MongoDB Atlas vs Azure Cosmos DB (and Azure DocumentDB) 2026

Quick verdict

Short answer

Choose MongoDB Atlas when you want the actual MongoDB server with its full query language, multi-document transactions, full-text search and MongoDB Vector Search, and the freedom to run on AWS, Azure or Google Cloud (including on Azure). Choose Azure Cosmos DB when you are committed to Azure and want turnkey multi-region writes, five consistency levels, serverless or autoscale throughput and a lifetime free tier, and your application fits Cosmos DB's documented MongoDB API limits or can use the native API for NoSQL. Note Microsoft's own guidance: for migrating an existing MongoDB application or using MongoDB Query Language features, it points to Azure DocumentDB, not to Cosmos DB for MongoDB (RU).

How we know: This comparison is research-based: products, naming, compatibility statements, limits, SLAs and prices were checked against MongoDB's pricing page, Atlas SLA and documentation, and against Microsoft Learn and Microsoft's Azure Retail Prices API, on 7 November 2026, and kept consistent with our MongoDB Atlas pricing and Azure Cosmos DB guides. Compatibility and cost claims made by either vendor are labelled as theirs. We have not run workloads on either service.

MongoDB Atlas is the managed cloud service from MongoDB, Inc. It runs the MongoDB server itself on AWS, Azure or Google Cloud, in more than 125 regions according to MongoDB, including multi-cloud clusters, and bills you directly. Its Atlas Core deployment option has Free (512 MB), Flex (up to 5 GB, capped at USD 30 a month) and Dedicated tiers (M10 and up); Atlas Infinite, which separates compute from storage, is in public preview. Features include full-text search and MongoDB Vector Search on the same data. See our MongoDB Atlas pricing guide and MongoDB hosting guide.

Azure Cosmos DB is Microsoft's fully managed NoSQL and vector database. Each account uses one API: the native API for NoSQL, or compatibility APIs for MongoDB, Cassandra, Gremlin and Table. All of these are billed in request units (RUs) and share Cosmos DB's partitioning, global distribution and consistency model. Microsoft states that it does not run MongoDB databases to provide the API for MongoDB and that Cosmos DB is not affiliated with MongoDB, Inc. See our Azure Cosmos DB guide.

Naming, precisely. Microsoft now documents two different MongoDB-compatible offerings. Azure Cosmos DB for MongoDB (RU) is the request unit API on the Cosmos DB engine. Azure DocumentDB (with MongoDB compatibility) is a separate vCore-based service built on the open source DocumentDB engine (MIT licence, governed by the Linux Foundation); Microsoft says it "was previously named vCore-based Azure Cosmos DB for MongoDB", that existing clusters and connection strings did not change, that it is not Amazon DocumentDB, and that it and Cosmos DB for MongoDB (RU) "are different services". This page compares Atlas with Cosmos DB and notes where DocumentDB changes the answer.

Side by side

Aspect	MongoDB Atlas	Azure Cosmos DB
What it is	Managed MongoDB server from MongoDB, Inc.	Microsoft's multi-model NoSQL database; MongoDB is one of several compatibility APIs
Clouds	AWS, Azure and Google Cloud, including multi-cloud clusters	Azure only
MongoDB compatibility	It is MongoDB: full query language and server features for the version you run	MongoDB wire protocol with documented gaps, for example $lookup without pipeline subqueries, no text indexes, no server-side sessions
Transactions	Multi-document ACID transactions, including across collections and shards	API for MongoDB: multi-document transactions only within one unsharded collection, fixed 5-second timeout
Capacity model	Cluster tiers billed hourly (Free, Flex, Dedicated M10 and up)	Request units: standard or autoscale provisioned RU/s, or serverless per million RUs
Sharding	You choose a shard key and add shards; storage up to 4 TB per node on Dedicated tiers	Automatic server-side partitioning by the shard key; no manual sharding commands
Global distribution	Multi-region and multi-cloud clusters; Global Clusters with geographic write zones (not on Flex)	Add or remove Azure regions at any time; multi-region writes; five consistency levels
Search and vectors	Full-text search and MongoDB Vector Search on Atlas clusters	Integrated vector search in Cosmos DB; text search operator not supported in the API for MongoDB
Free option	Free cluster: 512 MB, shared compute, no backups	Lifetime free tier: 1,000 RU/s and 25 GB per subscription (not for serverless)
Uptime SLA	99.995% for M10 and above	Microsoft cites a 99.999% SLA for high-scale multi-region use and points such workloads to the API for NoSQL
Microsoft alternative	Atlas is also sold on Azure	Azure DocumentDB: vCore-based, MongoDB-compatible, 32 GB free tier, 99.995% SLA

