Comparison
How to choose a data & AI platform
Five questions decide the choice, well before the feature list. The table answers them for graal and for the solutions we are most often compared with, in their own words, dated and sourced.
The criteria
Five questions that decide the choice
They matter more than the feature list, and answering them takes one meeting.
01
Where must the platform run?
In your datacenter, in your cloud, with a qualified hosting provider, or in the vendor’s cloud.
02
Who operates the control plane?
You, or the vendor: the answer determines the applicable law, reversibility and your conversations with your CISO.
03
How far do you need to go?
Exploring data, or taking workloads all the way to production: scheduled, governed, monitored.
04
What can your AI agents do?
Read, or act: create, run, schedule. And under which permissions, with which record.
05
How does the price evolve?
With usage, with the number of users, or on a fixed unit that you cap.
The table
What each one publishes
Every competitor cell links to its source and to the date we accessed it. Quotes stay in their original language.
| Criterion | graal | Databricks | Scaleway (formerly Saagie) | Onyxia | Palantir / ChapsVision |
|---|---|---|---|---|---|
| Where the platform runs | Your Kubernetes or OpenShift: on-premises, in a private cloud or with the hosting provider of your choice, with no cluster administration rights. | AWS, Azure and Google Cloud. “The control plane is located in the Databricks account, not your cloud account.” | Scaleway cloud: managed Data Orchestrator, in beta, available in the Paris region. | Any Kubernetes, through Helm. | Palantir: Palantir Cloud, or software installed on the customer’s infrastructure or cloud instance. ArgonOS: on-premises, public or private cloud, sovereign clouds, air-gapped environments. |
| Vendor | GRAAL.SYSTEMS SAS, Paris. | Databricks, San Francisco. | Scaleway (iliad group), which acquired Saagie’s assets on 20/06/2025. | Developed by INSEE, supported by DINUM; MIT license. | Palantir: United States. ChapsVision: France, more than 1,100 employees. |
| Promise, in their own words | From raw data to AI agents in production, on your infrastructure. | “the Data and AI company” | Pipelines built with drag-and-drop, or coded in YAML or Python. | “the glue between multiple open source backend technologies”, for data scientists. | Foundry: “foundational data operations platform”. ArgonOS: collection, preparation, modeling and visualization. |
| AI agents (MCP) | Built-in MCP server: your agents create, run and schedule under a service account scoped to one project, with human approval for sensitive actions. No destructive tool enabled by default. | Databricks-managed MCP servers, including Databricks SQL and Unity Catalog functions, in Public Preview. | The Data Orchestrator product page does not describe an MCP server. | No MCP server documented in the project repository. | Palantir MCP and Ontology MCP documented, with no published availability status. |
| Public pricing | Annual per-vCPU license, unlimited users, no usage-based billing. | Pay as you go, in DBUs, per second. | No price published on the product page. | Free (MIT license). | No public pricing. The budget of the DGSI’s OTDH project is estimated at around €40 million. |
A cell older than 90 days is re-checked or removed. Has something changed? Write to contact@graal.systems and we will correct it.
Depending on your situation
Each solution has its ground
Your data can go to a SaaS and your budget follows consumption
A SaaS platform such as Databricks is a natural choice, rich and quick to start.
Your teams explore data and share notebooks
A datalab such as Onyxia serves exploration very well, and it coexists with graal: Onyxia to explore, graal to industrialize.
You are looking for a managed offer in a French cloud
Scaleway offers its Data Orchestrator, in beta, in its Paris region.
Your data stays with you, all the way to production, and your AI agents act under control
This is the case graal is designed for: a complete platform, installed on your infrastructure.
Discover the platformCompare on your own data
The 8-week pilot puts graal to the test on your use case, in your infrastructure.