- BFSIBanking, Financial Services & Insurance
- IT & ITESIT services, BPM, GCCs & SaaS
- Retail & e-commerceStore chains, D2C brands & marketplaces
- HealthcareComing soon
- ManufacturingComing soon
Databricks Lakebase
Your app needs fresh features from the lakehouse. The export job ran last night, not now — Databricks Lakebase is managed Postgres inside your Databricks workspace — autoscaling computes that scale to zero, copy-on-write branches and synced tables to Unity Catalog, in AWS Mumbai and Azure Central India, from $0.092 per CU-hour.
Buy through TechBag
Same software. Better outcome — at a lower cost.
How it’s rated
Full scoreboard ↓Quick answer
The Databricks platform family
This page covers Databricks Lakebase — with Databricks Apps and Lakebase Search folded in. The rest:
Quick facts
30-second orientation- Product
- Managed Postgres for transactional work inside Databricks, with autoscaling, scale to zero, branching and recovery
- Maker
- Databricks, Inc., San Francisco; built on technology from Neon, which Databricks agreed to acquire in May 2025
- Status
- Public Preview from June 2025, generally available on 22 January 2026; Postgres 16, 17 (default) and 18
- Compute
- Autoscaling $0.092 per CU-hour on AWS Premium in US East, $0.104 in Mumbai; Azure $0.126 in Central India
- Offer
- 50% off Lakebase compute until 31 January 2027, according to Databricks’ Lakebase pricing page
- Storage
- 15 DSU per GB-month: $0.345 in US East, $0.375 in Mumbai and Central India; history and snapshots extra
- Size
- 0.5 to 112 Capacity Units of 2 GB RAM each; autoscaling up to 64 CU; up to 500 branches per project
- Security
- Customer-managed keys on AWS and Azure; PCI-DSS and HITRUST on AWS with a compliance security profile
- India
- AWS Mumbai and Azure Central India (Pune); on Google Cloud it is Beta, in US and EU regions only
- In India via
- TechBag — workspace and region fit, the promotion’s end date and the quote in INR with GST
Understand lakehouse Postgres before you buy it
Most product pages skip this. We start here — so you buy a capability, not a buzzword.
What is Databricks Lakebase?
It is managed Postgres inside your Databricks workspace, built for apps and agents that work beside lakehouse data.
An app database outside the platform vs Databricks Lakebase — the honest table
What consolidation actually replaces, dimension by dimension.
| Dimension | An app database outside the platform | Databricks Lakebase |
|---|---|---|
| App database | A separate Postgres outside the platform | Postgres inside the Databricks workspace |
| Data movement | Nightly exports in both directions | Synced tables to and from Unity Catalog |
| Test copies | Full clones, refreshed by hand | Copy-on-write branches, up to 500 a project |
| Idle cost | Servers billed around the clock | Computes that suspend when nobody uses them |
| Access control | Database grants managed on their own | Workspace and Unity Catalog permissions |
| What it is NOT | — | Oracle-compatible, or Postgres without Databricks |
The cheapest test is one branch per developer for a fortnight, with scale to zero on: read the CU-hours and storage before sizing production.
The four pieces of the platform
Vendors love diagrams; buyers need to know what they’re actually operating. Here’s the whole platform, demystified.
Compute
Each branch has a read-write compute that autoscales between a set minimum and maximum in Capacity Units of 2 GB RAM, and can suspend when idle.
Storage
Storage is kept apart from compute, as in Neon’s design, so branches are copy-on-write and history allows point-in-time recovery up to 30 days back.
Sync
Synced tables copy data between Unity Catalog and Postgres in either direction, run by a serverless Spark Declarative Pipeline with its own DBUs.
Clients
Databricks Apps, Agent Bricks agents and outside services connect over the Postgres protocol, APIs, CLI or SDK, under workspace permissions.
Autoscaling Postgres computes over copy-on-write storage, with synced tables to Unity Catalog — all inside your Databricks workspace.
Nine capabilities. Run, branch, connect.
Databricks Lakebase gives apps and agents a Postgres database that sits beside the lakehouse, under the same permissions.
