Book a consultation

HiveMQ Alternatives: EMQX, Mosquitto, AWS IoT Core and More Compared

HiveMQ vs AWS IoT Core: Self-Managed Broker or Cloud Service for a Plant

HiveMQ vs AWS IoT Core compared: QoS 0 and 1 only on AWS, retained-message limits, Sparkplug evidence, lock-in, Azure Event Grid and plant fit.

Vladimir Romanov


HiveMQ vs AWS IoT Core is a choice between a broker you control and a cloud service that happens to speak MQTT. AWS IoT Core is a managed device-connectivity service in Amazon Web Services. Its MQTT support is based on versions 3.1.1 and 5.0 “with some differences”, the biggest being QoS 0 and 1 only. HiveMQ, whether self-managed or as HiveMQ Cloud, is an MQTT-first broker with all three QoS levels and a listing on the Eclipse Sparkplug compatible software list. In our view, AWS IoT Core suits device fleets already sending data to AWS, and a plant namespace that depends on Sparkplug state is safer on a broker you can test and control. This page sits under HiveMQ alternatives.

HiveMQ vs AWS IoT Core compared on the same dimensions

Public facts as of 20 November 2026, from each vendor’s documentation. We checked the HiveMQ broker documentation and did not test HiveMQ Cloud’s limits separately.

DimensionHiveMQ broker (self-managed or Cloud)AWS IoT Core
What it isMQTT broker and platformManaged device-connectivity service
Where it runsYour network, your cloud account, or HiveMQ CloudAWS only
MQTT versions3.1, 3.1.1 and 5.03.1.1 and 5.0 with documented differences
QoS levels0, 1 and 20 and 1 only
Retained messagesSupportedSupported; wildcard filters do not receive them at subscribe time
Shared subscriptionsSupportedDocumented
Message orderingNot compared hereMessage and acknowledgement ordering not guaranteed
MQTT 5 AUTH packetPart of MQTT 5; not separately checkedNot supported
Sparkplug (Eclipse list)Listed: Community, Professional, Sparkplug Aware ExtensionNot listed; no AWS statement found
Features beyond MQTTExtensions for Kafka, Kinesis, databases; Data Hub schema validationRules, jobs, Device Shadow, thing objects
PricingPublished tiers, datedProprietary, pay per use

The HiveMQ side comes from its broker user guide and the AWS side from the AWS IoT Core MQTT guide.

What the AWS differences mean in a plant

QoS 2 is absent. AWS also does not support the PUBREC, PUBREL and PUBCOMP packets, and a QoS 2 request gets no acknowledgement. Most industrial telemetry uses QoS 0 or 1, so for many designs this is a non-issue. A bridge or client that defaults to QoS 2 must be capped.

Wildcard subscriptions do not receive retained messages at subscribe time. AWS added retained messages in August 2021, but only an exact-match filter receives them when it subscribes. A dashboard that subscribes with a wildcard to read last known values would see nothing until each value changes. In our reading, that is the sharpest difference for a namespace built on retained last-known state.

Ordering is not guaranteed. AWS documents that message and acknowledgement ordering is not guaranteed. Sparkplug depends on birth messages arriving before data. We did not test whether this bites in practice.

Enhanced authentication is out. For MQTT 5 the AUTH packet and server redirection are not supported, so clients that use challenge-response authentication need another method.

Sparkplug has no AWS statement. We found none, and AWS IoT Core is not on the Eclipse list, so a Sparkplug design on AWS needs its own test evidence. The wildcard-retained and ordering differences above are where to look first.

Wind turbines in the distance over a field of yellow flowers under scattered clouds
Remote assets that talk only to the cloud are what a managed MQTT service was built for. A plant with one namespace is a different job.

AWS features that have no HiveMQ equivalent

AWS IoT Core is a device-connectivity service first. Its rules, jobs, Device Shadow and thing objects are AWS features, and none of them is MQTT. They save work for a fleet that already lives in AWS, and they are the part you cannot take to another broker. HiveMQ’s counterparts are extensions and Data Hub, which are HiveMQ features in the same way. Neither side is portable, so list which of them your design uses before you call the choice a protocol decision.

