Direct answer: at Azure Stream Analytics public list prices on 10 November 2026 (eastus, USD, excluding tax and discounts), use SU V2: 1 SU V2 equals 1 streaming node, with a 1/3 SU minimum. V2 bills monthly aggregated SU-hours in three tiers: $0.33/SU-hour for 0–730 hours, $0.144474 for 730–5,840, and $0.120318 above 5,840. For a full 730-hour month: 1/3 SU $80.30, 2/3 SU $160.60, 1 SU $240.90, 2 SU $346.37, 3 SU $451.83, 6 SU $768.23, 12 SU $1,330.49, 24 SU $2,384.48. Legacy SU V1 is $0.11/SU-hour, but 6 SUs equal 1 node, so one node costs $481.80/month; V2 delivers the same node for $240.90, 50% less. At 6 nodes V2 is $768.23 vs V1 $2,890.80 (73.4% less), 12 nodes save 77.0%, and 24 nodes save 79.4%, close to Microsoft's "up to 80%" claim. V1 is marked to be deprecated in the official docs; start new jobs on V2. Edge is $1/device-month, billed once a job is deployed to devices regardless of running, failed or stopped state.
This guide is for manufacturing architects, data engineering, SRE, finance and procurement teams building device streams, production-line alerts, real-time dashboards or event-driven pipelines. It separates official price facts, editorial calculations and exclusions, covering V1/V2 unit conversion, V2 tiering, node-equivalent comparisons, job sizing, Edge and Dedicated clusters, hidden costs and a 90-day checklist. Prices come from the Stream Analytics pricing page, Azure Retail Prices API and Microsoft Learn (10 November 2026). This is not procurement advice.
(1) Don't compare $0.11 with $0.33: V1 needs 6 SUs per node, V2 needs 1; at equal compute V2 starts 50% cheaper. (2) Tiers apply to monthly aggregated SU-hours: 1 SU full-time is $240.90, but 6 SUs is not 6×—it is $768.23 (730 hours at $0.33 + 3,650 at $0.144474). (3) V1 is marked to be deprecated: start new jobs on V2, and migrate old jobs by node count, never 1:1 by SU number.
V1 vs V2: same compute, different units
V1 needs 6 SUs per streaming node, with 1 and 3 SU as fractional nodes; V2 simplifies to 1 SU = 1 node and adds 1/3 and 2/3 SU for low-throughput jobs. Capabilities are unchanged, and V2 is the recommended, cheaper-per-compute model. Migrate by node count: V1 36 SU = 6 nodes, which maps to V2 6 SU, not V2 36 SU. A 1:1 SU migration would multiply compute by six.
| Item | SU V1 | SU V2 (recommended) |
|---|---|---|
| SUs per node | 6 | 1 |
| Minimum unit | 1 SU (1/6 node) | 1/3 SU |
| Available steps | 1, 3, 6, 12, 18… | 1/3, 2/3, 1, 2, 3… |
| Unit price | $0.11/SU-hour | $0.33 (first tier) |
| One node/month (730h) | $481.80 | $240.90 |
| Status | To be deprecated | Recommended |
Sources: Stream Analytics pricing page and Learn "Understand and adjust Stream Analytics streaming units" (10 November 2026). Microsoft has not published a V1 retirement date; V2 has the same capabilities, at eastus PAYG prices.
V2 tiers: more SU-hours, lower unit price
V2 is not a single rate: it bills monthly aggregated SU-hours in three tiers—$0.33 for the first 730, $0.144474 for 730–5,840, and $0.120318 above 5,840. At 730 hours/month: 1/3 SU $80.30, 2/3 SU $160.60, 1 SU $240.90. A 2-SU job consumes 1,460 SU-hours, so 730 bill at $0.33 and 730 at $0.144474, totaling $346.37; 3 SU $451.83, 6 SU $768.23, 12 SU $1,330.49, 24 SU $2,384.48. The effective rate falls from $0.33 to $0.136/SU-hour as volume grows.
| SU V2 | Monthly SU-hours | Monthly cost | Effective $/SU-hour |
|---|---|---|---|
| 1/3 | 243.33 | $80.30 | $0.330000 |
| 2/3 | 486.67 | $160.60 | $0.330000 |
| 1 | 730.00 | $240.90 | $0.330000 |
| 2 | 1,460.00 | $346.37 | $0.237237 |
| 3 | 2,190.00 | $451.83 | $0.206316 |
| 6 | 4,380.00 | $768.23 | $0.175395 |
| 12 | 8,760.00 | $1,330.49 | $0.151882 |
| 24 | 17,520.00 | $2,384.48 | $0.136100 |
Editorial calculation: official three tiers, one job at 730 hours/month; the billing scope above 24 SU or across jobs follows the bill. If subscription-level aggregation applies, actual cost is lower; if per-job, each job is calculated separately. Excludes inputs/outputs, storage, Edge, tax and discounts.
