility needs.

What To Watch

Three developments will shape LPWAN over the next few years.

The first is how cellular IoT evolves in 5G and 6G. 5G RedCap, a lighter version of 5G for mid-range devices, has been launched commercially by 14 service providers, and an even lower-cost variant called eRedCap is expected to reach full commercial potential from 2028. Early 6G work is already reconsidering massive IoT. In a recent Ericsson analysis, the company argued that real deployments have shown traffic patterns that differ from what designers originally assumed.

The second is whether LoRaWAN can keep up its growth as it moves beyond utilities into logistics, buildings and safety, and whether easier onboarding makes very large deployments less labour-intensive.

The third is satellite. As low-orbit constellations add IoT services and standards bring satellite and terrestrial networks closer together, the line between “covered” and “not covered” will keep shifting.

None of this changes the basic picture. IoT is about turning physical objects into sources of data. For the many objects that have little to say and no power to spare, LPWAN remains the network that lets them say it.