Onomondo IoT Connectivity: The Limbo Problem

The dark part of every deployment

Switch on a cellular device for the first time in a place it has never been, a meter in a Romanian substation, a tracker in the steel hold of a ship, a sensor bolted to an irrigation pivot in a field with no road to it, and you have asked it to do something quietly difficult. It has no map. It cannot see the networks around it. It has to find one, attach, prove who it is, hold on long enough to say something useful, and then repeat all of that tomorrow, in the next country, after the next handover, for years, with no one ever touching it again.

A share of them always fails. And when a cellular IoT device fails, it usually does so in silence. No alert reaches anyone. The device is simply dark, and someone has to work out why, often with no view at all of what happened out on the network. That silence is the part of the industry the brochures skip over. Putting a SIM in a device is trivial. Delivering a service that stays up, across borders, at scale, for the working life of the product, is a different discipline, and it is the one where deployments quietly haemorrhage time and money.

Attached, but not connected

The clearest example is a fault every connectivity engineer knows. A device can be attached to a network, its SIM shown as connected on the dashboard, and yet pass no data at all. The registration is live but the data session behind it is dead. To anything watching the status field, the device looks fine. In reality it is stuck, and it will stay stuck until something forces the connection to re-establish: a reboot out in the field, or a reset of the session from the network side. That round trip, someone noticing, someone dispatched, someone power-cycling a box on a mast or a pump, is pure downtime.

Until it clears, the device sits in limbo, in the exact sense of the word: neither cleanly offline, where a monitoring system would raise an alarm, nor actually online. A surprising amount of IoT downtime lives in that gap between "shows connected" and "is connected", and it is invisible to anyone who can only see the status light rather than the traffic behind it.

Why global cellular fights you

Cellular networks were built for phones, which is to say they were built on the assumption that a person is present to notice a problem and fix it. An IoT device has no such person. It cannot read a message asking it to re-register. It cannot be told to switch flight mode off and on again. It has to survive alone in an environment designed around a completely different user.

That environment is fragmented by design. There is no single global network, only a patchwork of hundreds of operators stitched together by roaming agreements that were never meant for millions of unattended devices sitting still in one spot for a decade. Several familiar failure modes fall straight out of this. Steering can push a device onto a commercially preferred network that happens to carry the weakest signal exactly where the device is. An operator can drop a network onto a device's forbidden list and lock it out without a word. A growing number of countries restrict permanent roaming, so a device that never goes home can find itself deregistered. And beneath all of it sits the black box: with a conventional SIM, when something goes wrong inside another operator's network, you often cannot see it. You raise a ticket and you wait.

The industry has partial answers. Multi-IMSI SIMs carry several operator identities and hop between them. eSIM, and the newer SGP.32 standard written specifically for IoT, let a device be reprovisioned with a different operator profile over the air. These are real advances. But they refine the same underlying deal. You are still a guest on other operators' networks, still subject to their steering, still largely blind once your traffic leaves your sight. They make the guest better behaved. They do not make you the host.

A studio called Limbo

In 2010 a small Copenhagen studio released a game called Limbo. It is monochrome and almost wordless. You control a nameless boy who wakes alone in a dark, unmapped world and has to cross it. The world is full of traps, and the traps stay invisible until the instant they kill you. There is no dialogue, no instruction, no map. You move forward into the dark and try to survive what you cannot see.

It is hard to describe the life of a cellular device in a hostile roaming environment more exactly than that. A small thing, alone, crossing an unmapped world, meeting hazards it never sees coming, failing without warning and without explanation.

The rhyme runs deeper than mood. The studio that made Limbo, Playdead, spent its early life under a different name entirely: the entity was registered as Limbo ApS from 2008 until it was renamed Playdead in 2012.^[1] And one of the two people who went on to found Onomondo, Michael Freundt Karlsen, was Playdead's chief financial officer before he did.^[2] The reading of that we find hard to resist, and offer as our own rather than anything the company has claimed, is this: a man who once kept the books at the studio behind Limbo helped build a company whose whole reason to exist is getting a small, lone device safely through a dark and unmapped world.

What Onomondo IoT connectivity does differently

Onomondo was founded in Copenhagen in 2012, originally under the name Hello World Mobile, by Karlsen and Henrik Aagaard.^[3] Both came out of Tel42, a Danish network wholesaler, so they understood carrier infrastructure before they built anything, and its answer to everything above is to stop being a guest. Instead of reselling access to other operators' networks, Onomondo built and runs its own core: a software-defined cellular core designed from the start for machines rather than handsets. By the company's own figures that core reaches more than 600 carrier networks across over 180 countries at the radio level, and presents all of them to the device as a single network with a single identity.^[3]

Three things follow, and each one lands on a hazard named earlier.

One identity, no steering

Because the device carries one global identity rather than a wallet of operator profiles, there is no steering to fight and no forbidden-network trap to fall into. Onomondo describes its connectivity as non-steered: a device attaches to the strongest network available wherever it happens to be, rather than being herded towards a commercially preferred one. The problem of a device sitting on a weak partner network by design largely dissolves, because there is no preferred partner to be herded towards.