HiveMQ runs on AWS as well. Its documentation has EC2 and EKS guides, and its extension catalogue includes an Amazon Kinesis extension, a CloudWatch metrics extension and S3-based cluster discovery. A plant can therefore use AWS for analytics without using AWS IoT Core as its broker.

Azure Event Grid in brief

Microsoft’s Event Grid MQTT documentation lists MQTT 3.1.1 and 5 with WebSockets, QoS 0 and 1 only, TLS 1.2 or 1.3 required and a 1 MB message limit. On Sparkplug, a Microsoft Tech Community post of 23 March 2026 says Sparkplug B runs natively, with birth and death tracking. That is a vendor statement, and Event Grid is not on the Eclipse list. Azure IoT Operations runs a separate MQTT broker on Kubernetes at the edge. The fit is the same as AWS: Azure-centred estates.

Lock-in and the Google IoT Core lesson

What is portable. Standard MQTT clients connect to any compliant broker. What a cloud service adds around them is not: rules, registry entries, shadows and identity setup are specific to the service.

What happened to Google. Google announced the retirement of Cloud IoT Core in August 2022 and ended it on 16 August 2023, per Google’s services summary of July 2023, and TechCrunch reported the announcement at the time. Google pointed customers to its network of partners.

It is not only a cloud risk. HiveMQ’s documentation states that HiveMQ Cloud Serverless retires on 31 December 2026, with a free migration licence for the self-managed Enterprise broker offered to affected users. We found no second source for the date. A managed tier is the part a vendor can retire. See cloud vs self-hosted.

When each fits a plant

AWS IoT Core fits a fleet of remote devices over cellular or the internet, such as pumps or turbines, whose data goes straight to AWS analytics, where QoS 0 or 1 and a cloud-only footprint are acceptable. A self-managed HiveMQ broker fits a plant namespace whose publishers and consumers are on the plant network, where Sparkplug state matters and the plant must keep running when the WAN is down. Both fit in a layered design, which in our view is the common one.

A plant broker bridged to AWS IoT Core, with the checks at the boundary Two zones. In the plant network, PLCs, gateways and SCADA publish to a self-managed plant broker that supports QoS 0, 1 and 2, retained messages and Sparkplug. An outbound bridge forwards selected topics to AWS IoT Core in the AWS zone, which supports QoS 0 and 1, and passes data to a rules engine with device registry and shadows, then to cloud analytics and storage. A dashed box below lists the checks at the boundary from AWS documentation: QoS 2 is not supported, wildcard subscriptions do not receive retained messages at subscribe time, ordering is not guaranteed, and Sparkplug has no AWS statement so it must be tested. The pattern is our design view, not a documented HiveMQ configuration. Layered pattern: broker on site, selected topics to the cloud service Plant network PLCs, gateways, SCADA, historian Plant broker, self-managed QoS 0, 1 and 2 Retained messages, Sparkplug Keeps running if the WAN fails AWS AWS IoT Core MQTT endpoint QoS 0 and 1 only Rules engine, device registry, Device Shadow Cloud analytics and storage Bridge, outbound, QoS 0/1 Checks at the boundary, from AWS documentation QoS 2 is not supported. Wildcard subscriptions do not receive retained messages at subscribe time. Ordering is not guaranteed. No AWS statement on Sparkplug: test it. Pattern is our design view.
The plant broker keeps full MQTT behaviour on site and forwards only the topics the cloud needs. The AWS limits then apply at one known boundary, not across the whole namespace.

Illustrative: a beverage plant publishes line states on Sparkplug to a plant broker that its SCADA, historian and dashboards use. A data team in AWS wants production counts and downtime events. The bridge forwards only those topics, capped at QoS 1, and the cloud side does not need retained last-known values, so the AWS limits cost nothing. Moving the whole namespace into AWS IoT Core would test every difference above. The multi-site architecture page covers the pattern.

