The State of Enterprise Semantic Layers: A 2026 Market Overview

The enterprise semantic layer market has reached an inflection point

Where it's heading, what's driving it, and what we think enterprises should be paying attention to

Every enterprise has the same problem: data exists in abundance, but meaning is scarce.

Your data warehouse contains millions of rows, but when the CFO asks "what was revenue last quarter?" three different teams produce three different numbers. The data is correct. The definitions aren't aligned.

Semantic layers solve this. They sit between raw data and the people (or systems) that consume it, translating technical schemas into business concepts. Define "revenue" once, centrally, and every tool, every dashboard, every AI agent uses the same definition.

Two years ago, this was Palantir's differentiator and Looker's speciality. Today, every major platform has a semantic layer. Snowflake has Semantic Views. Databricks has Metric Views. Microsoft has Fabric IQ. Google has Looker. SAP has Datasphere.

The surface-level story: everyone converged on the same idea.

The real story is more interesting, and that's the subject of this article: the State of the Ontology Nation 2026.

The split that matters

Beneath the convergence, a fundamental divergence is emerging. The market is splitting into two camps.

Camp 1: Analytics consistency

Snowflake, Databricks, Looker, dbt. These platforms solve metric drift. They ensure Finance and Sales use the same revenue definition. They make data trustworthy for humans and AI to query.

They're read-only in the sense that matters here. You ask questions; you get answers. The semantic layer informs decisions. These platforms increasingly add light actioning capabilities: alerts, reverse ETL, triggered workflows. But they stop short of deep operational orchestration. The semantic model describes the business; it doesn't run it.

Snowflake's Semantic Views and Intelligence features are now generally available. Databricks' Metric Views provide governed metrics across AI and BI workloads. Looker's LookML has been doing this for years, and Google has now opened it up via SQL interfaces so external tools can query Looker-defined metrics directly.

Palantir can operate here too. Foundry SQL Server provides read-only connectivity via JDBC and ODBC, with certified integrations for Power BI, Tableau, and other BI tools. Organisations can use Contour and Quiver for governed analytics without invoking the Ontology's operational capabilities. The platform is modular—you can treat it purely as an analytics layer with governed metrics.

These are mature, production-ready options for organisations that need consistent analytics without operational complexity.

Camp 2: Workflow automation

Palantir Foundry and now Microsoft Fabric IQ. These platforms describe the business and act on it. Ontologies trigger workflows. Agents monitor conditions and take autonomous action.

Palantir's Ontology includes Actions and Functions. When an airline uses Foundry to manage fleet maintenance, the ontology schedules repairs, triggers alerts, and writes back to source systems. The semantic layer makes decisions.

The Fabric IQ announcement at Ignite 2025 is the inflection point. Microsoft has explicitly entered Camp 2. The Operations Agent watches live data, reasons over business rules, and takes action without waiting for human intervention. In a supply chain scenario, when real-time data shows traffic congestion, the agent can automatically reroute delivery trucks based on business rules defined in the ontology.

Palantir is production-proven at scale across defence, supply chain, aviation, and healthcare. Fabric IQ is in preview. But Microsoft has made its intent clear: workflow automation is now contested territory.

We think this split matters more than feature lists. Choosing a platform is choosing a philosophy. Do you want your semantic layer to inform decisions or make them?

The deployment question

I've written previously about the ontology gold rush, exploring why every major platform is now building semantic layer capabilities. That piece built on earlier work: the shift from process-first to purpose-first enterprise and the question of whether your digital twin should be alive.

The recurring theme: Palantir solved the deployment problem, not just the technology problem.

When an Ontology fails, it's typically because the organisation hasn't truly understood it. The technology works. The model doesn't reflect how the business actually operates. Forward Deployed Engineers act as a lubricant here. They are fully embedded with clients, learning the domain, and building models that reflect actual operations rather than idealised data structures. They bridge the gap between what the platform can do and what the organisation needs it to do.

The question for Microsoft: can Fabric IQ deliver workflow automation with a traditional enterprise software deployment model?

The technology looks promising. The implementation playbook is unproven. Fabric IQ lets business experts build and evolve ontology models using no-code visual tools, which lowers the barrier to entry. Organisations can jumpstart their ontology using existing Power BI semantic models. But whether this translates into production-grade workflow automation remains to be seen.

For enterprises, this creates a decision framework:

• High implementation capacity + workflow automation ambition: Foundry remains the proven choice. The investment is significant, but the outcomes are transformative when done well.