Less in the device

Because the intelligence sits in the network rather than the device, the hardware side gets simpler. Onomondo offers a cloud-downloadable software SIM, its SoftSIM, alongside conventional SIM, eSIM and eUICC form factors, and supports the SGP.32 IoT eSIM standard. The selection logic a multi-IMSI applet would normally carry is handled centrally instead. Less to go wrong in the field, and less to certify into every product.

Eyes on the network

And because Onomondo owns the core, it can show you what is happening inside it. This is the part that answers both the black box and the limbo state directly. The platform exposes live traffic monitoring and signalling logs, described by the company as a Wireshark-like view of how a device is behaving on the network.^[4] The attached-but-not-connected fault stops being invisible: an engineer can see a live registration with a dead session and reset it from the network, rather than dispatching someone to power-cycle the device or opening a carrier ticket and waiting for an answer that may never come. In a field defined by silent failure, turning the lights on is not a minor feature. It is close to the whole point.

The technical picture

Underneath the marketing, the architecture is the story. Onomondo operates a full telecoms core with direct integrations into the radio networks it uses, rather than buying wholesale access and layering a management portal on top. Global reach is carried over dedicated links, including a Tier 1 IPX partnership with Syniverse, which is how a single identity can appear local across so many countries.^[4] One SIM, one APN, one profile, with the complexity absorbed by the core instead of the device.

Where SGP.32 fits

SGP.32 is the GSMA architecture for eSIM in IoT, written to solve a specific problem: the consumer eSIM standard, SGP.22, assumes a person with a phone screen to scan a code and confirm a profile. An unattended sensor has neither a screen nor a user. SGP.32 removes the human from the loop by introducing two new elements, the eIM (eSIM IoT Manager), which orchestrates profile operations remotely and at fleet scale, and the IPA (IoT Profile Assistant), which carries them out on or near the device. The practical effect for a fleet operator is the ability to change which operator a deployed device uses, over the air, without visiting it or swapping hardware.

It is worth being precise about status, because the standard is still moving. SGP.32 v1.3 was published in May 2026, while v1.2 remains the baseline for certification, so support and certification are not the same thing and should not be read as such. Onomondo supports SGP.32 alongside its other form factors; we describe that as supported rather than asserting any particular certification state. For the full mechanics of SGP.32, the eIM and remote provisioning, see our dedicated coverage at sgp32.co.uk.

Proof at sea: the Maersk fleet

The clearest demonstration of what owning the core makes possible is not a slide deck. It is a shipping fleet. Under the name OneWireless, Onomondo and Maersk have built what the two companies call the world's largest private LTE network at sea, rolling out across around 450 Maersk vessels, roughly 100 of them time-chartered rather than owned outright. Onomondo's part is the piece that matters most for this story: a full software-defined LTE core, an Evolved Packet Core, installed and running locally aboard each ship.^[5]

Start with what it replaced, because that is where the case for the model begins. The fleet had been running on ageing 2G infrastructure. For something like refrigerated container monitoring, 2G means infrequent, low-resolution updates, which means poor visibility into exactly the cargo that most needs watching. A reefer full of pharmaceuticals or produce can drift out of temperature range for hours before anyone ashore has any way of knowing. The upgrade to LTE, with NB-IoT and LTE-M alongside it for the low-power sensors, lifts the data granularity sharply while keeping the power budget that thousands of continuously running sensors demand.^[6]

What makes the deployment more than a routine access-network upgrade is that each vessel effectively hosts its own private network. The core is on the ship. That is the detail that carries the resilience. Because the network intelligence is local rather than reaching back to shore for every decision, a vessel keeps a working private LTE environment even when its satellite link drops, which mid-ocean it routinely does. The blind spots that have long limited maritime IoT close up, because the network on board does not depend on a link to the outside world to keep functioning.^[5]

Then comes the part that proves the flexibility this whole profile rests on. Onomondo says devices move between the ship's private LTE network and public terrestrial networks automatically, with no SIM swaps and no manual reconfiguration, as a vessel leaves open water and comes into port. The same identity that worked on the private core at sea works on public networks on land. In Onomondo's framing this makes it, in effect, a maritime mobile operator in its own right, able to accept inbound roaming from other providers and support a bring-your-own-device model at sea.^[5] Set aside the labels and look at what is being described: a single connectivity identity that runs on infrastructure the provider owns and operates when that is the right answer, and rides other people's networks when that is. That is not a trick you can pull as a reseller. It is a direct consequence of owning the core, and it is the flexible-network argument of this piece made physical on 450 moving ships.

The payoff lands with the customer. Through Captain Peter, Maersk's cargo-visibility platform, its customers can now see a container's temperature, humidity, location and even its internal oxygen and carbon dioxide levels in close to real time, whether the box is mid-Atlantic or on a truck in Argentina. Maersk frames the value not as monitoring for its own sake but as earlier warning: the sooner a customer sees something drifting out of range, the sooner they can act on it.^[6] That is the difference between finding out a shipment spoiled and stopping it from spoiling.

