CloudCostCompare

Guides

Google Cloud IoT Core is gone: what to use instead

Updated 27 Oct 2026

Google shut down Cloud IoT Core on 16 August 2023. Teams that built on it had to move device connectivity somewhere else, and anyone starting a new IoT project on Google Cloud now has to choose a gateway themselves.

What Cloud IoT Core did

It gave each device an identity, accepted MQTT and HTTP connections, and passed telemetry into Pub/Sub. Everything after that (Dataflow, Bigtable, BigQuery) was ordinary Google Cloud, and that part still works.

Option 1: stay on Google Cloud with a partner broker

Run an MQTT broker, either a partner’s managed service or one you host yourself, and have it publish into Pub/Sub. The rest of a Google pipeline stays the same.

  • Cost: you pay the broker’s price (a separate bill or the VMs to run it) on top of Pub/Sub’s volume charges.
  • Effort: you take on device identity and certificate management, which IoT Core used to handle.

Option 2: use a managed IoT service on another cloud

Both Azure IoT Hub and AWS IoT Core are still offered and include device identity, MQTT and routing rules.

  • Cost: priced per unit (Azure) or per message (AWS). See current gateway prices.
  • Effort: if your analytics stay on Google Cloud, you also pay to move data between clouds.

Option 3: move the whole pipeline

If the gateway was the main reason for being on Google Cloud, it can make sense to price the entire pipeline elsewhere. Our Industry 4.0 estimator prices gateway, streaming, processing, storage and compute together on each cloud. It shows Google Cloud’s missing gateway as a gap, so you can see the cost of everything else.

How to decide

  1. Count your devices and message rate. Gateway pricing models diverge most at high volumes.
  2. Price the full pipeline, not just the gateway. Stream processing and storage often outweigh it.
  3. Add the cost of running a broker if you stay on Google Cloud: VMs, high availability and someone’s time.
  4. Include data transfer if devices connect to one cloud and analytics run on another.

See what a 1,000-device plant costs on each cloud as a starting point.