Databricks pricing in Singapore vs US East: how much more, and why
Databricks serverless pricing in Singapore runs 11–26% above US East, while classic DBU rates match. What it means for a cost estimate made from Vietnam.
Open the Databricks pricing page and the numbers you see are for AWS US East (N. Virginia). That is the default region of the page, and it is the figure most people copy into a first estimate. For a company in Vietnam it is the wrong figure. The nearest region on the list is AP (Singapore), and several products cost noticeably more there.
The difference is not spread evenly. It depends on whether the product is serverless or classic, and that distinction is worth understanding before you compare quotes or choose an architecture.
The numbers
List rates on AWS, read from the Databricks pricing pages on 16 September 2026:
| Product | US East (N. Virginia) | AP (Singapore) | Difference |
|---|---|---|---|
| SQL Serverless | $0.70 / DBU | $0.88 / DBU | +26% |
| Lakehouse RT (beta) | $0.55 / DBU | $0.69 / DBU | +25% |
| Jobs Serverless | $0.35 / DBU | $0.39 / DBU | +11% |
| SQL Classic | $0.22 / DBU | $0.22 / DBU | 0 |
| Jobs Classic | $0.15 / DBU | $0.15 / DBU | 0 |
These are list prices, before any discount, and Databricks changes them without notice. Check the current figures before you put them in a budget.
Why serverless moves and classic does not
A DBU is Databricks’ unit of consumption. What a DBU pays for depends on who owns the machines.
On classic compute, the clusters run in your own cloud account. Databricks charges for DBUs, and your cloud provider charges separately for the virtual machines, disks and network underneath. The DBU rate can stay the same everywhere because the regional cost of hardware lands on your cloud bill, not on the Databricks bill.
On serverless compute, the machines belong to Databricks. The DBU rate already includes the hardware, so it carries the regional price of that hardware with it. Singapore infrastructure costs more than US East infrastructure, and the serverless rate reflects that.
This has a practical consequence: comparing a classic rate with a serverless rate is comparing two different things. $0.22 for SQL Classic looks far cheaper than $0.88 for SQL Serverless, but the classic figure excludes the machines. The fair comparison is the classic DBU cost plus the matching cloud bill, against the serverless DBU cost alone.
Regional differences do not disappear on classic compute either. They move to the cloud bill. We do not have a published figure for how much more the same virtual machine costs in Singapore than in US East, so we do not quote one here; it is a line to read off your cloud provider’s price list for the instance types you actually use.
What it does to an estimate
Take a small three-week pilot: one serverless pipeline running two hours a day, and a small serverless SQL warehouse open three hours a day, over 15 working days. Assume 2 DBU an hour for the pipeline and 12 DBU an hour for the warehouse. Those DBU rates are our assumptions, not published figures.
| Item | DBUs | US East | Singapore |
|---|---|---|---|
| Jobs Serverless | 60 | $21 | $23 |
| SQL Serverless | 540 | $378 | $475 |
| Total | $399 | $499 |
The same workload costs about a quarter more in Singapore. On a pilot, that is a hundred dollars. On a production platform running all month, the same percentage decides whether a budget holds.
The mistake to avoid is not using the wrong region once. It is anchoring the business case on US East figures and discovering the gap in the second month of production.
Three more things that change the figure
Azure is priced by Microsoft. On Azure, Databricks rates are set and published by Microsoft, per region and per tier. In Southeast Asia, Serverless SQL on Azure Premium was $0.88 per DBU, the same as AWS Singapore, on the day we checked. Other products differ, and Azure’s tiers (Standard and Premium) do not line up one-to-one with AWS tiers, so compare product by product and tier by tier.
Promotions change the bill without changing the rate. Lakehouse RT (beta) is discounted by 30% until 31 January 2027. Databricks applies this by counting fewer DBUs, not by lowering the price: a workload that would use 10 DBUs an hour is billed as 7. When the promotion ends, the same workload goes back to 10, and the invoice rises with no change on your side. Put the end date in the budget now.
Committed use lowers the rate. Databricks offers discounts for committing to usage up front. On Azure the discount for pre-purchased units is published at up to 37% for one- or three-year terms. On AWS and Google Cloud it is negotiated and not published.
How we estimate it
For clients in Vietnam we price in AP (Singapore) by default, and we say so next to every figure, with the date the rate was read. In the discovery workshop we estimate both the Databricks line and the cloud line; in the pilot we measure them from the platform’s own billing tables, so that the production budget is built on usage rather than on a pricing page. The Start stage is where this happens, and the services page shows the rates we use today.
Sources: Databricks pricing pages (databricks.com/product/pricing and product sub-pages), AWS list rates for US East (N. Virginia) and AP (Singapore), read on 16 September 2026. Azure Databricks pricing for Southeast Asia, read on 16 September 2026. Lakehouse RT promotion: Databricks Lakehouse pricing page, read on 16 October 2026. The pilot figures are arithmetic on stated assumptions, not a quote.