Two things keep this honest. First, OneWireless is Maersk's platform and a multi-vendor build, not an Onomondo solo act. Onomondo provides the private LTE core, Nokia supplies the radio side, and 42com Sat, Complea and ZEDEDA fill out the wider stack.^[5] Onomondo is central to it, and is listed first in the companies' own materials, but it is one part of a larger system, and it would be wrong to imply otherwise. Second, this is a live programme rather than a finished monument: the rollout began in late 2024 and has run through 2026, with installations slotted into scheduled port calls to avoid taking ships out of service, and was described as approaching completion in early 2026.^[6] The scale is real, the architecture is real, and it is still being switched on.

Beyond the fleet

Maersk is the flagship, but it is not the whole story, and the model shows up at very different scales. Onomondo's named customers span connected beer dispensers for Carlsberg, shared pay-as-you-go bikes for Donkey Republic, and logistics for PostNord, alongside enterprise names such as Bosch and MAN Energy Solutions.^[7] A beer tap in a bar and a reefer container on the North Atlantic have almost nothing in common as devices, but they present the same problem to whoever has to keep them connected: a thing in the field, unattended, that has to stay reachable wherever it is. The consistency is the point. The same central core and single identity that lets Maersk hand a container off between sea and shore is what lets a drinks brand or a bike operator run a fleet of small devices across borders without rebuilding their connectivity logic for each one.

Owning the core, or renting it

It is worth being plain about where Onomondo sits and what its model costs, because the flattering version helps no one.

On the map of the IoT connectivity market, Onomondo belongs in the cloud-native layer: providers that run their own software-defined core and reach the world's radio networks through it, rather than acting as a channel for an incumbent's platform. That puts it directly opposite the resell-led end of the market, the MVNO-in-a-box propositions where a company can have its brand on connectivity within weeks and never touch a core at all. Both models answer the same question, how do I connect devices everywhere, with opposite architectures. One paints its name on the truck. The other built the road.

Building the road carries trade-offs a buyer should weigh honestly. A company running its own core is a smaller entity than the operators whose networks it reaches, so scale and longevity are fair questions, even with backers like Verdane and Maersk Growth behind it. Coverage in any given country still rests on which radio partners the provider has integrated there, so the single-network experience is only ever as good as the agreements beneath it. The regulatory limits on permanent roaming that trouble every roaming-based model do not vanish because the core is owned. And some buyers will reasonably prefer the multi-vendor flexibility of an enabler or an eSIM-led route, or direct operator relationships where data residency or regulation demands them. Onomondo's proposition is coherent and, on the evidence of the deployments, real. It is not the right shape for every project, and an independent read should say so.

So, out of limbo?

Every connected device begins the same way. It wakes somewhere, finds a network, and sends its first signal out into a world it cannot yet see, the machine equivalent of the first line every programmer ever writes, a hello addressed to the world. It is a nice accident of history that the company now called Onomondo began, in 2012, under a name that captured exactly that moment: Hello World Mobile.

The whole difficulty of this industry lives in what happens between that first hello and the ten-thousandth: across borders, through handovers, over years, unattended, in the dark. Onomondo's wager is that you do not solve it by being a better-behaved guest on someone else's network. You solve it by owning the ground the device stands on, and above all by refusing to let any of it stay dark. Whether that is worth the cost of building a core is a judgement each buyer has to make. And whether it truly guides IoT connectivity out of limbo, rather than into a smaller and better-lit version of it, is a question the deployments will keep answering. But as an argument about where the hard problem in IoT actually lives, it is one of the clearer ones on offer.

A note on how we tell these stories

At IoT Portal we spend nearly as long on how to explain something as on the thing itself. Some of the most consequential shifts in IoT connectivity, SGP.32, the eSIM, remote provisioning, are also the hardest to make anyone feel, because they live in acronyms and specification documents and rarely in pictures. A reader can be told what remote provisioning does ten times and still not hold why it matters.

The Onomondo story gave us a rare way in. Behind it sits the image from Limbo: a small figure, alone in the dark, crossing a hostile and unmapped world, trying to survive hazards it cannot see. That is not decoration. It is close to the most accurate picture we have found of what a SIM actually does, the perilous journey a single connection makes across borders and networks to deliver on the promise of a connected world. Hold that image and the abstractions turn concrete. Steering stops being jargon and becomes the choice of which dark path the device is pushed down. SGP.32 stops being a spec number and becomes the thing that decides whether a device stranded on a dead network can be handed a new identity over the air or simply left in the dark. Owning the core becomes what it means to switch the lights on in a world built to keep them off.

We will keep looking for images like this, because the industry has no shortage of facts and a real shortage of ways to hold them. The picture is the way in. For the machinery behind it, what SGP.32 and the eIM actually do, how remote provisioning really works, and where the limits still sit, see our dedicated SGP.32 coverage at sgp32.co.uk, and our wider explainer on IoT connectivity as a service.

Frequently asked questions