Cognite: our review
Cognite Data Fusion connects the sensors, ERP, and historians of a plant or industrial site into a single data model, so AI applications can use it without rebuilding connectors every time. The Norwegian vendor just agreed to a buyout by Schneider Electric, announced in late June 2026.
What we take away from it
What counts in its favor
- Certified ISO 27001, ISO 9001, ISO 27018, and audited SOC 2 Type 2 and Type 3, with audit reports available — rare at this level of detail from a vendor in this sector.
- Explicit public documentation: exchanges with the in-house AI (Cognite AI) aren't used to train models and aren't shared with third parties.
- Over $170 million in 2025 revenue and 36% growth in recurring bookings, a sign of a product in real use at large industrial accounts rather than a research project.
What counts against it
- Acquisition by Schneider Electric announced on June 30, 2026 for $3.1 billion, still subject to regulatory approval: the brand and product roadmap could change once integration with AVEVA takes effect.
- No public pricing page: the model is a quote after a demo, with no order of magnitude disclosed.
- The platform relies on public cloud providers (Google Cloud, Microsoft Azure): no fully on-premises hosting option appears in the public documentation.
A serious, widely certified industrial data platform, but don't commit without first asking the vendor what the Schneider Electric acquisition will change for your contracts and your data going forward.
The facts
- Data hosting
- Hosting not verified
- Categories
- Manufacturing,Data analysis
- Data verified on
- 30 July 2026
What happens to your data
- Is your data used to train their models?
- Subprocessors
- The vendor explicitly states that prompts and responses from Cognite AI aren't used to train a model. The privacy policy also names Google Cloud Platform and Microsoft Azure as hosts, along with HubSpot, Google Analytics, and several ad networks for website data; no exhaustive list of platform-side subprocessors is published.
- Retention
- Personal data tied to a contract is deleted once contractual obligations are fulfilled; client information is kept for five years under Norwegian accounting law.
Cognite in detail
Cognite publishes Cognite Data Fusion, a platform that pulls together the scattered data of an industrial site — sensors, production history, ERP, technical drawings — into a single model, so an application or an AI agent can then query it without rebuilding the connectors for every project. The vendor is Norwegian, and at the end of June 2026 accepted a buyout by the Schneider Electric group for $3.1 billion.
What it actually does
A factory or an oil platform accumulates data across dozens of systems that don't talk to each other: PLCs, maintenance histories, ERP, CAD drawings. Cognite Data Fusion builds a contextual model that links these sources together, then exposes it all to business applications (shutdown planning, robotics, conversational agents through Atlas AI) without every team rebuilding its own integration. The vendor claims more than $170 million in 2025 revenue and 36% growth in recurring bookings, a scale well beyond a pilot project.
Why it interests us
This is a clear case for the question that shapes industrial purchasing: where does the computing run, and where does the data go. Here, the documented answer is the public cloud — Google Cloud Platform and Microsoft Azure are named as hosts — with no publicly visible on-premises hosting option. In exchange, the certification level is high for this sector: ISO 27001, ISO 9001, ISO 27018, SOC 2 Type 2 and Type 3, with a written commitment that exchanges with Cognite AI aren't used to train models. That's the kind of guarantee we look for before recommending a data provider to an industrial operator; see our AI for industry page. On the actual AI share: Cognite Data Fusion remains primarily a data-structuring and contextualisation product (industrial "DataOps"), on top of which generative-AI building blocks (Atlas AI) are added — the product's value comes as much from the data plumbing as from the models themselves.
Where it falls short
The most important reservation in 2026 concerns the buyout: on 30 June 2026, Schneider Electric announced its intention to acquire Cognite for $3.1 billion and fold it into AVEVA, its industrial-software subsidiary. The deal was still subject to regulatory approval as of this profile's verification date; neither the eventual brand nor the exact fate of existing contracts is detailed publicly. Second, as is often the case in this sector, no pricing is published: sales go through a demo followed by a quote. Finally, the platform depends on public cloud providers, with no documented local-hosting option — a point that can matter for a site subject to data-sovereignty rules.
Our recommendation
A solid, well-certified technical platform for a large industrial site that already has sensors and an ERP but no common data model to plug AI into. Before signing, get in writing what the Schneider Electric acquisition changes for your contract, and check that going through a public cloud is compatible with your data-sovereignty rules.
Who it's for, and who it isn't
Suited for
- A large industrial site (energy, continuous production, power grid) that already has sensors and an ERP but no unified data model to plug AI into
- An organization whose compliance requirements (ISO, SOC 2) are non-negotiable for any industrial data vendor
Not a fit
- A small business or single site with no dedicated data team: the product assumes heavy integration of existing systems, not a self-service setup
- An organization signing a multi-year contract now that wants clear visibility on who will run the platform in two years
Frequently asked questions
- Does Cognite run on our servers or in the cloud?
- In the cloud: the privacy policy names Google Cloud Platform and Microsoft Azure as hosts. No option for hosting fully on the client's own infrastructure appears in the public documentation reviewed.
- Does the Schneider Electric acquisition change anything for a client today?
- The deal, announced on June 30, 2026 for $3.1 billion, wasn't finalized as of this review: it remains subject to regulatory approval. Schneider Electric has said it wants to integrate Cognite into its AVEVA subsidiary; what that means in practice for existing contracts isn't detailed publicly, and should be clarified before any long-term commitment.
- Does my data train Cognite's AI models?
- No, according to the vendor's documentation: prompts and responses exchanged with Cognite AI aren't used for training and aren't shared with third parties.
- How much does Cognite Data Fusion cost?
- No pricing is published. Access goes through a demo and then a quote, a common model for this type of platform sold into large industrial installations.
- What's the difference with Cosmo Tech or Sight Machine?
- Cognite structures a plant's raw data (sensors, ERP, historians) so AI applications can plug into it. Sight Machine does similar work on the production-analytics side. Cosmo Tech, meanwhile, simulates a system before a decision is made, rather than analyzing what's already happening.
Alternatives to Cognite
Manufacturing · Data analysis
Sight Machine
Digital twin of production, to spot where output is being lost.
Manufacturing · Data analysis
Braincube
Industrial process optimization from machine data.
Manufacturing · Data analysis
Cosmo Tech
French vendor of AI-simulated digital twins to optimize production and the supply chain of large industrial firms.
Our sources
- Security pageaccessed on 30 July 2026
- General privacy policyaccessed on 30 July 2026
- Trust center (documentation)accessed on 30 July 2026
- Announcement of the Schneider Electric acquisitionaccessed on 30 July 2026
Prices and hosting terms change often. We recheck them periodically, but the publisher's page remains the only up-to-date reference.