TILIO
tilio · explore · the platform landscape

Three platforms, one market.
Here's what the platforms won't show you.

The data platform market is converging fast. Databricks, Snowflake, and Microsoft Fabric now compete across the same categories. The features converge. The practices that make them work are still the hard part.

Convergence

The features are converging

Six categories where all three platforms now compete head-to-head. When features converge, the differentiator shifts from platform to practices.

Technology Approach Timeline
Databricks Delta Lake primary + UniForm for Iceberg/Hudi compatibility UniForm GA 2024
Snowflake Native Iceberg Tables + Polaris Catalog (open source) GA June 2024, Polaris OSS July 2024
Fabric OneLake shortcuts + Iceberg interop via mirroring Announced May 2024
Tilio take

Open format is commodity. The catalog and governance layer is the new battleground. The practices around Iceberg, not the format itself, are where the real decisions live.

Technology AI Brand Core Approach
Databricks Mosaic AI ML-native (MLflow heritage), compound AI, agent frameworks
Snowflake Cortex AI SQL-callable LLM functions, model marketplace
Fabric Copilot / IQ Copilot across workloads, IQ ontology layer
Tilio take

AI features are table stakes. The differentiator is data readiness: semantic models, business context, documentation, lineage. All three AI products assume context that most enterprises have not built.

Technology Product Maturity
Databricks Lakebase Announced June 2025. Earliest stage.
Snowflake Unistore + Snowflake Postgres GA but limited (1TB, AWS-only). Postgres newer.
Fabric SQL Database GA Nov 2025. SQL Server lineage gives OLTP depth.
Tilio take

None replace your production database yet. The consulting opportunity is evaluating which lightweight OLTP workloads can move to the analytics platform and which need dedicated databases.

Technology Product Approach
Databricks Genie Conversational BI integrated with AI/BI dashboards
Snowflake Intelligence Cortex Analyst, focused on structured data
Fabric IQ Data Agent Ontology-powered Q&A across enterprise data
Tilio take

Your text-to-SQL chatbot is only as smart as your data documentation. The convergence makes this a commodity feature. Quality is a function of metadata, not technology.

Category Databricks Snowflake Fabric
Ingestion Lakeflow Connect Openflow (NiFi-based) Data Factory
Transformation dbt + notebooks dbt native Dataflows + dbt
BI Genie + AI/BI dashboards Partner-dependent Power BI (20M+ models)
ML Platform MLflow + Mosaic AI Cortex + Notebooks Synapse ML + IQ
Catalog Unity Catalog Horizon + Polaris Purview + OneLake
Apps Databricks Apps Native Apps + Streamlit Power Apps + IQ agents
Tilio take

The "best-of-breed" era is ending. Platform selection is no longer about features. It is about organizational fit, team skills, and practice maturity. The tools converge. The practices that make them work don't.

Technology Product Approach
Databricks Agent Framework Agent evaluation, MLflow integration, compound AI
Snowflake Cortex Agents MCP server support, structured data focus
Fabric IQ Operations Agent Autonomous actions on enterprise data
Tilio take

Agents need even more context than chatbots. They act, not just answer. The governance and guardrails around autonomous data agents are the next frontier. Before your AI agent can act on your data, it needs to understand your business.

Divergence

Where technology choice actually matters

Seven areas where the platforms take fundamentally different positions. This is where selection decisions should focus.

Dimension Databricks Snowflake Fabric
Model Consumption (DBUs) Consumption (credits) Capacity (CUs)
Granularity Per-workload, per-second Per-workload, per-second Pooled capacity, hourly
Predictability Low Low Medium
Incentive Use more, pay more Use more, pay more Use more, same price
Tilio take

Consumption rewards efficiency but punishes experimentation. Capacity encourages experimentation but hides waste. Neither is inherently better. The right model depends on your organizational maturity and workload predictability.

Dimension Databricks Snowflake Fabric
Storage Delta Lake (OSS, vendor-controlled) Iceberg (Apache-governed) OneLake (proprietary)
Catalog Unity Catalog (proprietary) Polaris (fully open source) Purview (proprietary)
Compute Spark (OSS) + Photon (proprietary) Proprietary Proprietary
Tilio take

Databricks has the broadest ecosystem. Snowflake has the most portable catalog. Fabric is the most closed but offers the deepest integration. Help your team evaluate what actually needs to be portable versus what is marketing.