Key differences

The real MongoDB or a compatible API

This is the deciding difference. Atlas runs MongoDB's own server, so the query language, aggregation stages, indexes, transactions and tooling behave as MongoDB documents them for your server version.

Cosmos DB for MongoDB (RU) implements the MongoDB wire protocol on top of the Cosmos DB engine. Existing drivers, mongosh and tools connect, often with only a connection string change, but Microsoft's feature support page for server version 7.0 lists gaps: $lookup with let and pipeline returns an error, $graphLookup, $bucket and mapReduce are not supported, text, 2d and hashed indexes are not supported, sparse and case-insensitive index properties are not supported, server-side sessions and users and roles are not yet supported, and write concerns sent by the client are ignored because all writes are quorum by default. Test your actual queries before committing:

// MongoDB shell: a correlated $lookup with let and pipeline.
// Supported on MongoDB Atlas; Cosmos DB for MongoDB (RU) documents that let is not supported.
db.orders.aggregate([
  { $lookup: {
      from: "customers",
      let: { cid: "$customer_id" },
      pipeline: [ { $match: { $expr: { $eq: ["$_id", "$$cid"] } } },
                  { $project: { name: 1, tier: 1 } } ],
      as: "customer" } }
])

Microsoft itself points users who want to migrate an existing MongoDB application or use MongoDB Query Language features to Azure DocumentDB, for which it states support for 99.03% of MongoDB's query, projection, update and aggregation operators and MongoDB versions 4.0 to 8.0. That figure is Microsoft's own measure.

Clusters and tiers versus request units

Atlas sells capacity as cluster tiers. You pick a tier (for example M10 with 2 vCPUs and 2 GB of RAM, or M30 with 8 GB), Atlas runs a replica set, and you pay an hourly price plus backups, data transfer, extra storage and extra nodes. Performance tuning means classic database work: indexes, the working set in RAM, and tier sizing.

Cosmos DB sells throughput in request units. Microsoft's reference point is that reading a 1 KB item by ID and partition key costs one RU; writes, queries, larger items, more indexed properties and stronger consistency cost more. You provision RU/s (standard or autoscale) or pay per RU consumed (serverless), and storage per GB in every region. Each response reports its RU charge, so you can measure cost per operation. In our view, request units suit predictable key-based access and very elastic traffic well, but make ad hoc analytical queries and scans expensive and harder to predict.

Global distribution and consistency

Cosmos DB was built for global distribution: you add or remove Azure regions on an account at any time, can enable multi-region writes, and choose one of five consistency levels (strong, bounded staleness, session, consistent prefix, eventual) as the account default, relaxing it per request. Microsoft notes that strong and bounded staleness reads cost about twice the RUs of the relaxed levels.

Atlas offers multi-region and multi-cloud replica sets for availability and local reads, and global clusters that pin data to zones by shard key. Writes go to the primary of each replica set, and consistency is controlled with MongoDB's read and write concerns. If your requirement is active-active writes in several Azure regions with tunable consistency, Cosmos DB is designed around it; if it is surviving a whole cloud provider outage, only Atlas can span providers.

