Key takeaway
- Google Cloud IoT Core was retired in August 2023. See what Google documents today, how Pub/Sub, Dataflow and BigQuery fit, and how to choose.
What happened to the Google Cloud IoT platform?
Google Cloud IoT Core, the managed service that registered devices and bridged MQTT and HTTP traffic into Google Cloud, was retired. Google announced a retirement date of 16 August 2023, as reported by InfoQ. Google Cloud still offers the building blocks for an IoT platform, but you now assemble them yourself or with a partner.
This guide explains what IoT Core did, what Google Cloud documents today for IoT architecture, and how to choose a path. For the wider design view, read our guides to the IoT analytics platform and IoT edge analytics.
What did Cloud IoT Core do?
IoT Core had two main parts. The device manager registered devices, authenticated them and held a logical configuration for each one. The protocol bridge let devices connect over standard protocols such as MQTT and HTTP, so existing devices needed few firmware changes. Data then flowed to other Google Cloud services for processing and storage. Those functions still need an owner in any IoT design, which is why the replacement question matters.
What does Google Cloud document for IoT now?
Google's IoT platform architecture guide (last reviewed 9 August 2024) says IoT platform products typically provide basic MQTT and HTTPS data connectivity. It describes three parts, shown below.
| Part | What the guide describes | What it means for you |
|---|---|---|
| Ingestion | An MQTT broker or endpoint, plus an optional HTTPS endpoint, behind an external proxy Network Load Balancer | You run or buy the broker |
| Backend integration | Applications connect through Pub/Sub or the Dataflow MQTT connector (Apache Beam) | Messages reach your processing code |
| Device management | Credential storage, authentication, configuration, registry and update management | You or a partner keep the device inventory |
The same page links to Intelligent Products Essentials as a related guide for building IoT applications on Google Cloud. Partner platforms are another route: ClearBlade, a Google partner, announced a migration path from IoT Core, as reported by InfoQ.
Which Google Cloud services still fit an IoT pipeline?
| Service | Role in an IoT pipeline | Status and source |
|---|---|---|
| Pub/Sub | Decouples devices or gateways from the services that process messages | Current; Pub/Sub overview |
| Dataflow | Processes unbounded streams with the Apache Beam SDK, for example from Pub/Sub | Current; streaming pipelines |
| BigQuery | Serverless analysis with SQL and Python, with streaming ingestion and BigQuery ML | Current; BigQuery introduction |
| Cloud Run functions | Event-driven code without server management; Google suggests new functions are created in Cloud Run | Current under this name; overview |
| Android Things | Operating system for IoT devices | Turned down on 5 January 2022; Android Things |
| Cloud IoT Core | Device registry and MQTT/HTTP bridge | Retired 16 August 2023, as reported by InfoQ |
Pub/Sub, according to its documentation, is an asynchronous and scalable messaging service that decouples producers from consumers, with push delivery to webhooks. Console steps for creating an IoT Core registry no longer apply.
How should you choose a path after IoT Core?
| Option | When it fits | Trade-off |
|---|---|---|
| Self-run MQTT broker in front of Pub/Sub | You want control and already run containers | You own broker scaling, security and updates |
| Partner IoT platform | You want managed device registry and connectivity | Vendor dependence; check the partner's own pages for terms |
| Edge processing first | Latency or bandwidth limits make cloud-only designs costly | More fleet management at the edge |
| Another cloud's IoT service | Your estate already runs there | Data movement and a second platform to govern |
Adopt in this order: define the device identity and credential model, choose the ingestion endpoint, stream into Pub/Sub, process with Dataflow or Cloud Run functions, and land the results in BigQuery. Avoid building new designs on retired services, and avoid cloud-only designs where edge processing is needed. See our guide to edge computing and the use case on IoT for manufacturing.
What do governance, security and cost need?
Each device needs its own identity and credential, with a plan for rotating and revoking them. Decide who may act on device data, and keep an audit trail of automated actions. Costs come from message volume, stream processing and query use, so estimate them with the pricing pages of each service rather than a rule of thumb. Where AI agents act on IoT data, ElixirData Context OS keeps governed context, versioned policy and decision traces in one decision layer above AI agents, and NexaStack is an agentic operating platform for physical AI that runs across on-premises, cloud and hybrid environments, according to its site.
What will change in Google Cloud IoT over the next 3 to 5 years?
| Signal | Source | Likely effect |
|---|---|---|
| About 21.1 billion connected IoT devices at the end of 2025, growing 14% a year | IoT Analytics, State of IoT 2025, 28 October 2025 | More devices per fleet; identity and update management matter more |
| About 39 billion devices forecast for 2030 (13.2% compound annual growth rate for 2025 to 2030); more than 50 billion by 2035 | IoT Analytics, State of IoT 2025 | Ingestion and storage designs must scale without redesign |
| Google documents a build-your-own pattern (MQTT endpoint, Pub/Sub, Dataflow) in place of a single managed service | Google Cloud IoT platform architecture guide, reviewed 9 August 2024 | Expect IoT on Google Cloud to stay a composed architecture, with partners for device management |
| BigQuery describes itself as AI-ready, with in-place model building using SQL or Python | BigQuery introduction, Google Cloud docs | Analytics and models are likely to sit closer to the data |
This is an outlook based on the sources above, not a forecast from Google. Confirm product status on the official pages before committing.
To map one IoT decision your agents should make, and the context and policy it depends on, book a Context Workshop. See what happens in the workshop.
Frequently asked questions
Is Google Cloud IoT Core still available?
No. Google set the retirement of IoT Core for 16 August 2023, as reported by InfoQ. Google's current IoT platform architecture guide does not mention it. Check the Google Cloud documentation for the latest status before planning any design.
What replaces Google Cloud IoT Core?
There is no single drop-in service. Google's IoT platform architecture guide describes an MQTT broker or endpoint for ingestion, Pub/Sub or the Dataflow MQTT connector for backend integration, and separate device management for credentials, registry and updates. Partner platforms such as ClearBlade also announced migration paths, as reported by InfoQ.
Can Pub/Sub receive device data on its own?
Pub/Sub is a messaging service that decouples producers from consumers and works with Dataflow, BigQuery, Cloud Storage and Cloud Run functions, according to its documentation. Google's architecture guide places an MQTT or HTTPS endpoint in front of it for device traffic.
Is Android Things still supported?
No. According to the Android Things page, the console closed to new projects on 5 January 2021 and was turned down for all projects on 5 January 2022.
How many IoT devices will there be by 2030?
IoT Analytics, in its State of IoT 2025 report published on 28 October 2025, forecasts about 39 billion connected IoT devices in 2030, up from an estimated 21.1 billion at the end of 2025.
Where should an enterprise start?
Decide the device ingestion path first (self-run or partner MQTT, then Pub/Sub), then the analytics store, then the governance around who and what may act on the data.
Ready to put AI agents to work? See an ElixirClaw demo: the agentic OS to build, orchestrate and govern AI agents across your enterprise systems.
Explore the XenonStack platform family: ElixirData · ElixirClaw · ElixirHub · NexaStack · MetaSecure · Neural AI · AgentGRC · AgentSRE · AgentAnalyst · AgentInstruct · AgentRAI · AgentScrum · AgentSearch · AgentForce · AI Aviator · SAIF Aviator · CohortIQ · XenonStack AI · XenonStack UAE