Compute that follows load
Set a minimum and maximum in Capacity Units and the compute moves between them, up to 64 CU, with fixed sizes to 112 CU.
Idle computes suspend
Computes suspend after 60 seconds to 7 days without activity; new instances do so after 24 hours by default since May 2026.
HA and read replicas
High availability is optional per compute, and each branch can add up to six read replicas for reporting or for read scaling.
Copy-on-write branches
Up to 500 branches a project give each developer, test run or AI agent its own database copy without a full physical clone.
Point-in-time restore
History retention of up to 30 days supports point-in-time recovery, alongside automated backups and up to 10 manual snapshots.
Postgres 16, 17 and 18
Postgres 17 is the default and 18 arrived on 18 June 2026; Postgres APIs, CLI and SDKs are generally available since August 2026.
Synced with Unity Catalog
Synced tables keep Postgres and Unity Catalog in step both ways, so app records reach analytics and lakehouse features reach apps.
Lakebase Search
GA since 18 September 2026: lakebase_vector, pgvector-compatible, past a billion vectors per index, and lakebase_bm25 text search.
Databricks Apps on top
Databricks Apps hosts the app itself: Medium is 2 vCPU and 6 GB at 0.5 DBU an hour, with horizontal scaling since 29 September 2026.
Watch Databricks explain Lakebase
Three videos from Databricks’ official channel: Nikita Shamgunov’s June 2026 talk on how Lakebase works, a June 2026 branching demo and a December 2025 introduction.
Inside Lakebase: fully-managed serverless Postgres – Nikita Shamgunov, VP, Engineering, Databricks
A 9:21 talk from June 2026 by Nikita Shamgunov, Neon’s co-founder and now a Databricks VP of Engineering, on how Lakebase works.
Safe AI-Driven Development with Lakebase Branches
A 6:45 demo from June 2026 of using Lakebase branches so AI-written changes are tried on a copy, not on production.
Introduction to Lakebase: OLTP for Data Apps and AI Agents
A 9:49 introduction from December 2025 to Lakebase as the transactional database for data apps and AI agents.
Want a live, India-context walkthrough for your environment?
Book a guided demo →Most app databases sit a nightly export away from the analytics. Databricks Lakebase sits beside it.
Here’s what genuinely sets it apart — and exactly where it stops.
Postgres beside the lakehouse, not a copy away
Lakebase puts an operational Postgres in the same workspace and under the same Unity Catalog permissions as your analytics. Synced tables carry data between the two in either direction, so an app can serve features computed in the lakehouse and send its own records back, with no separate ETL estate to build and run.
Serverless economics, and a branch for every change
Computes scale between the limits you set and suspend when idle, so dev and test databases cost little between uses. Copy-on-write branches, up to 500 a project, let developers and AI agents work on a full copy safely, and point-in-time recovery reaches back as far as 30 days.
In Indian regions, with a launch discount
Lakebase runs in AWS Mumbai at $0.104 per CU-hour on Premium and in Azure Central India at $0.126, with compute at half price until 31 January 2027. Customer-managed keys cover AWS and Azure, and PCI-DSS and HITRUST apply on AWS where a compliance security profile is on.
Where it stops
It exists only inside a Databricks workspace, so it is no fit without one. Storage at $0.375 per GB-month in India is several times plain managed Postgres; synced tables bill pipeline DBUs; always-on computes follow minimum-usage rules after 24 hours. There is no Oracle compatibility layer, and Google Cloud has no India region.
The numbers behind the platform
What your Databricks Lakebase rollout looks like
Pick the first app
Choose one app or agent that reads lakehouse data and writes its own records, and list the tables to sync.
Create the project
Create a project in AWS Mumbai or Azure Central India, set autoscaling limits and scale to zero, then grant access.
Branch and sync
Give each developer a branch, set up synced tables from Unity Catalog, and measure CU-hours and storage per branch.
Harden production
Add HA or read replicas where needed, set history retention and keys, then load-test the production compute.
Commit and quote
Price the steady state after the 31 January 2027 promotion and get TechBag’s quote itemised in INR with GST.
The review scoreboard
Modelled on Gartner Peer Insights structure. *Counts and breakdowns are illustrative pending verified review collection.
