AI assistant — not human

T-SQL native + OneLake + serverless compute + medallion architecture + Copilot for Fabric. Migration from Synapse + Snowflake.
Last updated July 7, 2026 by Errin O'Connor, Founder & Chief AI Architect, EPC Group
Fabric Data Warehouse: T-SQL native, OneLake integration, serverless compute (auto-scaling), cross-warehouse queries, row-level + column-level security, dynamic data masking, Copilot for Fabric SQL generation. Replaces Synapse dedicated pools. Medallion architecture: Bronze (raw) → Silver (cleaned) → Gold (business-aggregated). Synapse to Fabric migration: 20-48 weeks. Cost via Fabric F-SKU capacity: F64 supports 500GB+moderate; F128 supports 5TB+heavy; F256+ multi-TB. Copilot for Fabric productivity gain: 20-40% for query authoring. EPC Group Warehouse Build $100K-$300K; Synapse to Fabric Migration $200K-$600K; Snowflake to Fabric $250K-$700K.
Fabric Data Warehouse is Microsoft's cloud-native T-SQL warehouse within Microsoft Fabric, providing: (1) T-SQL native — write T-SQL directly, no Spark required. (2) OneLake integration — data stored in Delta Parquet on OneLake, accessible to all Fabric workloads. (3) Serverless compute — auto-scales with query load, no cluster management. (4) Cross-warehouse queries — join across warehouses + lakehouses + KQL databases. (5) Row-level security + column-level security + dynamic data masking. (6) Copilot for Fabric — natural-language SQL generation + optimization. Replaces separate Synapse dedicated pools + serverless SQL for Fabric customers.
Six differences: (1) Compute model — Fabric serverless auto-scaling vs Synapse dedicated DWU pools (must pre-provision). (2) Storage — OneLake Delta Parquet vs proprietary Synapse storage. (3) Cost model — Fabric F-SKU capacity vs Synapse per-DWU-hour. (4) Feature parity — Synapse still has some warehouse features Fabric doesn't (specific PolyBase scenarios, some CETAS options). (5) Investment — Fabric gets new features, Synapse stagnant. (6) Ecosystem — Fabric integrated with Power BI + Copilot + Purview natively. Migration path: 12-24 months. For most Synapse customers, Fabric Warehouse is the strategic destination.
Medallion architecture is a data engineering pattern for organizing data quality tiers in Fabric: (1) Bronze — raw ingested data, minimal transformation, source of truth for reprocessing. (2) Silver — cleansed + conformed + deduplicated, aligned business entities. (3) Gold — business-aggregated + aggregated + optimized for consumption + analytics. Fabric implementation: (1) Bronze in Lakehouse (Parquet + Delta). (2) Silver in Lakehouse (cleaned Delta tables). (3) Gold in Warehouse (T-SQL modeled tables). Or all three in Warehouse for simpler stacks. Data flows Bronze → Silver → Gold via Data Factory pipelines + Spark notebooks + T-SQL procedures.
Migration approach: (1) Assessment (2-4 weeks) — inventory Synapse workloads, schema, ETL, Power BI dependencies. (2) Design (2-4 weeks) — target Fabric F-SKU sizing, medallion architecture, workload placement. (3) Foundation deployment (2-4 weeks) — Fabric tenant + OneLake + workspace hierarchy + governance. (4) Schema migration (4-8 weeks) — T-SQL schema recreation + adjustments for Fabric warehouse limitations. (5) ETL rebuild (4-16 weeks) — Data Factory pipelines + Notebooks replacing Synapse pipelines. (6) Power BI validation (2-4 weeks) — reports point to new warehouse + validate accuracy. (7) Cutover + parallel run (4-8 weeks). Total: 20-48 weeks depending on complexity.
Copilot for Fabric warehouse capabilities: (1) Natural-language T-SQL generation — "show me top 10 customers by revenue last quarter" → SQL query. (2) Query optimization suggestions — Copilot identifies performance issues + rewrites. (3) Documentation generation — auto-generate table + column descriptions. (4) Schema exploration — natural-language schema discovery. (5) DAX generation for Power BI — from warehouse metrics. (6) Migration assistance — help translating Synapse SQL to Fabric SQL syntax variations. Included in Fabric F64+ SKUs. Productivity gain: 20-40% for data analysts + engineers writing queries.
Fabric Warehouse cost components (as of 2026): (1) F-SKU capacity — F2 $260/mo through F256+ $33.6K/mo. Compute + storage costs count against capacity units. (2) OneLake storage — $0.023/GB/month. (3) No separate query fees — CU consumption tracked against capacity. Autoscale + burst supported. Compared to Synapse: for steady workloads Fabric is 20-40% cheaper; for bursty workloads Synapse serverless can be cheaper. Rough sizing: F64 supports 500 GB warehouse + moderate query volume; F128 supports 5 TB + heavy queries; F256+ for multi-TB warehouses. EPC Group Fabric Migration Assessment includes F-SKU sizing.
EPC Group Fabric Warehouse portfolio: (1) Warehouse Discovery + Design ($35K, 4 weeks) — current-state + target-state + F-SKU sizing + fixed-fee proposal. (2) Foundation Deployment ($75K, 6 weeks) — F-SKU + OneLake + workspace hierarchy + governance. (3) Warehouse Build ($100K-$300K, 12-20 weeks) — schema + ETL + Power BI models + operations. (4) Synapse to Fabric Migration ($200K-$600K, 20-48 weeks) — full migration from Synapse dedicated pools + serverless. (5) Snowflake to Fabric Migration ($250K-$700K, 24-48 weeks) — Snowflake schema + workload migration. (6) Ongoing Warehouse Governance Retainer ($10K-$25K/month). All led by senior data architect.
$35K/4wk current-state + target-state + F-SKU sizing. Call (888) 381-9725.
Monday-Friday, 8 AM - 7 PM CT
We respond to all inquiries within one business day