Node-equivalent comparison: V2 saves 50%–79%
Align by node count: 1 node = V1 6 SU = V2 1 SU, $481.80 vs $240.90, 50.0% less; 6 nodes = V1 36 SU = V2 6 SU, $2,890.80 vs $768.23, 73.4% less; 12 nodes $5,781.60 vs $1,330.49, 77.0% less; 24 nodes $11,563.20 vs $2,384.48, 79.4% less. That is where Microsoft's "up to 80%" comes from—not the first-tier unit price alone, but V2's tier discount combined with the simplified unit.
| Nodes | V1 SU | V1/month | V2 SU | V2/month | Saving |
|---|---|---|---|---|---|
| 1 | 6 | $481.80 | 1 | $240.90 | 50.0% |
| 6 | 36 | $2,890.80 | 6 | $768.23 | 73.4% |
| 12 | 72 | $5,781.60 | 12 | $1,330.49 | 77.0% |
| 24 | 144 | $11,563.20 | 24 | $2,384.48 | 79.4% |
Editorial calculation: V1 $0.11/SU-hour × 6 SUs/node × 730 hours; V2 at the three tiers; 24 nodes estimated as multiple jobs or a Dedicated cluster. Excludes inputs/outputs, storage, Edge, tax and discounts.
Job sizing: start at 1/3 SU
Microsoft's guidance: start non-PARTITION BY queries at 1 SU V2 and tune by SU% utilization; 1/3 and 2/3 SU suit low-throughput, single-stream, non-partitioned jobs. Common mapping: single low-throughput stream 1/3 SU ($80.30); single-node partitioned query 1 SU ($240.90); multi-partition/window aggregation 3 SU ($451.83); high throughput 6 SU ($768.23); larger 12 SU ($1,330.49). Don't buy by event count: run real data, then decide from SU% utilization, watermark delay and backlog.
| Scenario | Recommended SU V2 | Monthly | Fit |
|---|---|---|---|
| Single low-throughput stream | 1/3 | $80.30 | Non-partitioned, simple filter/aggregate |
| Single-node partitioned | 1 | $240.90 | PARTITION BY, one node |
| Multi-partition/window | 3 | $451.83 | Windowed aggregates, temporal joins |
| High throughput | 6 | $768.23 | Multiple partitions, higher event rate |
| Larger scale | 12 | $1,330.49 | Multiple jobs or cluster aggregation |
Editorial summary: SU guidance from Learn "Understand and adjust Stream Analytics streaming units"; monthly costs at eastus, 730 hours, official tiers. Actual sizing depends on SU% utilization, latency and backlog tests.
Edge and Dedicated clusters
Stream Analytics on IoT Edge is $1/device-month: 100 devices $100, 1,000 $1,000, 5,000 $5,000. Billing starts when a job is deployed to devices, regardless of whether it is running, failed or stopped—delete unused deployments to stop charges. A Dedicated V2 cluster has a 12-SU minimum and uses the same V2 tiers: 12 SU full-time is $1,330.49. It adds VNet and Availability Zone isolation for multiple jobs, compliance or dedicated capacity; Standard V2 jobs support VNet but not Availability Zones.
| Item | Billing | Example/month |
|---|---|---|
| Edge 100 devices | $1/device-month | $100 |
| Edge 1,000 devices | $1/device-month | $1,000 |
| Edge 5,000 devices | $1/device-month | $5,000 |
| Dedicated V2 12 SU | V2 tiers | $1,330.49 |
| Inputs/outputs/reference data | Billed by Event Hubs/IoT Hub/Storage | Separate |
Sources: Stream Analytics pricing page and Retail API (eastus, 10 November 2026). Edge bills on deployment; above 5,000 devices contact Microsoft; private endpoints may be charged in future.
Hidden costs and a 90-day checklist
Seven leaks: input/output services (Event Hubs/IoT Hub/Storage/Synapse bill separately), reference-data storage, unused Edge deployments (billed on deployment), SU overprovisioning (you pay for peak, not the trough), multiple jobs (each buys its own SUs), private endpoints/egress (possible future charges), and V1→V2 dual-running during migration. 90 days: weeks 1–2 inventory jobs, event rates, partition keys, windows and outputs; weeks 2–4 run real data at 1/3 or 1 SU and record SU% utilization, latency and backlog; weeks 4–6 map V1 nodes to V2 and evaluate tiers; weeks 6–8 consolidate small jobs, delete unused Edge deployments and optimize PARTITION BY; weeks 8–10 load-test peaks and failover; weeks 10–12 reconcile SU-hours, tiers and input/output charges against the bill.
