Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →The $14k weekend described here is the author’s own figure. The total and its cause come from that account and are not independently verified in this article. What Databricks’ documentation does establish is how serverless charges accrue, which billing records show who ran what, and which controls limit spend and which do not. A team that wants to explain a surprise bill, or prevent one, needs those three things first.
How a serverless weekend bill builds up
Serverless compute is billed in DBUs for the workloads that consumed them. A weekend bill is therefore the sum of every serverless workload that ran, priced at the rate for each SKU. Three behaviors described in Databricks documentation explain how that sum can grow without anyone deliberately starting a large job:
As an Amazon Associate I earn from qualifying purchases.
- Scale-up limits apply to each workload. Notebook, job, and pipeline scale-up limits cap the maximum cost per workload per hour. They do not prevent new serverless workloads from launching, so several workloads running at the same time can add up well beyond any single workload’s ceiling.
- Notebook queries have a default timeout. Databricks’ cost-management documentation (2026) sets a default execution timeout of 2.5 hours for serverless notebooks. Workspace admins can change that default in Compute settings, and an individual notebook can override it by setting
spark.databricks.execution.timeout. Every raised or overridden limit lets a single query run longer. - Some features bill under the serverless jobs SKU. Data quality monitoring and predictive optimization can appear as serverless jobs SKU usage, even when a team did not knowingly run a serverless notebook or job. Those charges will not appear in a list of jobs someone recently created.
Find the spend in system.billing.usage
The system.billing.usage table is the starting point. Databricks documents it as the source for serverless notebook and job usage, with identity and workload metadata that connects records to users, jobs, and notebooks. Work through it in this order.
Recommended Free Tools
- Set the window to the weekend and include one day on either side, so workloads that started late on Friday or finished on Monday are captured.
- Group by SKU and by the identity that ran the workload, using
identity_metadata.run_as. This field identifies the user or service principal whose credentials ran it. - Group by workload identifiers:
usage_metadata.job_run_id,usage_metadata.job_name,usage_metadata.notebook_id, andusage_metadata.notebook_path. - Sum the DBUs per run. Do not read one row as one run. Databricks says its distributed architecture can create several records for the same job ID, run ID, or name within a timeframe, so the quantities must be added together.
- Wait before you finalize. Databricks says usage records can take up to 24 hours to appear in the billable usage table. A query run on Sunday evening can undercount Sunday’s work, so rerun it on Monday before comparing totals.
- Map each high-usage row back to the workspace. Immutable job and notebook IDs from the billing record locate the corresponding item in the UI, even if it was renamed or moved since it ran.
SELECT
usage_date,
sku_name,
identity_metadata.run_as AS run_as,
usage_metadata.job_id,
usage_metadata.job_run_id,
usage_metadata.job_name,
usage_metadata.notebook_id,
usage_metadata.notebook_path,
SUM(usage_quantity) AS total_dbus
FROM system.billing.usage
WHERE usage_date BETWEEN '2026-10-03' AND '2026-10-04'
GROUP BY ALL
ORDER BY total_dbus DESC;Change the dates to the weekend under investigation. The query returns DBUs, not dollars.
#1 Best Overall
Converting DBUs to dollars
Multiply the summed DBUs by the rate that applies to your contract, cloud, region, and SKU. Databricks’ cost-query guidance uses list prices as an estimate. Discounts can require a custom pricing table, so a list-price figure can differ from an invoice. Confirm the rate before presenting a dollar total as fact.
Also separate what the table does not cover. For non-serverless compute, the Databricks usage table does not include cloud infrastructure spend, which must be checked in the cloud provider’s console. If the $14k figure came from a cloud bill rather than Databricks usage, the two sources need to be reconciled line by line before the total can be attributed to serverless compute.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Controls compared: what each one actually limits
Databricks lists several cost-management tools. They do different jobs, and only some of them stop spending. The table below separates what each control does from what it does not do.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware match| Control | What it does | What it does not do | Availability noted in documentation |
|---|---|---|---|
Billing records and dashboards (system.billing.usage) |
Show and attribute usage by identity, job, and notebook | Reduce usage on their own; records can lag up to 24 hours | Not stated |
| Budgets and alerts | Notify about spending against a target | Stop workloads running, according to the documentation’s description of alerts | Not stated |
| Tags and serverless usage policies | Attach tags to serverless usage for attribution | Cap spend by themselves | Serverless usage policies are labeled Public Preview |
| Compute policies | Constrain which compute choices users can make | Set a dollar limit | Not stated |
| Notebook execution timeout | Limits how long one serverless notebook query runs; default 2.5 hours | Limit total weekend spend across many runs; can be raised by admins or overridden per notebook | Default set in Compute settings |
| Scale-up limits | Cap the maximum cost per notebook, job, or pipeline workload per hour | Stop new serverless workloads from launching | Not stated |
| SQL warehouse quotas | Restrict the number of serverless resources that can exist at once in a region | Stop warehouses that already exist | Not stated |
| Governance Hub cost page | Surfaces cost information within Governance Hub | Not stated | Labeled Beta |
Databricks is explicit about quotas. Its quota documentation states: “Quotas are not intended as a capacity planning mechanism and are not a general purpose way to manage or limit spend.” Treat quotas as safety rails for resource scale, not as a budget ceiling for a weekend.
Quick Recap
Best Value
What to put in place before the next weekend
- Create budgets and alerts so that a spending threshold produces a notification within the same weekend, not on Monday’s invoice review.
- Tag serverless usage so that each workload maps to an owner, using the tagging and usage policy options that your account exposes. Confirm whether usage policies are available to you, since they are in Public Preview.
- Review the 2.5-hour notebook default in Compute settings, and audit notebooks that override
spark.databricks.execution.timeout. - Schedule the billing query from the workflow above to run daily, so a spike is visible on Saturday rather than reconstructed after the fact.
- Benchmark the workload you plan to run. Databricks recommends “running and benchmarking a representative or specific workload and then analyzing the billing system table” before estimating cost, and this step gives you a per-run DBU figure to multiply by your rate.
- Keep cloud infrastructure costs in the cloud console and reconcile them against Databricks usage each month, so the two totals are never mixed.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
