Azure Databricks Standard to Premium: what the tier change costs you

Short answer: Azure Databricks Standard tier is gone. New workspaces have had to be Premium since April 1, 2026, and every remaining Standard workspace was upgraded to Premium automatically on October 1, 2026 (Microsoft Learn). Your VM bill didn't change, but your DBU rates did: in East US, All-Purpose Compute +38%, Jobs Compute +100%, Jobs Light Compute +214%. Depending on how much of your bill is VMs, that means a total increase of roughly 24% to 40% in our examples, before you change anything.

What happened, with dates

• April 1, 2026: all new Azure Databricks workspaces must be created on Premium (Microsoft Learn).
• October 1, 2026: "Any remaining Standard tier workspaces on October 1, 2026, will be automatically upgraded to the Premium tier" (Microsoft Learn). Microsoft's Azure Databricks pricing page now lists Premium prices only.
• AWS and Google Cloud went through the same change a year earlier, on October 1, 2025 (AWS docs, GCP docs).

If a Standard workspace was still running on October 1, it was upgraded for you, and the new rates apply from then. To confirm a workspace's tier, check the Workspaces page in your account console (Microsoft Learn).

Standard vs. Premium rates

Microsoft's pricing page no longer shows Standard prices, but the Azure Retail Prices API still returned the Standard meters as of October 17, 2026, so here are both side by side for East US, straight from today's data:

Rates vary by region. Use the price table and filter by Azure and your region.

Note that the Photon variants carry the same $/DBU as standard, so the increase is the same. Photon still burns more DBUs per hour (see Is Photon worth it?), which means Photon clusters feel the increase in more dollars.

Worked examples

Examples, not quotes. Assumptions: East US, 5 nodes (1 driver + 4 workers) of D8ds v5 (8 vCPU, 32 GB) at Microsoft's pay-as-you-go Linux price of $0.452/hour, rated at 2 DBU/hour, no Photon. The VM price doesn't change with the tier.

Example A: a shared interactive cluster, 8 hours × 22 days

The DBU line goes up 38%, but the whole bill goes up 24% ($264.00 a month) because the VMs don't change. That's why a DBU-only comparison overstates the hit for all-purpose clusters.

Example B: a nightly jobs cluster, 2 hours × 30 days

Jobs Compute is where the percentage hurts most: DBUs go up 100% and the total bill 40%. Jobs Light goes up even more in percentage terms (214%), but from a low base.

Is Azure now more expensive than AWS or Google Cloud?

Not in the way the sticker suggests. Azure Premium Jobs Compute lists at $0.30/DBU in East US, against $0.15 for AWS Premium. But Databricks says "the Premium tier on Azure Databricks corresponds to the Enterprise tier on AWS and GCP" (Databricks pricing page), and AWS Enterprise Jobs Compute is $0.20. Instances also carry different DBU ratings on each cloud. Compare total cost per workload, not $/DBU. Our pricing explainer runs the same cluster on all three clouds.

Region matters too. Premium Jobs Compute lists at $0.30/DBU in 44 of the 47 Azure regions in our data, but not everywhere: US Gov AZ, US Gov Virginia at $0.375; IN Central Jio at $0.975.

What you get for the money

Microsoft's end-of-life page lists what Premium adds over Standard, and you lose nothing (Microsoft Learn):

• Unity Catalog
• Databricks SQL
• Lakeflow pipelines
• Serverless compute
• Mosaic AI, Lakebase and Genie Code
• Managed services like Lakehouse Monitoring and predictive optimization
• Advanced enterprise security

Some of these are real cost levers, not just features. Serverless and SQL warehouses were not available to you on Standard, and for bursty or short jobs serverless can cost less than a classic cluster (see serverless vs. classic). Unity Catalog is also a prerequisite for serverless jobs (Databricks docs).

How to claw some of it back

1. Move scheduled work to job clusters. Premium Jobs Compute at $0.30/DBU is still cheaper than Standard All-Purpose was at $0.40. Moving scheduled notebooks off all-purpose clusters onto job clusters can more than offset the tier change for those workloads. Databricks says non-interactive workloads "will cost significantly less" on job compute than on all-purpose compute (Databricks docs).
2. Set auto-termination on every interactive cluster and enforce it with compute policies. Databricks suggests something like an hour (Databricks docs). Idle all-purpose clusters are now billed at the higher rate.
3. Test serverless for short, frequent or spiky jobs, which you couldn't do on Standard.
4. Revisit Photon defaults. Photon is on by default for classic compute created in the UI (Databricks docs). On jobs that don't speed up much, it's now more expensive to leave on.
5. Pre-purchase DBUs. Azure offers "up to 37% savings over pay-as-you-go DBU prices" with 1- or 3-year Databricks Commit Units (Azure pricing page).
6. Cut the VM side. Reservations, savings plans and spot apply to the VMs, which didn't go up, and they lower your total bill no matter the tier (Azure pricing page).

Check your actual impact

Query system.billing.usage for the 30 days before and after October 1, grouped by sku_name and billing_origin_product (docs). Multiply by the rates above, or join system.billing.list_prices, which keeps a dated price history per SKU (docs). Watch for clusters with heavy all-purpose usage. That's where the money is, and usually where the easiest fix is.