FAQ
How is Azure Stream Analytics billed?
By SU-hour: V2 is $0.33/SU-hour for the first 730 hours, $0.144474 for 730–5,840, and $0.120318 above 5,840; V1 is $0.11/SU-hour but 6 SUs equal 1 node. Edge is $1/device-month; inputs/outputs, storage and reference data are billed by their services.
How do V1 and V2 convert?
1 node = V1 6 SU = V2 1 SU. V1 36 SU is 6 nodes, which maps to V2 6 SU; a 1:1 SU migration would multiply compute by six. V2 also offers smaller 1/3 and 2/3 SU steps.
What does 1 SU V2 cost per month?
At 730 hours and the first tier, $240.90. 1/3 SU is $80.30 and 2/3 SU $160.60. Once monthly SU-hours exceed 730, the excess moves to $0.144474, so 2 SU costs $346.37, not $481.80.
Why isn't 6 SU six times the 1 SU bill?
Because of tiers: 6 SU full-time is 4,380 SU-hours, with 730 hours at $0.33 ($240.90) and 3,650 at $0.144474 ($527.33), totaling $768.23; six times the 1-SU bill would be $1,445.40. Effective unit cost falls with volume.
Can I still use V1?
The official docs mark V1 as to be deprecated but do not publish a date. Start new jobs on V2 and migrate old jobs by node count, with a dual-run and validation window. V2 has the same capabilities as V1.
How is Edge billed?
$1/device-month: 100 devices $100, 1,000 $1,000, 5,000 $5,000. Billing starts when a job is deployed to devices regardless of running, failed or stopped state; delete unused deployments to stop charges.
When should I use a Dedicated cluster?
For VNet/Availability Zone isolation, multiple jobs, dedicated capacity or compliance. Dedicated V2 has a 12-SU minimum, $1,330.49 at full time; Standard V2 jobs support VNet but not Availability Zones.
How do I reduce cost?
Four steps: (1) migrate V1 to V2 by node count; (2) use 1/3 or 2/3 SU for small jobs; (3) consolidate small jobs to reach lower tiers; (4) delete unused Edge deployments and optimize PARTITION BY/windows to avoid overprovisioning. Optimize inputs/outputs and storage separately—don't look only at the SU price.
Research method and limitations
Official facts (pricing page, Retail API, Learn): V2 1 SU = 1 node with 1/3 and 2/3 steps; V2 tiers $0.33/$0.144474/$0.120318; V1 $0.11/SU-hour, 6 SU = 1 node, marked to be deprecated; Edge $1/device-month and billed on deployment; Dedicated V2 minimum 12 SU. Editorial calculations: 730 hours/month, per-job SU-hours, official tiers; effective rate = monthly cost ÷ (SU × 730).
Seven limitations: (1) only eastus public list prices, excluding tax/CSP/enterprise discounts; (2) event rate, partitioning and windows are editorial assumptions, with no real throughput benchmark; (3) tier aggregation scope (per job vs subscription) follows the bill—this article uses per-job math, so subscription-level aggregation would be lower; (4) excludes inputs/outputs, storage, reference data, Edge, private endpoints, egress and tax; (5) Microsoft has not published a V1 retirement date, so migration needs buffer; (6) 24 nodes is estimated as multiple jobs or a cluster, not a single-job cap; (7) prices and roadmaps change—recheck official pages and bills before purchase.
Sources and basis
- Microsoft Azure, Azure Stream Analytics Pricing: azure.microsoft.com/en-us/pricing/details/stream-analytics. V1/V2 structure, three tiers, Edge, Dedicated cluster. Accessed 10 November 2026.
- Azure Retail Prices API (productName=Stream Analytics, eastus): prices.azure.com/api/retail/prices. $0.33/$0.144474/$0.120318, V1 $0.11, Edge $1/device-month. Accessed 10 November 2026.
- Microsoft Learn, Understand and adjust Stream Analytics streaming units: stream-analytics-streaming-unit-consumption. 1 SU V2 = 1 node, 1/3 and 2/3, V1 6 SU = 1 node, deprecation note, SU% utilization. Accessed 10 November 2026.
- Microsoft Learn, Scale Azure Stream Analytics jobs: stream-analytics-scale-jobs. SU adjustment, PARTITION BY, throughput and latency trade-offs. Accessed 10 November 2026.
Data cut-off 10 November 2026; prices/availability follow Microsoft official pages and bills; this is not procurement, commercial or legal advice.