HiveMQ Cloud: Tiers, Hosting, Serverless Retirement and Plant Fit
HiveMQ Serverless Retirement: Cloud vs Self-Hosted Decision Guide
HiveMQ Serverless retires on 31 December 2026. What users must do, Cloud Starter vs self-managed Lab, Launch, Run and Enterprise, and a checklist.
HiveMQ Serverless is the pay-as-you-go and free family of HiveMQ Cloud plans, and HiveMQ’s documentation states it retires on 31 December 2026. If you run a cluster on it, the question is not whether to move but where to, and how to do it without stopping a line. This page covers what Serverless users must do before that date, how HiveMQ Cloud Starter and Enterprise compare with the self-managed Lab, Launch, Run and Enterprise plans, a decision for a plant, a migration checklist and where it gets hard. It is current as of 21 November 2026.
It is the deeper page under HiveMQ Cloud and part of the HiveMQ hub.
What HiveMQ Serverless users must do before 31 December 2026
HiveMQ’s documentation states that after 31 December 2026, clients cannot connect to Serverless, and clusters and data are deleted. It also says no new Serverless clusters can be created, and that affected users can claim a free migration licence for the self-managed HiveMQ Enterprise broker if they received the retirement email. We found the date in one primary source, the Cloud documentation. Treat it as firm and ask HiveMQ to confirm it for your account.
- Confirm you are affected. Find every Serverless cluster in every account and project, including the test and pilot ones that nobody owns.
- Read the retirement email and the documentation. Note the migration licence conditions, which we did not read.
- Choose a destination. Cloud Starter, Cloud Enterprise, or a self-managed broker. The decision tree below is how we would choose.
- Inventory clients and data. List every client, its credentials, topics, QoS and whether it relies on retained messages or persistent sessions.
- Build the destination in parallel and run both until the move is proven.
- Cut over with a rollback, well before the end of December.

Cloud Starter, Enterprise or a self-managed broker
The table compares the options on the dimensions a plant cares about, using HiveMQ’s pricing page as listed on 21 November 2026.
| Cloud Starter | Cloud Enterprise | Self-managed Lab, Launch, Run | Self-managed Enterprise | |
|---|---|---|---|---|
| Price listed | From $0.34 per hour plus $0.80 per million messages | Quote | Lab free; Launch $299 per month; Run $499 per month | Quote |
| Who operates it | HiveMQ | HiveMQ | You | You |
| Connections | 10,000 | Custom | Lab 25; Launch and Run 10,000 | Custom |
| Throughput | 1 MB per second | Custom | Lab 25, Launch 500, Run 1,000 messages per second (5 KB normalised) | Custom |
| Brokers | Managed | Managed | One broker each | Multi-broker, multi-site |
| Support | Silver | Silver to Platinum | Lab and Launch community; Run Silver | Silver to Platinum |
| Where the data sits | EU or US | Choose your cloud | Wherever you run it | Wherever you run it |
The Lab plan is a 30-day licence with manual renewal, so it suits a test and not production. All self-managed plans include TLS, authentication, authorization and the Enterprise Extensions. Whether the Enterprise plan is priced by connection is not stated publicly. The migration licence is a separate offer for Serverless users, not one of these rows. The editions and pricing page explains the self-managed plans in more depth.
Illustrative arithmetic, not a total cost. If a Cloud Starter cluster ran for a whole month of 730 hours at the listed $0.34 per hour, the hourly part would be about $248, before message charges and before any larger size. A Run plan is $499 per month for the licence alone, before the server, backups, patching and the engineer’s time. The two prices are not comparable without those, and neither is the answer by itself.
Choosing: where data stays, latency and who operates it
Three questions decide most plants.
- Where must the data stay? If a rule or contract fixes it to a site or a country, and “EU and US” is too coarse, a broker you run is the safer reading. If either region is acceptable, Cloud is open.
- What happens when the link drops? A line that must keep working when the internet fails needs a broker at the plant. Starter, being cloud-hosted, adds a WAN to every message. Our HiveMQ Cloud page shows the failure points.