Quick poll — what’s driving your evaluation?
Talk to an advisor“Our credit-scoring app reads features from the lakehouse through synced tables, so we retired a nightly export job.”
“Every pull request gets its own Lakebase branch, so testers use real-shaped data and nobody clones production by hand.”
“Dev databases scale to zero overnight, which kept the pilot bill small; the production compute we left always on.”
“Storage surprised us: at India rates it costs several times what plain managed Postgres charged us for the same data.”
“Running it in Mumbai inside our workspace meant the same Unity Catalog access reviews covered the app database too.”
“Moving an Oracle-backed app meant rewriting the PL/SQL, not porting it; Lakebase suited our new services, not that one.”
Where everyone sits — the grids
Analyst firms bury this view behind paywalls, and review-site grids score satisfaction, not fit for your estate. So here’s TechBag’s synthesis of the managed and serverless Postgres market — tap any vendor to see why it sits where it does.
TechBag Managed Postgres Grid
Execution strength vs product vision — the classic market map, minus the paywall.
GA January 2026; Databricks reports a run-rate above $100M.
Elasticity × Analytics Link
The grid nobody publishes — how elastic the database is (autoscaling, scale to zero, branching) vs how tightly it links to an analytics platform.
Autoscale to 64 CU, scale to zero, 500 branches; Unity Catalog sync.
Positions are TechBag’s illustrative synthesis of public review-platform data and vendor documentation — not a reproduction of any analyst graphic. Verify before relying on it.
Databricks Lakebase vs the managed Postgres field
Against Snowflake Postgres, Azure Database for PostgreSQL, Amazon Aurora PostgreSQL, EDB Postgres AI and Supabase — on scaling, branching, HA, analytics, vectors, price, storage and India regions.
| Dimension | Databricks Lakebase | Snowflake Postgres | Azure Database for PostgreSQL | Amazon Aurora PostgreSQL | EDB Postgres AI | Supabase |
|---|---|---|---|---|---|---|
| What it is | Postgres in workspace | Postgres from Snowflake | Azure-managed Postgres | Aurora Serverless v2 | Postgres AI platform | Postgres dev platform |
| Where it runs | Inside Databricks only | Inside Snowflake only | Azure only | AWS only | SaaS or on-premises | Supabase-hosted |
| Postgres versions | Postgres 16, 17, 18 | Postgres 16–18 | Postgres 14–18 | Postgres-compatible | Oracle-compatible option | Version set by Supabase |
| Scaling to zero | 0.5–112 CU, to zero | Fixed sizes by family | Resize by vCore tier | 0–256 ACUs | Not published | Fixed compute sizes |
| Branching and recovery | 500 branches, 30 days | Forks, 10-day PITR | Backups 7–35 days | Fast clones | Not documented | Branches, PITR add-on |
| High availability | HA, 6 read replicas | Standby doubles rate | Zone-redundant, 99.99% | Up to 15 readers | PGD active-active | Read replicas |
| Analytics link | Synced to Unity Catalog | Beside the warehouse | Build a pipeline | Zero-ETL to Redshift | Analytics on Postgres | Your own pipeline |
| Vector search | Lakebase Search | pgvector, ivfflat | pgvector extension | pgvector supported | Vector built in | pgvector on Postgres |
| Pricing model | CU-hours plus DSUs | Credits per instance-hr | vCore-hours plus disk | ACU-hours plus GB | Quoted | Plans plus usage |
| Published entry price | $0.092 per CU-hour | About $0.014 an hour | B1ms ~$18 a month | $0.18 per ACU-hour | Not published | Free, then $25/month |
| Storage price | $0.375 per GB-month | $117.76 per TB-month | $0.131 per GB-month | $0.11 per GB-month | Yours or quoted | $0.125 per GB |
| India data region | Mumbai and Pune | Mumbai and Pune | Four regions in India | Mumbai and Hyderabad | Unverified | Mumbai region |
| Lock-in and exit | Postgres out, apps in | Bill stays in Snowflake | Nothing proprietary | Aurora features stay | EDB features bind | Postgres plus extras |
| Best fit | Databricks app builders | Snowflake-first teams | Azure-first estates | AWS-native app teams | Self-hosting estates | Startups and app devs |
Which approach fits you?
