Event Hubs and streaming cost control.

Streaming feels like a usage based service, but most of the bill is capacity you decided to hold. Here is how to size it deliberately.

Which tier should you actually run on?

Event Hubs has four tiers and the gap between them is large. Basic is for simple ingest with one day retention and no consumer group flexibility. Standard is the workhorse, billed per throughput unit hour with auto inflate available. Premium reserves isolated capacity in processing units and suits steady, latency sensitive workloads that outgrow shared Standard. Dedicated reserves a whole cluster in capacity units for the largest, most predictable estates.

The decision rule is sustained throughput. If your traffic spikes occasionally around a low baseline, Standard with a sane auto inflate ceiling is usually cheapest because you are not paying for idle reserved capacity. Once throughput is high and steady enough that you would hold many throughput units around the clock anyway, Premium or Dedicated can lower the effective rate while adding isolation. Buying Dedicated for a workload that is still finding its shape is the classic overspend.

What is auto inflate doing to your bill?

Auto inflate raises the throughput unit count automatically when ingress approaches the current ceiling, which protects you from throttling during spikes. The trap is that it never scales back down on its own. A single busy hour can push the count up and leave it there, so you quietly pay peak rates every hour after that until someone notices and resets it.

Treat auto inflate as a safety net, not a capacity plan. Set the maximum throughput unit count to a defensible number rather than the platform maximum, and review the floor on a regular cadence so it tracks real baseline load rather than the high water mark of one incident. The same discipline applies to processing units on Premium.

Where does retention and Capture cost hide?

Standard tier includes one day of message retention. Extending retention adds a storage charge per GB held, and on a busy hub that compounds fast. If the reason you want long retention is durable archive rather than replay, the Capture feature writes the stream straight to Azure Blob storage, where a cool or archive tier costs a fraction of in hub retention. Reserve long in hub retention only for genuine replay windows that consumers actually use.

Where streaming meets the rest of the Azure bill

Event Hubs rarely lives alone. The consumers downstream are often the same PaaS services that drive the rest of the estate, so size them together rather than in isolation, as covered in PaaS choices that cut the Azure bill. Streaming pipelines also generate heavy diagnostic data, and the governance tooling around them carries its own cost, which we work through in Purview and governance tool costs. The full estate view lives in the Azure cost optimization guide, and the cross cloud picture in the cross cloud cost optimization guide.

Talk through your streaming estate

We help enterprises size Event Hubs and the streaming estate around it so capacity tracks real load rather than the last incident. Our guarantee: we reduce your cloud spend or we reimburse our service fee. Pricing is either a Fixed Fee scoped up front or Gainshare, a share of verified savings with no retainer and no risk.