IoT platform retired? How to choose the next one and migrate

If your IoT platform is being retired, you need to get three things out of it: the history, the fleet and the logic that runs it. Choose the next platform by how it lets you leave, not just by how it lets you in. And migrate in groups, with both platforms running in parallel until the data matches.

Why do IoT platforms get retired?

Because to the company selling them they are one product among many, and products get shut down. Between 2023 and 2026, several large cloud providers have retired managed IoT services. Their users have had to move their fleets with a shutdown date hanging over them.

The lesson isn’t that the cloud is bad. It’s that your devices and your data outlast the platform that manages them. A meter, a pumping station or an inverter is installed for many years; the service behind it can change its terms, change its price or disappear sooner.

So the useful question isn’t “which one has the most features?”, but “what happens to me if I have to leave it?”.

What does the law say about taking your data with you?

The EU Data Act is on your side. Since September 2025 it has required providers to make switching easier, and since September 2026, to give access to the data generated by new connected products.

That’s a good basis for negotiating, but it doesn’t replace the practical part: an export you’ve never tested isn’t an exit. Before you sign with a new platform, export some test data and check you can read it with another tool.

What should your next IoT platform have?

Choosing in a hurry because of a shutdown is the quickest way to end up on another platform that ties you down. This table sums up what to require and what to ask to check it:

There’s one more criterion that doesn’t fit in a table: the platform should be light to run. If keeping it alive needs someone dedicated to queues and clusters, you’ve swapped one dependency for another.

Why AI should travel inside the data rather than be bolted on afterwards is the subject of AI in the data path.

How do you migrate an IoT platform without losing data?

An orderly migration follows five steps. Order matters more than speed:

1. Inventory. List every device type, with its protocol and firmware; the rules and alarms; the integrations; the users; and who consumes the data downstream. Whatever isn’t in the inventory breaks on cutover day.
2. Export the history. Get it out in a standard format and check record counts and date ranges against the source. Do it early: leave it to the last few weeks and there’s no room to repeat an export that fails.
3. Run in parallel. Send the same data to both platforms for a while and compare readings, alarms and dashboards. If they don’t match, you still have time to fix it.
4. Cut over in groups. Start with a small, representative group, with a plan to roll back. Expand once the previous group has been stable for a while.
5. Decommission the old one. When the last group is running on the new platform, switch off the old integrations, revoke credentials and store the exported history wherever you decide.

If some devices speak a protocol the new platform doesn’t understand, an edge gateway handles the translation without touching the device.

What isn’t data has to move too

Rules, alarm thresholds, who gets each alert and through which channel, dashboards and the scripts someone wrote years ago are as much part of the platform as the history. They tend to be the last thing inventoried and the first thing missed. Treat them like code: write them down, review them and check them in parallel, just like the data.

What if you don’t need to switch?

You might not. If your platform isn’t being retired but has fallen short or become expensive, migrating doesn’t always pay off. We have our own platform, but you don’t have to replace yours: we add AI, edge, per-device security or offline operation to what you already have. Our services page explains how we improve existing AIoT platforms.

Switching pays off when the platform is going away, when its cost grows with every device or when it won’t let you take your data. If none of those apply, start by improving what already works.

Where does NovaeOps fit?

NovaeOps, our AIoT platform, is built to meet that list:

• It receives data over HTTP and MQTT, polls equipment over HTTP, ICMP and SNMP, and reads from databases such as PostgreSQL or MongoDB.
• It runs in our cloud, on your infrastructure or at the edge, with the same product. Edge mode runs on modest hardware, such as a Raspberry Pi 5, and on sites without a stable connection it keeps collecting data and firing alarms on its own.
• No per-device or per-user fees. AI is billed per delivered verdict: an idle model costs nothing.
• Your data exports to Parquet, CSV or JSON.
• Every reading can carry an AI verdict, provided by CogniX: the decision, how far to trust it, a warning if the data falls outside what it knows and which variables weighed in.
• Remote updates through campaigns, with rollout windows, and each device reports which phase it’s in.
• More than 180 operations through the command line and MCP, so your AI assistant can monitor and run the platform. The ones that delete anything ask for explicit confirmation.

NovaeOps is available for pilots, and we are looking for the first reference customers.

Behind it is a team with more than two decades in IoT platforms. In previous roles, it has kept fleets of more than a million devices running in production and replaced a hyperscaler’s edge runtime at a fifth of the total cost, working offline.

Where do you start?

With the inventory, even if the shutdown date feels a long way off. If you’d like to do it with us, we’ll show you NovaeOps with a case like yours.