Where it gets hard

  • A retained-state design on AWS needs a test. Subscribing with a wildcard to read last known values does not work on AWS IoT Core. Test the exact subscriptions your consumers use.
  • Defaults that exceed QoS 1. Client libraries and bridges that default to QoS 2 fail or stall against AWS. Cap them.
  • The cost model differs. AWS is pay per use and HiveMQ publishes tiers with connection and throughput caps (editions and pricing). Model your real message counts for both. We did not compare AWS prices.
  • Quotas. We did not review AWS IoT Core’s service quotas, and they apply at fleet scale.
  • Network design. A bridge to the cloud is outbound traffic from the plant. Where it crosses the zones is a design decision: see OT network segmentation.
  • Exit. List the AWS rules, registry entries and shadows your design depends on before committing, because those are what you rebuild on exit.

Compare the siblings: HiveMQ vs EMQX and the enterprise-broker view in Solace vs HiveMQ. Back to HiveMQ alternatives and the HiveMQ hub. For an independent view on a cloud-versus-plant broker decision, see our IT/OT integration work.

Frequently asked questions

Does AWS IoT Core support MQTT QoS 2?

No. AWS's developer guide says AWS IoT Core supports QoS 0 and 1 only, and that it does not support the PUBREC, PUBREL and PUBCOMP packets QoS 2 needs. Most industrial telemetry uses QoS 0 or 1, so this often does not matter, but a design that relies on exactly-once delivery has to change. HiveMQ's documentation lists QoS 0, 1 and 2.

Can AWS IoT Core carry Sparkplug B?

We found no AWS statement on Sparkplug support, and AWS IoT Core is not on the Eclipse Sparkplug compatible list. The documented retained-message and ordering differences are the points a Sparkplug design must test, because Sparkplug relies on retained STATE messages and ordered birth messages. Anyone claiming it works or fails should say what they tested.

Does AWS IoT Core support retained messages?

Yes, since AWS announced retained-message support in August 2021, with a documented limit: a subscription to a wildcard topic filter does not receive retained messages at subscription time, only an exact-match filter does. A client that subscribes with a wildcard to read last known values will see nothing until the next publish.

Is Azure Event Grid a better fit than AWS IoT Core for a plant?

It has the same QoS 0 and 1 limit and a similar fit: cloud-centred estates. Microsoft's documentation lists MQTT 3.1.1 and 5, and a Microsoft post of 23 March 2026 says Sparkplug B runs natively, which is a vendor statement. The Eclipse list does not include it. Choose by where your data platform already is, then test.

Why did Google shut down Cloud IoT Core, and does it matter?

Google announced the retirement in August 2022 and ended the service on 16 August 2023. We did not find a reason in the sources we read. It matters as a dependency lesson: an MQTT endpoint tied to one cloud's roadmap can disappear. Standard MQTT clients are portable, but rules, registries and device state in the cloud service are not.

References and further reading

  1. AWS IoT Core MQTT developer guide · Amazon Web ServicesQoS, retained message and ordering differences from the MQTT specification.
  2. AWS IoT Core supports MQTT retained messages · Amazon Web Services
  3. Azure Event Grid MQTT support · Microsoft
  4. Azure Event Grid MQTT broker: enterprise-grade messaging · Microsoft Tech CommunityThe 23 March 2026 post carrying the Sparkplug claim.
  5. Sparkplug Compatible Software · Eclipse Foundation
  6. Google Cloud will shutter its IoT Core service next year · TechCrunch
  7. Google Cloud Platform Services Summary (July 2023) · Google CloudGives 16 August 2023 as the Cloud IoT Core discontinuation date.
  8. HiveMQ Enterprise MQTT Broker user guide · HiveMQ
  9. HiveMQ Cloud documentation · HiveMQ

← Alternatives

Have a project that needs technical ownership?

Send us the situation. If it is not something we should take on, we will say so.

The Plant-Side Brief

The weekly brief for the people who own a plant’s control systems.

One email every Wednesday for engineering and operations leaders responsible for controls, SCADA, MES and OT networks. Skim the top in two minutes; read the Deep Dive when you have fifteen.

  • One modernization problem, worked through to cost, risk and a schedule you can defend. The Deep Dive.
  • End-of-support dates before they reach your plant Windows, servers, controllers and HMI software. Lifecycle Watch.
  • A checklist your engineers can run that week not a framework to admire. Field Kit.
  • Every number sourced, no vendor agenda we are paid by plants, never by vendors.