• Microsoft ecosystem + emerging operational needs: Fabric IQ offers a lower-friction path. Existing Power BI investments accelerate adoption. But it's preview software, and the operational capabilities are new.

• Analytics consistency without operational complexity: Snowflake, Databricks, Looker deliver mature options. Define metrics once, query everywhere, govern centrally. No workflow automation, but also no deployment complexity.

We've seen both approaches succeed and fail. The pattern isn't which platform. It's whether the organisation commits to building the semantic model as a living asset rather than a one-time project.

The tension we see everywhere: governance vs agility

IT wants control. Business wants speed. This tension has existed for decades, but semantic layers bring it into sharp focus.

The promise is that you define things once, centrally, and everyone benefits. But who defines? Who approves changes? How fast can the model evolve as the business changes?

We see this playing out in every client conversation. The technical decision (which platform) is often easier than the organisational decision (who owns the semantic model and how does it evolve).

Platforms are taking different approaches:

• SAP Datasphere uses a two-tier model. IT governs the foundation in a controlled layer, then shares it to business spaces where domain experts can extend the model within guardrails. This preserves the business logic embedded in decades of ERP configuration while enabling self-service.

• Fabric IQ offers no-code ontology builders so business experts can evolve the model without waiting on specialist engineers. The pitch is democratisation: the people closest to the business define how it's represented. IT secures, approves, and versions, but doesn't bottleneck.

• Palantir Foundry takes a collaborative approach. Building the Ontology involves Forward Deployed Engineers working alongside client teams, which takes time upfront. The trade-off is granular control and genuine workflow automation capability. Once built, business users interact through Workshop applications that expose complexity through intuitive interfaces.

• Snowflake and Databricks lean on familiar SQL and YAML patterns, keeping the semantic layer close to engineering teams. This works well when analytics engineering is mature, but can leave business users dependent on technical resources for changes.

The organisations getting this right are the ones who've stopped treating the semantic model as either an IT deliverable or a business wish list. They've built shared ownership: business defines meaning, engineering implements structure, and both are responsible for keeping it current. The governance model evolves with the business rather than constraining it.

The governance model matters as much as the platform choice.

Where this is heading

We're confident about three directions.

• Semantic layers become the interface for AI agents. Every client wants AI agents. Agents need semantic context to reason about the business, not just access to raw data. The platforms with strong semantic layers are becoming the foundation for agentic architectures. Whoever controls the semantic model controls how AI understands the business. Gartner’s recent research (as quoted by Neo4j) argues that "a unified semantic layer is required to enable AI agents to achieve contextual understanding and advanced reasoning." Neo4j's reflection on the 2025 Gartner Magic Quadrant captures the shift well: traditional databases store data; semantic layers provide the context that AI needs to reason accurately. This is where evals and guardrails become critical. Agents operating on semantic models will make mistakes. The question is whether you can detect those mistakes before they propagate. Structured evaluation frameworks become essential infrastructure: testing agent outputs against known-good answers, monitoring for drift, implementing circuit breakers when confidence drops. Guardrails encoded in the semantic layer itself can constrain agent behaviour, preventing hallucinations from becoming operational decisions.

• Workflow automation becomes contested. Microsoft's entry means Palantir faces real competition for the first time in the operational space. Expect acceleration from both. Foundry will likely move faster on AI-native capabilities; Fabric IQ will likely move faster on accessibility and integration with the broader Microsoft ecosystem.

• Open vs integrated remains the defining tension. Databricks and Snowflake bet on open formats and tool flexibility. Microsoft and Palantir bet on integrated platforms. SAP bets on preserving existing investments. Timbr.ai offers a headless approach for organisations that want ontology capabilities without platform commitment. All are valid strategies. None will win outright. Most enterprises will run multiple platforms and need to manage semantic consistency across them.

We're building our practice around the assumption that semantic infrastructure is foundational. The enterprises that treat it as a strategic asset, governed and evolved deliberately, will have a structural advantage in the agentic era.

For readers wanting the specifics on each platform, the reference guide at the end of this article breaks down capabilities, maturity, deployment models, and cost structures.

How we think about this

We work with Palantir. We also work with clients on Google, Databricks, Snowflake, and Microsoft stacks.

Our view: the platform matters less than the commitment. Building a semantic model that reflects how your business actually operates, keeping it current as the business evolves, and embedding it into decision-making processes. That's the hard part. The technology is the easy part.

If you're navigating this decision, we're happy to help you drive.

The players: a reference guide

The thematic argument above covers how to think about semantic layers. This section provides the reference material on each platform for readers who want the specifics.