Where Azure DocumentDB fits

Teams on Azure that want MongoDB semantics now have three options: Atlas on Azure, Cosmos DB for MongoDB (RU), and Azure DocumentDB. Azure DocumentDB is priced by vCores and storage (cluster tiers from M10 upwards, scaled independently), includes backups and text and vector search without separate charges, offers a free tier with 32 GB of storage, and carries a 99.995% availability SLA, according to Microsoft. Microsoft also claims it costs 40% less on average than MongoDB Atlas on AWS; that is a vendor comparison, not an independent result.

DocumentDB is not the MongoDB server either, so the same advice applies as for Cosmos DB: run Microsoft's free compatibility assessment and your own tests before migrating. Because the DocumentDB engine is open source under MIT, Microsoft notes that MongoDB's Server Side Public License does not apply to it.

Pricing and licensing

MongoDB Atlas (USD, listed on mongodb.com/pricing on 7 November 2026, AWS tab): Free at USD 0 with 512 MB of storage; Flex from USD 0.011 per hour, up to USD 30 a month, with 5 GB of storage; Dedicated from USD 0.08 per hour for an M10 (2 vCPUs, 2 GB RAM, 10 to 128 GB storage), shown as starting at USD 56.94 a month, with M20 at USD 0.20, M30 at USD 0.54, M40 at USD 1.04 and M50 at USD 2.00 per hour. Clusters are billed hourly and invoiced monthly; backups, data transfer, extra storage, nodes and regions are added on top, and prices differ by cloud provider and region.

Azure Cosmos DB (East US, USD, from Microsoft's Azure Retail Prices API on 7 November 2026, single region): standard provisioned throughput USD 0.008 per 100 RU/s per hour, which is USD 5.84 per 100 RU/s per month; autoscale USD 0.012 per 100 RU/s per hour (USD 8.76 a month, billed at the highest RU/s reached each hour); serverless USD 0.25 per million RUs; transactional storage USD 0.25 per GB per month in each region. The lifetime free tier gives one account per subscription 1,000 RU/s and 25 GB free. As arithmetic, 400 RU/s provisioned with 10 GB stored is 4 x 5.84 + 10 x 0.25 = USD 25.86 a month in one region.

Azure DocumentDB is priced per node by cluster tier (vCores and memory) plus storage per GiB, with one- and three-year reserved capacity discounts; its pricing page loads tier prices dynamically, so check the page or calculator for your region. Its free tier is one cluster per subscription with 32 GB of storage, without high availability or backups.

Pricing checked on the vendors' official pages on 7 November 2026. Prices change; confirm before buying.

Where each one leads

MongoDB Atlas strengths

The real MongoDB server, so queries, indexes and transactions behave as MongoDB documents them
Runs on AWS, Azure and Google Cloud, including multi-cloud clusters
Multi-document transactions across collections and shards
Full-text search and MongoDB Vector Search on the same cluster
Simple tier-based pricing with a free cluster and a capped Flex tier

Azure Cosmos DB strengths

Turnkey multi-region distribution with multi-region writes
Five consistency levels, settable per account and relaxable per request
Serverless and autoscale throughput for very elastic or intermittent traffic
Lifetime free tier of 1,000 RU/s and 25 GB
One Azure bill, Azure RBAC and native integration with Azure services; several APIs in one service

Limitations

MongoDB Atlas limitations

Billed by MongoDB rather than as a first-party Azure, AWS or Google Cloud service
Backups, data transfer, extra nodes and regions add noticeably to the base tier price
Free and Flex tiers have tight storage and throughput limits, and Free clusters cannot be backed up
Multi-region writes are not active-active in the Cosmos DB sense; writes go to each replica set's primary

