An IoT analytics platform collects data from connected devices, cleans and stores it, and turns it into dashboards, alerts and predictions. The useful ones do this both in the cloud and at the edge, so that time-critical decisions such as stopping a failing pump happen next to the machine.
Updated October 2026
What changed in 2026
- Azure IoT Central is closing. New IoT Central apps could not be created from 23 September 2026; existing apps end on 20 September 2029. Microsoft recommends IoT Hub plus Microsoft Fabric Real-Time Intelligence for analytics (Microsoft Learn).
- Industrial suites added 3D twins. At Hannover Messe in April 2026, ABB added NVIDIA Omniverse 3D views to its Genix IoT and AI suite on Azure (ABB).
- More analytics moved to the edge. IDC reports $265 billion of edge computing spending in 2025, heading to about $450 billion by 2029 (IDC).
- Device security reporting became law. From 11 September 2026 the EU Cyber Resilience Act requires manufacturers to report actively exploited vulnerabilities within 24 hours (McCann FitzGerald).
What is IoT analytics?
IoT analytics is the analysis of time-series data from sensors and machines. It runs at four levels:
| Level | Question it answers | Example |
|---|---|---|
| Descriptive | What happened? | Energy use per line, per shift |
| Diagnostic | Why did it happen? | Temperature spike before each stoppage |
| Predictive | What will happen? | Bearing failure expected in 12 days |
| Prescriptive | What should we do? | Schedule maintenance in Tuesday's planned stop |
The 4-layer architecture
- Device layer: sensors, PLCs and gateways.
- Connectivity layer: protocols such as MQTT, OPC UA, CoAP, AMQP and HTTP move data reliably.
- Data layer: ingestion, stream processing and a time-series store, at the edge and in the cloud.
- Application layer: dashboards, alerts, ML models and integration with ERP, CMMS and MES.
Which protocol to use
| Protocol | Best for |
|---|---|
| MQTT | Lightweight publish/subscribe telemetry; the default for most IoT |
| OPC UA | Industrial machines and PLCs, with rich data models |
| CoAP | Very constrained devices on lossy networks |
| AMQP | Reliable enterprise messaging |
| HTTP/REST | Simple, low-frequency device updates |
Types of IoT analytics platform
| Type | Examples | Choose it when |
|---|---|---|
| Hyperscaler IoT services | AWS IoT Core and SiteWise, Azure IoT Hub and IoT Operations | You are already on that cloud and want managed services |
| Industrial IoT platforms | Cumulocity, PTC ThingWorx, Siemens Insights Hub | You need OT connectivity and asset models out of the box |
| Open-source | ThingsBoard, Eclipse projects, InfluxDB + Grafana | You want control and low licence cost and have engineers to run it |
| Edge AI platforms | NexaStack | You need models and decisions running on-site across many locations |
Migration note: if you still run workloads on Google Cloud IoT Core (retired 2023) or AWS IoT Analytics (end of support), plan a move to a supported ingestion and storage service now.
How to choose: a checklist
- Connects to your existing PLCs and protocols without custom code.
- Processes data at the edge when the network is slow or down.
- Stores time-series data at the volume and retention you need, at a known cost.
- Runs ML models and lets you update them across sites.
- Has role-based access, audit logs and encryption in transit and at rest.
- Exports data openly, so you are not locked in.
- Integrates with your maintenance and ERP systems.
Where AI changes IoT analytics
Dashboards tell people what happened. AI models and agents act on it: they predict failures, flag quality drift and open maintenance tickets. Running those models at the edge cuts latency and data transfer costs. NexaStack deploys and monitors AI models across edge sites, and connects them to your existing IoT data. See edge computing platforms, digital twin technology and NexaStack Edge AI.
Where it's heading: 2026 → 2030
| When | What it looks like |
|---|---|
| Around 2020 | All-in-one cloud IoT platforms (Google Cloud IoT Core, AWS IoT Analytics, Azure IoT Central) with dashboards and threshold alerts. |
| 2026 (now) | Those services are retired or closing; teams assemble ingestion (IoT Hub, MQTT), streaming analytics (such as Fabric Real-Time Intelligence) and edge processing, and add AI models and agents on top. |
| By 2028 (expected) | Gartner expects more than two-thirds of enterprise-managed data to be created and processed outside the data centre or cloud (Gartner via ZEDEDA). |
| By 2030 (expected) | Azure IoT Central apps stop on 20 September 2029, and Gartner expects physical AI to assist 80% of frontline workers at international companies by 2030, so analytics shifts from reporting to triggering actions (Microsoft Learn, Gartner). |
Frequently asked questions
What is IoT analytics?
The analysis of data from connected devices to describe, diagnose, predict and act on what is happening in equipment and environments.
What are the top IoT platforms?
There is no single top 10. Hyperscaler services (AWS, Azure), industrial platforms (Cumulocity, ThingWorx) and open-source tools (ThingsBoard) each fit different needs. Choose by your protocols, edge needs and team skills.
What is the best platform for IoT?
The one that connects to your devices with the least custom work, processes data where you need it, and fits your cloud and budget.
Is Google Cloud IoT Core still available?
No. Google retired it in August 2023.
What are the latest IoT analytics trends in 2026?
All-in-one IoT platforms are being retired in favour of composable stacks (IoT Hub plus streaming analytics), more processing happens at the edge, and analytics increasingly feeds AI agents that act rather than dashboards alone.
Turn IoT data into actions at the edge
Already collecting sensor data but stuck at dashboards? We will review your devices, protocols and sites and show where AI models and agents can act on that data. Book a 30-minute IoT analytics review →
