An open source alternative to Cognite Data Fusion, built in the open

Schneider Electric has agreed to acquire Cognite. If you hold a Cognite Data Fusion contract, you are working out what that means for renewal, for where your data lives, and for the roadmap you planned against. We build a platform in the same category, so read this page knowing that. The difference we offer is not a longer feature list. It is that everything we build, the code, the issues, the discussions and the documentation, is public, and this page is about what that changes.

The problem

What both Cognite and IntelliStream fix for you

From 100 data points, only 5 become actionable insight. Both platforms exist to change that split. Illustrative split based on Splunk 2019, Seagate/IDC 2020 and HBR 2017.

The difference

Same category, built in the open

What an operational data platform has to do

Cognite Data Fusion is more mature and has more features than we do. More than a decade of real deployments, deep vertical tooling, and a support organisation we do not match. That is simply true, and a page that pretended otherwise would deserve your scepticism.

So the case for us is not the feature list. It is the structure. The same shape of platform, one connected model of assets, processes and business context with time-series, events and lineage attached, developed as a public open source product rather than behind a licence wall.

The bet we are making

The platform is open source under AGPL-3.0, free to self-host, standard components throughout, and yours to fork if we ever disappear. Our bet is that an open platform accumulates what a closed one has to buy: more installations, more running hours, more contributors, and features that close the gap over time because anyone who needs one can build it.

That is a strategy, not a promise, and you do not have to take our word for its progress. The repository, the issue tracker and the release history are public, so you can watch whether it is working.

Every answer is public, including for your AI agents

The code, the documentation, the issue history and every pull request are public. That is exactly the foundation a language model needs, so ask any capable AI assistant about IntelliStream DataHub and it reasons from the same sources our own engineers use, from the architecture down to a single configuration flag.

It changes what support means, too. Most questions are answered in the repository before you ever talk to us, and every question answered there stays public for the next team, and the next model, to learn from. Knowledge about a closed platform sits behind a support portal, where no assistant can reach it.

The numbers

Measured on our setup, reproducible on yours

~100x

time-series compression

what we routinely see

~25x

event-data compression

on real operational events

10B+

rows per second

what the query engine reads in our setup

2 PB

storage we own

40 servers in Stavanger, no rented capacity

AGPL-3.0

the whole platform

every line public, yours to fork

Familiar ground

The same protocols, and a familiar API

A different platform does not have to mean starting over, and a migration is a smaller project than it used to be. The instrumentation side speaks the standards your plant already uses, the API is designed so that experience with Cognite Data Fusion transfers, and the integration plumbing in between has become agent work.

OPC-UA, Modbus and similar standards

IntelliStream DataHub connects to OPC-UA, Modbus and similar instrument standards, just as Cognite Data Fusion does. Getting data off your instruments and control systems is standard protocol work, not a proprietary integration project.

An API inspired by the CDF API

The DataHub API is deliberately inspired by the Cognite Data Fusion API. Assets, time-series, events, files: if your team has written code against CDF, the shape of ours will look familiar, and much of that integration thinking carries over. The API reference and the SDKs for Java, Python and Rust are open, so you can check that claim before you talk to anyone.

Agents change what integrations cost

Integration work has always been more code plumbing than computer science. You understand the data contract at the source, the contract at the destination, and write the mapping between them. That is precisely the job AI agents do better than people, and it has made data integrations almost free to build and maintain. The plumbing that used to be the expensive part of adopting a platform no longer is.

Questions we get asked

Straight answers, including the unflattering ones

Is there an open source alternative to Cognite Data Fusion?

Yes. IntelliStream DataHub is an operational data platform released under AGPL-3.0. It covers the same core ground: one connected model of assets, processes and business context, with time-series and events linked to it and lineage on every derived value. It does not match the enterprise platforms on vertical breadth or on support organisation, and this page says exactly where.

Is IntelliStream a drop-in replacement for Cognite Data Fusion?

No, and you should be sceptical of anyone who says their platform is. The data itself moves without much drama, and agents have made the integration plumbing cheap. The asset and process model is the real work of a migration, because it encodes years of decisions and has to be restated deliberately. Plan it as a project with one question you genuinely answer first, not as a switch you flip.

Can we keep our operational data in Norway?

Yes. You can run the platform in your own environment, air-gapped if you need that, or on infrastructure we own and operate in Stavanger. IntelliStream is a Norwegian company with a Norwegian engineering team and no United States entity anywhere in its structure, so the CLOUD Act obligations that attach to United States providers do not attach to us.

What happens if IntelliStream stops existing?

You keep the platform. AGPL-3.0 means you already hold the source, the right to run it and the right to fork it, and your data sits in standard components rather than a proprietary format. That is the practical answer to vendor risk, and it is the same answer whether the risk is a small supplier disappearing or a large one being acquired.

Start by reading the code

The quickest way to find out whether this fits is to look at it. The platform is open source under AGPL-3.0, so you can read it, run it in your own environment and form a view without speaking to us at all. When you do want a conversation, you reach an engineer rather than a queue.

See the open source platform → Talk to us See the playground

Cognite, Cognite Data Fusion, Schneider Electric and AVEVA are trademarks or registered trademarks of their respective owners. IntelliStream is not affiliated with, endorsed by or sponsored by Cognite or Schneider Electric. The acquisition facts on this page are as the companies announced them on 30 June 2026, and the transaction had not closed when this page was last updated on 25 August 2026.