Best Lakehouse Analytics Platforms (2026): Top 10 Compared

Databricks describes its platform as a lakehouse. Tables are stored in Delta Lake format in the buyer's cloud object storage, and the same data serves Spark data engineering, Databricks SQL warehouses for BI, streaming jobs and machine learning with MLflow. The vendor's lakehouse page stresses that data stays in open formats under the buyer's control.

Billing is by the Databricks Unit, a normalised unit of processing. Each workload type has its own DBU rate, so jobs compute costs less per DBU than serverless SQL. Microsoft's published Azure Databricks Premium rates in East US show the spread: $0.30 for jobs, $0.55 for all-purpose and SQL Pro, and $0.70 for serverless SQL. On classic compute, the cloud provider also bills the instances.

Databricks also appears on the StatWharf data warehouse page, through Databricks SQL.

Dremio is a lakehouse platform built around Apache Iceberg, Apache Arrow and Apache Polaris. Its query engine runs SQL directly on Iceberg tables in object storage, and Reflections accelerate repeated queries without copies in a separate warehouse. Open Catalog, built on Polaris, lets other Iceberg engines read the same tables, and an AI Semantic Layer and AI Agent are listed as core capabilities.

Dremio Cloud is fully managed on AWS at $0.20 per DCU. The DCU measures engine runtime, so a 20-DCU engine running for ten hours consumes 200 DCUs; the vendor states that infrastructure costs are included. Dremio Enterprise is self-managed for buyers with on-premises or strict compliance needs, and is priced by sales.

In May 2026 SAP announced an agreement to acquire Dremio, and Dremio's site now states that it is part of SAP.

Starburst sells commercial editions of Trino, the open-source distributed SQL engine. Starburst Galaxy is the fully managed service, which the documentation describes as a data lake analytics platform that handles provisioning, maintenance and security. Starburst Enterprise is the self-managed edition for buyers who run their own infrastructure.

Galaxy prices compute in credits. The credit is the same in every tier, so a two-worker cluster uses 12 credits an hour whether the account is on Pro or Mission-Critical; what changes is the price per credit, from $0.50 to $1.00, and the features and support attached. Pro adds flexible execution modes and streaming ingest. Enterprise adds autoscaling, ABAC and SCIM access controls and PrivateLink. Mission-Critical adds elite support and the highest uptime commitments.

The free tier stays available after the trial, with three clusters for ad hoc queries.

Microsoft Fabric is a SaaS analytics suite in which the lakehouse is one item type among several. Microsoft's documentation describes a Fabric lakehouse as combining data lake scale with warehouse querying: data is stored once in OneLake in Delta Lake format, processed with Apache Spark in Python, Scala, SQL or R, and queried with T-SQL and Power BI.

All Fabric workloads draw on one capacity. Azure's retail price list shows $0.18 per capacity-unit hour pay-as-you-go in East US, and reservations save about 41%. Storage is billed separately in OneLake by tier, from $0.026 per GB a month hot down to $0.004 cold. Autoscale Billing for Spark moves Spark onto its own pay-as-you-go meter.

Fabric also appears on the data warehouse page through Fabric Data Warehouse.

Snowflake began as a cloud data warehouse and now supports Apache Iceberg tables as its route into the lakehouse. Its documentation describes Iceberg tables as combining the performance and query semantics of standard Snowflake tables with external cloud storage the buyer manages. They suit existing data lakes that a team cannot, or prefers not to, load into Snowflake storage.

Two storage choices exist. With Snowflake-managed storage, Snowflake stores the files, protects permanent tables with Fail-safe and bills for storage. With an external volume, files sit in the buyer's Amazon S3, Google Cloud Storage or Azure bucket, the cloud provider bills storage and the buyer handles recovery. Snowflake can be the Iceberg catalog, or a catalog integration connects it to an external one.

Compute is billed in credits, with rates set by edition and region. Snowflake also appears on the data warehouse page.