Dimension Databricks Snowflake Fabric
GTM motion Engineering-led, bottom-up Sales-led, top-down Installed-base, top-down
Primary buyer Data engineering team CDO / data leadership CIO / Microsoft relationship
Entry point Spark workloads SQL analytics Power BI
Tilio take

Engineering-first organizations gravitate to Databricks. Microsoft-native to Fabric. Analytics-first to Snowflake. The starting point differs. The practices that make any of them succeed are the same.

Dimension Databricks Snowflake Fabric
Owned BI Genie, AI/BI dashboards None (partner-dependent) Power BI (20M+ semantic models)
Philosophy Replace dashboards with AI chat Best-in-class analytics engine BI is the platform on-ramp
Tilio take

The BI layer reveals platform philosophy. Databricks believes dashboards are dying. Snowflake believes BI should be best-of-breed. Fabric believes BI is the entry point. For Norwegian enterprises, Power BI is dominant. That is Fabric's structural advantage in this market.

Dimension Databricks Snowflake Fabric
Cross-org sharing Delta Sharing (open protocol) Marketplace + Clean Rooms OneLake shortcuts + mirroring
Clean Rooms None (partner-dependent) Native + Samooha (market leader) None
Tilio take

Clean rooms are genuinely differentiated. If cross-organizational data collaboration matters, Snowflake is the only real option today. But clean rooms require bilateral adoption, and in smaller markets that limits their value.

Dimension Databricks Snowflake Fabric
Cloud support AWS, Azure, GCP AWS, Azure, GCP Azure only
True multi-cloud Yes (same product, all clouds) Yes (same product, all clouds) No (Azure-native)
Tilio take

Most Norwegian enterprises are Azure-primary. Multi-cloud matters less here than globally. But for organizations with international operations or regulatory requirements, Fabric's Azure-only position is a real constraint.

Technology Primary Lock-in How It Works
Databricks Unity Catalog Governance scope expands every release. Deeper UC adoption means higher switching costs.
Snowflake Compute specialization Credits burned on Snowflake-specific features create workload dependency.
Fabric OneLake data gravity Mirroring pulls data in. Direct Lake ties BI to storage. Each step deepens dependency.
Tilio take

All three platforms create lock-in. The question is not whether you will get locked in. It is choosing the lock-in you can live with and making that choice consciously, before the architecture decides for you.

Persistent Gaps

Six problems no platform will fix

These gaps are structural. They persist because solving them conflicts with how platforms make money. Structural gaps get fixed by practices, not platforms.

The Foundation Gap
No platform can tell you "your data engineering needs to be better before our features help you." Every product launch assumes clean, well-modeled data underneath.
Nearly nine in ten enterprises have adopted AI. Fewer than one in ten capture value at scale.
Every platform sells on the assumption your data is ready. It usually is not.
The Context Gap
AI features across all three platforms need business context: semantic definitions, documentation, lineage, business rules. Making AI accessible does not make it useful.
Nearly every organization surveyed reports context gaps between their data and their AI tools.
AI without business context is expensive autocomplete.
The Cost Visibility Gap
All three pricing models obscure per-workload cost attribution to varying degrees. No platform offers transparent, actionable cost optimization guidance because it conflicts with their revenue model.
Infrastructure costs keep dropping. Enterprise AI bills keep rising. The gap is not the technology. It is how it is used.
Cost transparency is structurally anti-revenue. No platform will build tooling that reduces their own.
The Governance Complexity Gap
All three platforms are expanding governance scope. None can say "you probably need less governance, not more." Governance expansion is their lock-in mechanism.
Governance is the most common blocker to AI adoption. Most teams cannot measure whether their governance investment is paying off.
More governance tools, same missing layer underneath: well-engineered data.
The Migration Reality Gap
All three platforms publish case studies showing smooth adoption. None of them reflect the organizational and process change that migration actually requires.
Every platform launch promises seamless migration. The first 90 days tell a different story.
Honest migration estimates would lengthen sales cycles. That is why platforms do not give them.
The Organizational Readiness Gap
All three platforms sell capabilities. None assess whether the buying organization can actually use them. The gap between capability and maturity keeps widening.
Leadership adopts AI tools. Frontline teams do not follow. The gap between executive enthusiasm and operational reality keeps widening.
Platforms grow at Moore's Law speed. Organizations do not.
What Now?

Every platform starts with
a conversation.

Whether you are evaluating platforms, renegotiating contracts, or trying to get value from the technology you already have, we would like to hear about it.

Start a Conversation