The ontology gold rush and why everyone’s building semantic layers

A response to the Palantir vs Microsoft Fabric IQ discourse:

TLDR: Palantir Technologies has been 10 years ahead because…

Nicolas Daveau's recent post on Palantir's decade-long head start sparked the predictable response: Microsoft is playing catch-up, Fabric IQ copies the airline ontology example, and traditional enterprise software vendors can't replicate Palantir's Forward Deployed Engineering model.

He's right. And he's missing the bigger picture.

The Market Has Fractured

Every major data platform now offers some form of semantic layer or ontology capability. Snowflake launched Semantic Views. Databricks embedded Metric Views in Unity Catalog. Microsoft announced Fabric IQ with operational ontologies and autonomous agents. Google opened Looker's LookML to external tools.

They're all addressing the same problem: raw data is useless without business context. A column named amt_usd_net means nothing until someone defines it as "Net Revenue after discounts, excluding returns, in reporting currency."

But here's what Daveau’s narrative slightly misses: these platforms aren't all solving the same problem.

Three Distinct Problem Spaces

1. Analytics Consistency Snowflake, Databricks, and Looker primarily solve metric drift. When Finance calculates revenue differently from Sales, you get boardroom arguments instead of decisions. These semantic layers enforce a single definition, queryable by any tool. Snowflake's Semantic Views and Intelligence features are now generally available; Databricks' Unity Catalog Metric Views provide governed metrics across AI and BI workloads.

2. Workflow Automation Palantir's Ontology goes further. It includes Actions and Functions: the ability to trigger workflows, not just read data. So called “Kinetics”. When Airbus uses Skywise to manage maintenance schedules, the ontology describes both the aircraft fleet and crucially, it also it operates on it.

Microsoft is now competing directly in this space. At Ignite 2025, they announced Fabric IQ with an Operations Agent that continuously monitors business data in real time, reasons over live conditions, evaluates trade-offs, and automatically takes actions to advance business outcomes. Fabric IQ is in preview; Foundry is production-proven (I should of course clarify that I mean Palantir’s Foundry). But the gap is closing.

The Deployment Model Is The Product

Daveau's core insight holds: Palantir's success comes from Forward Deployed Engineers who embed with clients, learn the domain, and build ontologies that reflect actual operations rather than idealised data models.

Microsoft, Snowflake, and Databricks sell software. Palantir sells transformation delivered through software. These are different businesses.

But that distinction cuts both ways. The FDE model doesn't easily scale to mid-market and nor do Palantir really want it to. It requires a new level of involvement from operational decision makers that’s not been required before or in other paradigms. And it assumes the vendor's embedded engineers will eventually leave, transferring knowledge to client teams who may not have the depth to maintain what was built.

For enterprises that need analytics consistency without operational workflow integration, the lighter-touch semantic layers from Snowflake or Databricks may be the right tool. For those who want the full operational ontology, Palantir remains unmatched. That being said, the implementation cost and commitment are naturally, proportionally higher.

What this means for enterprise buyers

The ontology and semantic layer market is no longer a Palantir monopoly. It's a spectrum:

• Analytics consistency: Snowflake Semantic Views, Looker, dbt Semantic Layer, Databricks Metric Views

• Operational intelligence (emerging): Microsoft Fabric IQ

• Operational intelligence (mature): Palantir Foundry

Choosing between them isn't about who invented ontologies first. It's about matching the solution to your actual problem, and being honest about your deployment capacity.

Next: A comprehensive comparison of semantic layer and ontology platforms. I dig into what each actually does, where they overlap, and how to choose.