How IoT platform pricing works on Azure, AWS and Google Cloud
Connecting devices to the cloud looks similar on every platform: devices publish telemetry over MQTT, and the platform routes it to storage and analytics. The bills look nothing alike, because each vendor counts usage differently. Understanding the counting rules matters more than comparing headline prices.
Azure IoT Hub: capacity units with a daily quota
Azure sells IoT Hub in units. Each unit of a tier (S1, S2 or S3 in the Standard tier) comes with a fixed number of messages per day, and you pay a flat monthly price per unit whether you use the quota or not.
Two details drive the bill:
- Messages are counted in 4 KB blocks. A 6 KB message uses two messages of quota.
- Quota resets daily. You size for your busiest day, not your monthly average.
This model is predictable and gets cheaper per message at higher tiers, but you pay for headroom. A workload that sits just above a unit’s quota pays for a whole extra unit. The estimator picks the cheapest tier and unit count for your load.
AWS IoT Core: pay per message, connection and rule
AWS IoT Core has no capacity units. It charges separately for:
- Messages, counted in 5 KB blocks, with lower prices at higher monthly volumes;
- Connectivity, per minute each device stays connected;
- Rules engine use, per rule triggered and per action executed, which is how messages get routed to other services.
This model scales smoothly from a handful of devices to millions, and you pay nothing for unused capacity. The trade-off is that several small charges add up, and routing every message through a rule adds a cost per message.
Google Cloud: no managed IoT service
Google retired Cloud IoT Core on 16 August 2023. There is no first-party managed device gateway on Google Cloud today. The common replacement is a partner MQTT broker that forwards data into Pub/Sub, which charges by the volume of data published and delivered.
That means a Google Cloud IoT design has an extra component to choose, run and pay for, which a like-for-like price comparison can’t capture. Our estimates show this as a gap rather than pretending the gateway is free.
Which model suits which workload
| Your situation | Tends to favour |
|---|---|
| Steady, predictable telemetry from a fixed fleet | Unit pricing (Azure IoT Hub), if the load fits a unit well |
| Spiky traffic, or a fleet that grows month to month | Per-message pricing (AWS IoT Core) |
| Large messages (over 4–5 KB) | Check both: block sizes change the count sharply |
| Already all-in on Google Cloud | Pub/Sub plus a partner broker, priced separately |
These are tendencies, not rules. The gateway is often not the largest line on the bill: stream processing and time-series storage frequently cost more. See example factory costs for a full breakdown.
Compare current prices
Or estimate your own setup with your device count and message rate.