Honest fit signals — because the fastest way to lose your trust is to pretend one product wins every scenario.
Choose Databricks Lakebase if…
- ✓You already run Databricks and want an app database beside the lakehouse, not a separate estate
- ✓Developers or AI agents need branch-per-change copies of a Postgres database
- ✓You need it in AWS Mumbai or Azure Central India under the same Unity Catalog permissions
Compare alternatives if…
- ✓You have no Databricks workspace — Neon, Supabase or Azure Database for PostgreSQL stand alone
- ✓Storage dominates your bill — Azure and Aurora list it at about a third of Lakebase’s India rate
- ✓You are leaving Oracle — EDB offers an Oracle compatibility layer
Do not expect…
- ✓Lakebase without a Databricks workspace, or an Indian region on Google Cloud
- ✓Cheap storage — $0.375 per GB-month in India, before history and snapshots
- ✓The 50% compute discount after 31 January 2027
Databricks Lakebase is one of 28 databases & data tools products TechBag carries. The Databases & Data Tools guide narrows them to a shortlist and shows the reasoning. →
What does keeping an outside app database in step cost you?
Drag the sliders (export jobs between your app database and the lakehouse; engineer-hour cost). Estimates model the engineering time spent building, running and fixing jobs that copy data between an app database and analytics, at an assumed 1.5 hours per job a month, with 70% of it removed by synced tables. Both figures are assumptions. Illustrative.
Loaded cost = salary + overheads per productive hour. Illustrative only — your TechBag quote models your actual environment and modules.
Three ways to consume it
Published list prices, USD, as of October 2026. Autoscaling compute is $0.092 per Capacity Unit-hour on AWS Premium in US East and $0.104 in Mumbai ($0.111 and $0.126 on Enterprise), and $0.126 on Azure in Central India, with 50% off compute until 31 January 2027. Database storage is 15 DSU per GB-month: $0.345 in US East and $0.375 in Mumbai and Central India; point-in-time history and snapshots are billed apart. Always-on computes follow minimum-usage rates after 24 hours. TechBag quotes in INR with GST.
Lakebase compute
Best for apps and agents beside lakehouse data
- $0.092/CU-hour US East, $0.104 Mumbai
- 50% off compute to 31 January 2027
- Storage $0.375/GB-month in India
+ Platform add-ons
Best for a broader rollout
- Scoped to your estate
- Add-on modules as needed
- Phased, right-sized deployment
+ Databricks Apps
Best when the app is hosted on Databricks too
- App Capacity at $0.75/DBU (AWS Premium)
- Medium app: 0.5 DBU an hour
- Five Medium apps ≈ $1,350 a month
Buy it for less — TechBag pricing beats list
Whatever the list prices above, TechBag negotiates a significantly better deal — with GST-compliant INR invoicing and local support. Ask us for your discounted quote.
Get an India-ready quote
Tell us your requirements and current tools — we’ll model it against what you spend today.
The 8 questions to ask every vendor
Take this into your next vendor call — including ours.
Do you already run a Databricks workspace on AWS or Azure? Without one, Lakebase is not an option at all.
Will the project sit in AWS Mumbai or Azure Central India? Lakebase on Google Cloud has no Indian region.
What will compute cost at full price after 31 January 2027, and is that already in the business case?
How many GB, how much history, how many snapshots? At $0.375 per GB-month in India storage can lead the bill.
Which computes will run with scale to zero off for over 24 hours, and so fall under always-on minimums?
How many synced tables will run, and what will their serverless pipeline DBUs add to the monthly bill?
Any Oracle PL/SQL in scope? Lakebase has no compatibility layer, so budget a rewrite or keep that app.
What tier, region and term does the quote cover? Ask TechBag for INR with GST, storage and Apps itemised.
Questions buyers ask
Ready to evaluate Databricks Lakebase?
Name the app and the tables it needs, or let a TechBag advisor model CU-hours and storage at Mumbai or Pune rates, before and after the promotion, and get the quote itemised in INR.
Stats, ratings, review counts and pricing are illustrative and sourced from public materials; verify before purchase.