Cloudera's Open Data Lakehouse is built on Apache Iceberg and aimed at enterprises with data in both public clouds and their own data centres. The product page describes unified security, governance, metadata and lineage across both, with built-in ingestion and pipelines and zero-copy sharing between business units.

Cloudera Data Warehouse provides the SQL layer. Its documentation covers Apache Hive, Apache Impala and Trino in virtual warehouses, plus the Hue editor. Data engineering runs on Spark, and Data Hub manages clusters for Spark, Hive, Impala, Kafka, Flink and other open-source services.

Pricing on cloud is hourly per CCU, with a separate rate for each service, from $0.04 for Data Hub to $0.20 for all-purpose data engineering. Cloudera states that on-premises annual subscriptions are aligned with those rates for hybrid use.

IBM describes watsonx.data as a hybrid, open data lakehouse for analytics and AI. It reads and writes Apache Iceberg, Hive and Delta Lake tables on Amazon S3 and IBM Cloud Object Storage, and applies consistent access controls, policies and lineage across them.

Instead of a single engine, buyers provision the engines they need. The pricing page lists Presto in C++ and Java versions for SQL, Spark with the Gluten accelerator for data processing, Milvus for vector search and Cassandra for operational data, each sized in resource units an hour.

The managed service on AWS starts at $1.00 per RU pay-as-you-go, metered per second, or $83.30 a month on subscription. A core support charge of 3.00 RUs an hour is added per account. Self-managed software for on-premises use is also offered.

Onehouse was founded by the team behind Apache Hudi and Apache XTable, and sells a lakehouse platform that runs in the buyer's own cloud account or Kubernetes cluster. Data never leaves that account, which lets buyers keep cloud discounts they have already negotiated.

The platform has several parts. Oneflow ingests from cloud storage, database change data capture and streams. Quanton runs Spark and SQL ETL. Table Optimizer handles compaction and maintenance. Open Engines runs managed Trino, Ray and other clusters. Lakegres adds a Postgres-compatible, autoscaling SQL endpoint for interactive and agent-driven queries on Hudi and Iceberg tables. OneSync keeps one copy of data readable as Hudi, Iceberg and Delta in every catalog.

Onehouse states pricing as pay-as-you-go or a flat yearly fee, without published rates.

e6data describes itself as a lakehouse compute engine. Its Query Engine is a Kubernetes-native SQL engine that runs inside the buyer's cloud account and reads Iceberg, Delta Lake, Hudi and Hive tables where they already live. It connects through Python, JDBC, the PostgreSQL protocol and a REST API, and the documentation covers catalog registration, row filters and column masking.

The main distinction is sizing. Instead of fixed cluster sizes, e6data allocates compute in single vCPUs, so cost follows load. The engine sits beneath existing platforms such as Databricks, Snowflake and Trino rather than replacing them, and deploys on AWS, Google Cloud, Azure or on premises.

Pay-as-you-go is $0.175 per vCPU-hour, all infrastructure included. Buyers moving from a warehouse can instead sign an outcome-based contract that guarantees a percentage saving.

Qlik Open Lakehouse is a feature of Qlik Talend Cloud rather than a standalone product. Qlik's documentation describes it as a fully managed service to ingest, process and optimise data in an Iceberg-based lakehouse, built on AWS EC2 and S3 in the buyer's account.

A data movement gateway captures changes from databases, SAP and mainframes and lands them in S3. Lakehouse clusters, EC2 auto-scaling groups that Qlik manages on Spot Instances by default, process that data into Iceberg tables registered in a catalog such as AWS Glue. The Adaptive Iceberg Optimizer then compacts and cleans the tables continuously.

Querying happens in other engines. Qlik lists Snowflake, Amazon Athena, Apache Spark, Trino and Presto, and offers zero-copy mirroring into Snowflake. Pricing follows the Qlik Talend Cloud plans, from Standard upward, by quote.