Azure Cosmos DB limitations

API for MongoDB is a compatibility layer with documented gaps, not the MongoDB server
Multi-document transactions only within one unsharded collection, with a 5-second timeout
Request unit costs make scans and ad hoc analytical queries expensive and harder to predict
Azure only
Microsoft itself recommends Azure DocumentDB for MongoDB migrations, which means choosing between two Microsoft services

When to choose each

Choose MongoDB Atlas if

Your application uses MongoDB features such as $lookup pipelines, text indexes or transactions across collections and shards
You want to run on more than one cloud provider, or keep the option to move
You want vector and full-text search inside the same MongoDB cluster
Your team already operates MongoDB and wants the same server, tooling and documentation
You prefer sizing by cluster tier rather than by request units

Choose Azure Cosmos DB if

You are committed to Azure and want a first-party service on your Azure bill
You need multi-region writes and tunable consistency across Azure regions
Your access pattern is mostly point reads and writes by key and partition
Traffic is spiky or intermittent and serverless or autoscale RU billing fits
You are building new and can use the native API for NoSQL, which Microsoft recommends for high-scale workloads

When neither is right

You want MongoDB compatibility on Azure with vCore pricing and fewer gaps than the RU API: evaluate Azure DocumentDB, described above.
You are on AWS and want a key-value or document service from AWS itself: see DynamoDB vs MongoDB.
Your data is highly relational, or you need SQL joins and ad hoc reporting: see MongoDB vs PostgreSQL and Relational vs document database.
You need wide-column storage for very high write rates: see Cassandra vs MongoDB.

Final recommendation

Bottom line

If you want MongoDB, MongoDB Atlas is the direct choice: it is the MongoDB server, managed by the company that builds it, on whichever cloud you use, including Azure. Pick Azure Cosmos DB when you are building for Azure and its strengths matter more than full MongoDB compatibility: global distribution with multi-region writes, tunable consistency, request unit billing with serverless and autoscale, and a generous free tier, ideally on the native API for NoSQL. If you are on Azure, want MongoDB compatibility and cluster-style pricing, add Azure DocumentDB to the shortlist, since that is the service Microsoft now recommends for MongoDB migrations.

Frequently asked questions

Is Azure Cosmos DB the same as MongoDB?

No. Cosmos DB is Microsoft's own NoSQL database. Its API for MongoDB implements the MongoDB wire protocol so MongoDB drivers and tools can connect, but Microsoft states that it does not run MongoDB databases to provide the service and that Cosmos DB is not affiliated with MongoDB, Inc. Feature support is documented per server version and has gaps.

What happened to vCore-based Azure Cosmos DB for MongoDB?

Microsoft renamed it. Its documentation says Azure DocumentDB (with MongoDB compatibility) "was previously named vCore-based Azure Cosmos DB for MongoDB", that existing clusters, connection strings and APIs did not change, and that Azure DocumentDB and Azure Cosmos DB for MongoDB (RU) are different services. It is also not Amazon DocumentDB.

Can I run MongoDB Atlas on Azure?

Yes. Atlas deploys clusters on AWS, Azure and Google Cloud, and can span providers in multi-cloud clusters. You are billed by MongoDB, and Atlas prices differ by provider and region.

Which is cheaper, MongoDB Atlas or Cosmos DB?

It depends on the workload. Atlas charges by cluster tier per hour plus extras; Cosmos DB charges per RU/s provisioned or per million RUs consumed, plus storage per GB in each region. Key-based access with modest throughput can be cheap on Cosmos DB, while query-heavy workloads can consume many RUs. Microsoft claims Azure DocumentDB costs 40% less on average than Atlas on AWS; treat that as a vendor claim and price your own workload.

Does Cosmos DB for MongoDB support transactions?

Partly. Microsoft documents multi-document transactions only within a single unsharded collection, not across collections or in sharded collections, with a fixed 5-second timeout. MongoDB Atlas supports multi-document transactions across collections and shards.

Sources

Checked November 2026.

How we research comparisons: our editorial method.