Databricks Genie converts natural-language questions into SQL and runs the resulting queries on metered compute.
An analyst can ask "what were our top ten accounts by revenue last quarter, split by region" and receive a table without seeing a cost estimate or scan size. The generated SQL runs on a warehouse that charges by the second.
Genie is the natural-language analytics component of Databricks AI/BI. It answers questions using tables governed through Unity Catalog. The generated SQL consumes compute when the associated query executes.
Genie does not have a separate per-seat line item. Databricks divides its cost between the large language model (LLM) that interprets the question and the SQL warehouse that executes the generated query. Free usage applies only to Genie LLM usage, while Genie compute such as SQL Serverless is billed separately.
Databricks has offered promotions that make model reasoning free or inexpensive for human users during defined periods. These promotions have effective dates and exclude service principals, which are billed at the normal rate. Current terms on the Databricks pricing page determine whether an allowance applies.
Every Genie answer executes on a SQL warehouse. Many deployments use a serverless SQL warehouse for fast startup. Serverless has the highest per-DBU rate among the SQL warehouse tiers, above Pro and Classic. A Genie query consumes the same warehouse resources as equivalent analyst-written SQL, including the full cost of wide scans or inefficient query plans.
Natural-language access can increase query volume by allowing users without SQL experience to submit repeated follow-up questions. In this example, nine follow-up questions can produce nine separate scans. Questions may also request more columns or wider date ranges than required. Across multiple users, this activity can keep a warehouse active, especially when its auto-stop interval is long.
The system.billing.usage table records DBU consumption by warehouse. SQL warehouse query history and monitoring record the queries executed, their duration, and scan volume. Filtering query history to the warehouse used by Genie allows the warehouse usage to be associated with Genie-generated queries.
The following query prices the warehouse's DBUs at current list rates:
-- daily list-price cost of the sql warehouse genie runs on
select u.usage_date, sum(u.usage_quantity * p.pricing.default) as list_cost_usd
from system.billing.usage u
join system.billing.list_prices p
on u.sku_name = p.sku_name
and p.price_end_time is null
where u.usage_metadata.warehouse_id = 'your-warehouse-id'
group by u.usage_date
order by u.usage_date;
Not as a standalone line item. The model reasoning bills on the Genie LLM meter, often promotional-free or on a small allowance for human users, and the SQL it generates bills as ordinary SQL warehouse compute. Your per-question cost is the warehouse cost of the query Genie ran.
The compute. Databricks states that free usage applies to Genie LLM usage only and that Genie compute, such as SQL Serverless, is billed separately. A free natural-language layer does not make the query execution free.
Serverless SQL warehouses hold the highest per-DBU rate of the SQL warehouse tiers, above pro and classic, in exchange for fast startup and no infrastructure management. Genie leans on that speed, so Genie queries tend to land on the priciest tier.
Query system.billing.usage for the warehouse Genie uses and cross-reference SQL warehouse query history and monitoring. That surfaces the queries Genie ran, their duration, and their scan size, which is the spend the Genie interface never displays.