No capacity units.
No throttling.
Fabric makes you buy an F-SKU capacity, size it, reserve it, and babysit it — smoothed and throttled whether you're busy or idle. Databasin is a per-minute data ecosystem on open Apache Iceberg. Four engines, one lake you own, an AI analyst built in. Stop the cluster, the meter stops.
A platform that just runs.
Not a capacity you babysit.
Fabric is a bundle of proprietary experiences over a capacity you provision and manage. Databasin is one open ecosystem, metered by the minute — no CUs to size, no smoothing window, no throttle when you get busy.
No F-SKU to size, no reservation, no annual commit. Run a cluster and pay for the minutes it runs. Stop it, and the meter stops.
Fabric smooths your usage over time and throttles interactive queries when you exceed the capacity you bought. Databasin gives you the compute you asked for, when you asked for it.
Every table is real Apache Iceberg on Parquet, in your lake — not Delta with Iceberg bolted on as a virtualized overlay. Any engine reads it. Leaving is not a migration.
Trino for federated SQL, Doris for sub-second OLAP, Spark for ML, DuckDB for in-process speed — all pointed at the same tables. Pick the right engine per job, no copies.
75+ certified connectors ingest on a schedule and build your gold views for you — not Data Factory pipelines you design, wire, and maintain by hand.
SaaS, or self-installed into your own Azure tenant with your keys and VNets. Same ecosystem, same features, zero vendor logical access if you want it. HIPAA-ready with BAA available.
The capacity model,
line by line.
Microsoft Fabric bills a fixed capacity you size in advance. Databasin bills the minutes you actually run. Here's the same work, side by side.
| Capability | Databasin | Microsoft Fabric |
|---|---|---|
| Billing model | ✓Per-minute, post-pay · no commit, or flat-rate unlimited | Fixed F-SKU capacity (CUs), reserved or pay-as-you-go |
| Idle cost | ✓$0 when stopped · the meter stops with the cluster | You pay for the capacity around the clock unless you pause it |
| Capacity sizing | ✓None · scale compute per workload, on demand | You size the F-SKU up front; resize is a capacity change |
| Throttling / smoothing | ✓None · you get the compute you launched | Usage is smoothed over time; overages throttle then reject queries |
| Query engines | ✓Trino · Doris · Spark · DuckDB · open, one catalog | Spark · T-SQL Warehouse · KQL Eventhouse · Direct Lake — per-workload |
| Table format | ✓Native Apache Iceberg + Parquet · any engine reads it | Delta on OneLake; Iceberg via metadata virtualization overlay |
| Data pipelines (ELT) | ✓75+ certified connectors · gold views built for you | Data Factory pipelines — you build and maintain them |
| AI / natural-language analyst | ✓Databasin One, built in · $15 / 1M tokens | Copilot — requires an F64+ capacity to enable |
| Data ownership | ✓Your open lake · query from anywhere, leave anytime | OneLake, Microsoft-managed inside the Fabric tenant |
| Deployment | ✓SaaS or self-install in your own Azure tenant | SaaS on Microsoft-managed capacity |
Microsoft Fabric behavior reflects published capacity, smoothing, and throttling documentation and F-SKU pricing for US regions as of early 2026. Copilot in Fabric requires an F64 or higher capacity. Actual Fabric cost and throttling vary by SKU, region, and reservation term. Verify at azure.microsoft.com/pricing and Microsoft's Fabric documentation.
The same analytics workload.
Without buying a capacity.
A team that needs roughly an F64 of production capacity, but only actually queries 8 hours a day, 20 business days a month. Pipelines running 60 min/day. 500 GB stored. AI analyst in daily use.
Storage: $12 · AI (Databasin One): $45
BI & dashboards: $0 (included)
Paid 24/7 whether idle or busy
Pay-as-you-go: materially higher · Copilot: needs F64+
Open where it matters.
Nothing to babysit.
Fabric bundles proprietary engines over a capacity you manage. Databasin gives you open engines over a lake you own, metered by the minute — with the analyst built in.
Open storage you own.
Every table is native Apache Iceberg on Parquet, in your lake. Trino, Doris, Spark, DuckDB, or any engine you bring reads it directly — no copies, no conversion.
No capacity to babysit.
Scale compute per workload, on demand, and pay by the minute. Stop a cluster and the meter stops. No F-SKU to size, no reservation to true up.
Pipelines that run themselves.
75+ certified connectors ingest on a schedule and build your cleaned, business-ready gold views for you — bronze → silver → gold, handled.
Intelligent, no capacity required.
Databasin One is an agentic analyst built into the ecosystem — Claude + GPT with full catalog awareness. It picks the tables, writes the SQL, renders the chart, and cites its source. Included, metered per token.
The capacity math Fabric
hopes you don't run.
Fabric's sticker is the F-SKU. The bill you actually sign is that capacity — paid around the clock, smoothed and throttled at the ceiling — plus the pipelines you maintain and the F64 floor to turn Copilot on. Databasin replaces the capacity with a per-minute meter and an ecosystem that just runs.
Try the meter that stops.
$50 in credits. No card. No capacity to size. No commit. Load real data, run a real query, and ask Databasin One to build you a real dashboard — before lunch.