| Ideal workload profile | Most workloads, especially spiky, AI-driven, agentic, or high-growth applications where demand can change without warning. | Predictable, steady-state workloads with well-understood, consistent traffic patterns.
|
| Compute and storage relationship | Compute and storage scale independently and rapidly, automatically absorbing sudden spikes in reads and writes.
| Compute and storage are provisioned together in an attached configuration as fixed cluster sizes.
|
| Storage capacity per high availability (HA) shard
| Up to 128 TB per HA shard to support up to petabyte-scale clusters.
| Up to 4TB per HA shard, sized for steady, well-scoped datasets.
|
| Elasticity and scaling speed | Rapid vertical and horizontal scaling designed to absorb unexpected bursts in traffic.
| Vertical and horizontal scaling supported, sized and executed as planned resize operations.
|
| Security model | Multi-tenant architecture with dedicated compute in your VPC plus zero-knowledge storage where data and metadata are encrypted before reaching the shared layer.
| Single-tenant clusters with strong isolation and standard Atlas encryption.
|
| Background operations impact | All background operations, including backups, restores, replication, and maintenance, run on the storage layer, so they never compete with live traffic.
| Background operations share cluster resources and should be scheduled around peak times.
|
| Financial model | Available on Dedicated clusters. Storage is purely consumption-based, metered by actual data stored. Compute runs on a reserved instance tier, billed for the running shard, not query volume.
| Provision-based: capacity reserved up front for consistent monthly charges. Free and flex tier clusters are also available.
|
| Multi-cloud portability | Available on AWS today, with additional clouds planned.
| Run a single cluster across AWS, Microsoft Azure, and Google Cloud.
|
| Developer experience | Same MongoDB API, drivers, and tooling as Atlas Core, ready for AI and large-scale growth.
| Familiar Atlas experience focused on steady-state workloads.
|
| Search and Vector Search | Supported. Search and Vector Search run natively on Atlas Infinite with no separate search engine or vector database to manage. | Supported. Search and Vector Search run natively on Atlas Core with no separate search engine or vector database to manage. |
| Limitations | Some Atlas features are not available during the Atlas Infinite public preview. See the full list in the documentation. | No limitations except for new Atlas features that may be in a public preview state.
|