An IoT platform is the software that connects, identifies, updates and secures your devices and moves their data to the applications that use it. In 2026, choose one on five tests: device identity and provisioning, safe over-the-air updates at fleet scale, an edge runtime that works offline, security that meets new regulation, and how easily it feeds AI models running at the edge.
Updated October 2026
Most IoT programmes don't fail on sensors. They fail at device 5,000, when firmware updates brick units in the field, certificates expire unnoticed, and nobody knows which devices still run vulnerable code. Each failure means truck rolls, downtime and, increasingly, regulatory exposure. The platform you choose decides whether that happens.
Who this is for: CTOs, IoT product owners and OT/IT leaders in manufacturing, utilities, smart buildings, logistics and connected-product companies choosing or replacing an IoT platform.
What changed in 2026
- AWS ends support for IoT Greengrass V1 on 1 June 2026 and recommends V2 (AWS IoT Greengrass), forcing many fleets to re-evaluate their edge runtime.
- Microsoft's edge direction centres on Azure IoT Operations: an MQTT broker, OPC UA and Akri connectors and data flows on Arc-enabled Kubernetes, with up to 72 hours offline operation.
- The EU Cyber Resilience Act entered into force on 12 November 2024 and applies fully from 11 December 2027; it expects automatic security updates by default for connected products. Device management is now a compliance capability.
- AI moved onto devices and gateways, so platforms must deliver models as well as firmware.
Looking for analytics rather than device operations? Read our IoT analytics platform guide. This post covers platform selection and device management.
The five capabilities that matter
| Capability | What good looks like | Red flag |
|---|---|---|
| Identity and provisioning | X.509 certificates, zero-touch bulk provisioning | Shared keys, manual onboarding |
| OTA updates | Staged rollouts, rate control, automatic rollback | All-at-once pushes |
| Fleet visibility | Search by attribute, health dashboards, alerts | Spreadsheets of device IDs |
| Edge runtime | Local processing, offline buffering, containers | Cloud-only logic |
| AI readiness | Deploys and monitors models next to firmware | Models handled by a separate manual process |
Reference architecture: an IoT platform with edge AI
From device to decision
Sensors, PLCs, cameras
MQTT, OPC UA, local buffer
Run, monitor and govern models and agents on site
Identity, OTA, fleet health
MES, CMMS, dashboards
Figure: Devices connect through an edge gateway; AI runs on site; the IoT hub manages identity, updates and fleet health; business apps consume events.
AWS vs Azure vs open source: a fair comparison
| Option | Strengths | Watch for |
|---|---|---|
| AWS IoT Core + Device Management + Greengrass V2 | Fleet search, jobs and OTA with rollout control, deep AWS integration | V1 migration effort; AWS-centric |
| Azure IoT Hub + IoT Operations | X.509 identity, Device Provisioning Service, scale to millions of devices, industrial connectors | Arc and Kubernetes skills needed |
| Open source (e.g. ThingsBoard, KubeEdge) | Control, no per-device cloud fees, on-prem friendly | You own upgrades, security and support |
Many enterprises use a hyperscaler hub for identity and OTA, and a separate edge AI layer for models, so they can move between clouds or run air-gapped sites.
Choosing or replacing a platform? Get a free IoT platform fit assessment covering device management, edge AI and CRA readiness. Book an assessment.
How to start in 30 days
- Week 1: Inventory device types, firmware versions, connectivity and who updates what today.
- Week 2: Score two or three platforms against the five capabilities above, using your hardest device class.
- Week 3: Pilot on 20–50 devices: provision with certificates, run a staged OTA with forced rollback, deploy one edge model.
- Week 4: Measure update success rate, time to onboard, and operator effort; map gaps against CRA obligations.
Where it's heading: 2026 → 2030
| Year | IoT platforms |
|---|---|
| 2020 (history) | Cloud hubs collect telemetry; dashboards are the main output. |
| 2026 (now) | Kubernetes-based edge runtimes; runtime migrations; regulation drives update discipline. |
| 2027 (law) | EU Cyber Resilience Act fully applies from 11 December 2027. |
| 2030 (our view) | Devices host AI agents; platforms manage firmware, models and policies as one release. |
Where NexaStack fits
NexaStack is not a device hub. It is the AI layer that works with your IoT platform: it deploys, runs, monitors and governs models and agents on edge nodes, private cloud and air-gapped sites, with centralised orchestration across thousands of nodes. See the Edge AI solution, our industrial IoT applications guide and predictive maintenance with IoT.
FAQ
What is an IoT platform?
Software that connects devices, manages their identity and updates, and routes their data to applications, often with an edge runtime for local processing.
Is AWS an IoT platform?
AWS offers a set of IoT services, including IoT Core, IoT Device Management and IoT Greengrass, which together form a platform. Azure offers IoT Hub and IoT Operations.
What are the four types of IoT platforms?
Common groupings are connectivity platforms, device management platforms, application enablement platforms and analytics platforms. Most enterprise products now span several.
Is it risky to switch IoT platforms mid-programme?
It is, but staying on an unsupported runtime is riskier. Reduce risk by moving device identity first, piloting OTA on a small group, and keeping AI models portable.
What are the latest IoT platform trends in 2026?
Runtime migrations such as Greengrass V1 to V2, Kubernetes at the edge, Cyber Resilience Act preparation and AI models managed alongside firmware.
Want AI on your IoT fleet without a rebuild? Start a 30-day edge AI pilot alongside your current platform. Book a NexaStack demo.
