BI dashboards
High-concurrency, small queries run on right-sized compute.
Databricks SQL cost optimization
Databricks bills by the DBU. Yuki makes every one count.
Cut Databricks SQL spend without touching a single query. Yuki sends every query to the right SQL warehouse and consolidates workloads, so you only pay for the compute you actually need.
Cost over time (DBUs)
Purple: cost actually used. Dashed: savings.
Yuki for Databricks is a real-time cost optimization layer for Databricks SQL. It analyzes each query as it arrives and routes it to the most cost-efficient SQL warehouse, then consolidates workloads so fewer warehouses sit idle. Teams connect by changing connection details only, with no changes to SQL, dbt models or pipelines. Yuki reads usage metadata only and runs privately in the customer’s cloud.
Databricks SQL bills per DBU for every minute a SQL warehouse is up, not per query. Three patterns drive most waste:
How it works
Yuki acts as an intelligent routing layer across your Databricks SQL warehouses. It continuously analyzes query behavior and runs each query where it costs the least. Your teams keep building. Yuki handles the rest.
Query volume only goes up: more dashboards, more pipelines, more AI agents. As it grows, Yuki routes every query to the most cost-efficient compute, so your DBU spend doesn’t climb with it.
dbt and Lakeflow Jobs
Fast when you need it, cheap when you don’t.
Yuki routes every dbt run to the right SQL warehouse automatically. Heavy transformations and incremental merges get the compute they need to finish on time. Lightweight models and tests run on less. Your dbt project stays exactly as it is.
Yuki deploys privately in your cloud and connects to Databricks through secure APIs, so your data never leaves your workspace. Stay in control with role-based access, budget guardrails, routing policies and audit logs. Yuki decides where queries run, never what they can access.
AI workloads
AI agents and automated pipelines can fire thousands of queries on their own. Yuki manages that traffic in real time, so automation runs at full speed without unpredictable DBU spikes.
| Capability | Databricks built-in controls | Yuki for Databricks |
|---|---|---|
| Warehouse sizing | Set manually per warehouse | Chosen per query, automatically |
| Auto stop | Fixed idle timeout per warehouse | Consolidates workloads so fewer warehouses stay up |
| Routing | Users or tools pick a warehouse | Each query routed to the cheapest warehouse that meets its needs |
| Adapts to workload changes | Only when someone re-tunes settings | Continuously, as query patterns change |
| dbt runs | One warehouse per profile/target | Heavy and light models routed separately |
| AI agent traffic | Same controls as human queries | Managed in real time to prevent DBU spikes |
| Governance | Budgets, tags, usage dashboards | Role-based access, budget guardrails, routing policies, audit logs |
| Effort | Ongoing manual tuning | Set up once, no maintenance |
Yuki works alongside Databricks’ native controls. It adds the per-query decision they don’t make.
High-concurrency, small queries run on right-sized compute.
Heavy merges get power, tests and light models run on less.
Exploratory queries stop landing on the biggest warehouse.
Automated query bursts stay inside budget guardrails.
Update the connection details in your BI tools and pipelines. No code changes, no migration.
Yuki learns your SQL warehouses, query patterns and DBU usage.
Optimization starts right away and runs in the background.
Yuki analyzes each query in real time (scan size, runtime patterns and current load) and routes it to the most cost-efficient SQL warehouse. It also consolidates workloads onto fewer warehouses, so less compute sits idle.
Databricks gives you the building blocks: warehouse sizing, auto stop and usage dashboards. Those settings are static. Yuki makes the call query by query, all day, as your workloads change.
No. Yuki runs in your cloud, and your data and query results never leave your Databricks workspace.
Most teams cut Databricks SQL costs by 30-50% in the first month. Results depend on your usage. A free cost analysis will show your number.
Routing adds less than 50 milliseconds per query, with no slowdowns.
No. Once connected, Yuki learns your query patterns and adjusts routing on its own.
Yes. Yuki is built for thousands of queries a day across multiple workspaces and SQL warehouses.
No. Your SQL, pipelines and tools stay as they are. Only the connection details change.
A few minutes. Optimization starts as soon as you're connected.
Start with a free cost analysis. We'll show where your DBU spend goes and estimate your savings. No commitment.
Yuki works with serverless, pro and classic SQL warehouses.
Yes. Yuki connects to Databricks SQL on all three clouds.
Yes. Yuki works with the dbt-databricks adapter. Point your dbt profile at Yuki's connection details and every run is routed automatically, with no changes to models.
No. Yuki uses query and warehouse usage metadata only. Unity Catalog permissions are unchanged, and Yuki decides where a query runs, never what it can access.
Serverless autoscaling adds or removes capacity inside one warehouse. Yuki decides which warehouse each query should run on in the first place, across your whole Databricks SQL estate.
Contact us for pricing based on your Databricks SQL environment and optimization requirements.
Yes. Yuki runs the same routing approach for Snowflake and BigQuery, so multi-platform teams manage cost in one place.
Also explore Yuki for Snowflake and Yuki for BigQuery.
Share read-only usage metadata and get a report in minutes: what drives your DBU spend, where compute sits idle and what Yuki can automate.
Yuki estimated annual savings