Camp 1: Analytics consistency

Snowflake Semantic Views

Snowflake's native semantic layer, integrated with Cortex AI for natural language querying.

• Core capability: Governed metrics and dimensions defined in SQL, queryable across BI tools and AI applications

• Workflow automation: No

• Maturity: GA (2024)

• Deployment model: Self-service; analytics engineering teams define and maintain

• Cost model: Consumption-based (compute + storage)

• Best fit: Organisations with Snowflake as primary warehouse wanting consistent metrics without additional tooling

Databricks Metric Views

Unity Catalog's semantic layer providing governed metrics across the Databricks lakehouse.

• Core capability: Metric definitions in Unity Catalog with lineage, access controls, and AI/BI integration

• Workflow automation: No

• Maturity: GA (2024)

• Deployment model: Engineering-led; integrates with existing Databricks workflows

• Cost model: Consumption-based (DBUs)

• Best fit: Organisations standardised on Databricks wanting governed metrics within the lakehouse architecture

Looker / LookML

Google's semantic modelling language, now accessible via SQL interfaces for external tool connectivity.

• Core capability: Declarative data modelling with metrics, dimensions, and relationships; SQL interface for external access

• Workflow automation: No

• Maturity: Established (10+ years)

• Deployment model: Engineering-led; LookML requires technical resources to author and maintain

• Cost model: Licence-based (per user)

• Best fit: Organisations wanting a mature, well-documented semantic layer with broad BI tool compatibility

dbt Semantic Layer

MetricFlow-powered semantic layer for dbt projects, enabling governed metrics accessible via APIs and integrations.

• Core capability: Metrics defined in YAML alongside dbt models; queryable via semantic layer APIs

• Workflow automation: No

• Maturity: GA (2023)

• Deployment model: Engineering-led; extends existing dbt workflows

• Cost model: dbt Cloud subscription tiers

• Best fit: Organisations already using dbt wanting semantic consistency without platform migration

Camp 2: Workflow automation

Palantir Foundry

Integrated data platform with Ontology layer enabling both analytics and operational automation.

• Core capability: Object-centric data model with Actions, Functions, and real-time operational workflows

• Workflow automation: Yes (mature, production-proven)

• Maturity: Established (10+ years in production at scale)

• Deployment model: FDE-embedded or partner-led; requires significant implementation investment

• Cost model: Platform licence + implementation services

• Best fit: Organisations with complex operational workflows, high-stakes decision environments, and capacity for deep implementation

Microsoft Fabric IQ

Microsoft's operational ontology layer with autonomous agent capabilities, announced at Ignite 2025.

• Core capability: Business ontology with Operations Agent for automated decision-making and workflow execution

• Workflow automation: Yes (emerging)

• Maturity: Preview (2025)

• Deployment model: Self-service with no-code ontology builder; leverages existing Power BI semantic models

• Cost model: Fabric capacity units (expected)

• Best fit: Microsoft-native organisations wanting operational automation with lower implementation friction than Foundry

Other players

SAP Datasphere

SAP's data fabric with semantic layer preserving ERP business logic.

• Core capability: Two-tier model with governed foundation and business-managed extension spaces

• Workflow automation: Limited (via SAP BTP integration)

• Maturity: GA (2023)

• Deployment model: IT-governed foundation with self-service business spaces

• Cost model: Capacity units + data integration credits

• Best fit: SAP-centric enterprises wanting to expose ERP semantics to modern analytics and AI

Timbr.ai

Headless semantic layer enabling ontology capabilities across existing data infrastructure.

• Core capability: Virtual ontology layer over existing databases without data movement

• Workflow automation: No

• Maturity: GA (2022)

• Deployment model: Engineering-led; connects to existing data sources

• Cost model: Subscription-based

• Best fit: Organisations wanting ontology capabilities without committing to a platform migration

Neo4j

Graph database with knowledge graph capabilities increasingly positioned as semantic infrastructure for AI.

• Core capability: Native graph storage with semantic relationships; GraphRAG integration for AI grounding

• Workflow automation: No (but enables agent reasoning via knowledge graphs)

• Maturity: Established (10+ years)

• Deployment model: Engineering-led; AuraDB for managed cloud

• Cost model: Consumption-based (AuraDB) or self-hosted licence

• Best fit: Organisations building knowledge graphs as the semantic foundation for AI applications

Valliance is an AI consultancy specialising in enterprise transformation. We help organisations evaluate, implement, and operationalise data platforms including Palantir Foundry.