- Who will operate it, and can they? A self-managed broker needs patching, monitoring, backups and an on-call answer. If nobody at the plant can own that, a managed service is the honest choice, with a local buffer.
In our view, most plants end up with both: a broker or gateway at the site and a managed service for the enterprise side, joined by a bridge. The HiveMQ Platform page covers the self-managed side, and our Unified Namespace guide explains why the plant broker is usually the source of truth.
A HiveMQ Serverless migration checklist
- Every Serverless cluster and owner identified.
- Retirement email found and migration licence conditions read.
- Destination chosen and approved, with the reason written down.
- Client inventory: identity, topics, QoS, will messages, retained and session use.
- Access rules and credentials recreated on the new broker, and tested with a denied case.
- TLS trust, hostname and port changes listed per client.
- Message size, topic alias and connection limits compared with the old broker.
- Retained messages: which topics need them, and who republishes.
- Capacity checked against the reconnect peak, not the average.
- Monitoring and alerts live on the new broker before cutover.
- Parallel run, then a cutover window with a rollback to the old cluster.
- Cutover finished before the end of December, with a buffer for surprises.
Where it gets hard
- Clients you cannot reach. A device with a baked-in hostname, or a vendor-managed gateway, may need a site visit or a supplier ticket. These set the schedule.
- Retained messages and sessions. We found no documented export between brokers. A UNS that relies on retained values to show current state looks empty until publishers republish.
- Limits that differ. A new broker may cap message size, topic aliases or connections differently. Find out in the parallel run, not at cutover.
- A single date. The 31 December 2026 date rests on one documentation page. A year-end freeze at the plant also removes the working days you need.
- Migration licence terms. Unread by us. A free licence that expires at an awkward date is not free.
- Self-managed means you operate it. Clustering, discovery on a cloud network and Kubernetes are real work. See the HiveMQ cluster documentation, and plan for failure at the plant.
For help choosing a destination and writing the cutover plan, our IT/OT integration service covers broker placement and client migration. We are independent of HiveMQ and have no commercial relationship with the company.
Back to HiveMQ Cloud, the HiveMQ hub, or across to a sibling: editions and pricing.
Frequently asked questions
What must HiveMQ Serverless users do before 31 December 2026?
HiveMQ's documentation states that after that date clients cannot connect and clusters and data are deleted. Before it, export or recreate your configuration, choose a destination (Cloud Starter or a self-managed broker), repoint and test every client, and cut over with time to spare. Do not leave the cutover to the last week of December.
Is the HiveMQ migration licence really free?
HiveMQ's documentation states that affected Serverless users can claim a free migration licence for the self-managed Enterprise broker, and that the claim follows the retirement email. We did not read the licence terms, so ask HiveMQ for its duration, connection limits and what happens when it ends.
Should a plant move from Serverless to Cloud Starter or to a self-managed broker?
It depends on where the data must stay, how the line behaves if the link to the cloud drops, and who will operate the broker. Starter is the lighter move, with HiveMQ operating the broker. A broker at the plant keeps the line independent of the WAN but adds patching and monitoring work.
Do my MQTT clients need to change when the broker moves?
Usually only the hostname, port, credentials and TLS trust settings, because the protocol is the same. Check the details a new broker can change: client limits, topic alias and message size limits, retained messages, persistent sessions and access rules. Test every client class, not a sample.
Can I move retained messages and sessions across?
We did not find a documented HiveMQ export of retained messages or persistent sessions between Serverless and another broker. Plan to have clients republish retained values after the move and to start sessions fresh, and confirm the procedure with HiveMQ.
References and further reading
- HiveMQ Cloud documentation · HiveMQThe source for the 31 December 2026 retirement and the migration licence.
- HiveMQ Pricing · HiveMQCloud Starter and Enterprise, and the self-managed Lab, Launch, Run and Enterprise plans, as listed on 21 November 2026.
- HiveMQ Cloud Serverless vs Starter · HiveMQHow the two Cloud plan families differed.
- HiveMQ cluster documentation · HiveMQWhat a self-managed cluster needs.