AI assistant — not human

F2-F2048 tier selection · Workload sizing methodology · Burst-and-smooth capacity mechanics · Copilot for Fabric cost implications · Azure reservation strategy for 25-45% year-over-year savings.
Last updated July 10, 2026 by Errin O'Connor, Founder & Chief AI Architect, EPC Group
Microsoft Fabric F-SKU is the unified capacity model spanning F2-F2048 tiers that replaces Power BI Premium P-SKU + Azure Synapse dedicated pools. Six architectural differences: hourly billing + pause/resume + all-workload access + burst-and-smooth + wider tier range + Azure reservation eligible. F64 is the pivotal tier — unlocks Fabric Free per-user consumption saving Power BI Pro/PPU licensing for enterprises with 200-500+ users. Copilot for Fabric consumes 15-40% baseline CU rising to 60-80% at maturity — plan +30-60% headroom. Azure reservations deliver 30-41% baseline savings. Six-phase EPC Group engagement 8-16 weeks $85K-$585K. Anchored by Microsoft Solutions Partner Data & AI + Modern Work + Security + Infrastructure designations.
Microsoft Fabric F-SKU is the unified capacity model that replaces separate Power BI Premium P-SKU + Azure Synapse dedicated pools + standalone Fabric compute. Six differences from Power BI Premium P-SKU: (1) F-SKU is billed per hour with per-second granularity + pause/resume support vs P-SKU monthly commitment. (2) F-SKU capacity units (CU) power ALL Fabric workloads — Data Factory + Data Warehouse + Real-Time Intelligence + Data Science + Data Engineering + Power BI + Copilot for Fabric — vs P-SKU only powering Power BI + Dataflows. (3) F-SKU implements burst-and-smooth architecture: workloads can burst above their allocated capacity for short periods with automatic smoothing over 24 hours vs P-SKU hard-capped v-cores. (4) F-SKU tiers span F2 through F2048 (11 tiers) vs P-SKU only P1-P5. (5) F-SKU capacity is Azure resource + Azure reservation eligible for 1-year or 3-year commit discounts (up to 41% savings). (6) F-SKU includes OneLake storage + workload autoscaling vs P-SKU billing per-user + per-Premium-workspace. Enterprise pattern: retire Power BI Premium P-SKU + Azure Synapse dedicated pools + provision Fabric F-SKU capacity per workload domain.
Seven-step F-SKU tier sizing decision framework: (1) F2 or F4 — pilots, proof-of-concept, dev/test environments only. Not production. (2) F8 or F16 — small enterprise Power BI-only workloads under 500 users, no significant ETL or Real-Time workloads. (3) F32 — mid-market Power BI Premium replacement + light data engineering (Dataflows Gen2, Notebooks, Warehouse queries). Suitable for 500-2,500 users. (4) F64 — the minimum tier that includes Power BI Premium Per User (PPU) license inclusion — every user reading reports needs a PPU or Fabric Free license below F64. Above F64, "Fabric Free" users can access reports at no per-user licensing cost. F64 is the practical minimum for enterprise Power BI deployments avoiding per-user PPU costs. (5) F128 through F256 — mid-to-large enterprise with meaningful data engineering + warehouse + Real-Time Intelligence workloads. (6) F512 through F1024 — large enterprise with multiple data domains, high concurrency, Copilot for Fabric at scale. (7) F2048 — largest enterprise scale, multi-tenant delivery, dedicated capacity per business unit. Starting tier depends on: (a) user count for Power BI consumption + PPU savings, (b) data engineering workload intensity, (c) Copilot for Fabric planned rollout, (d) Azure reservation appetite.
F64 is the pivotal Fabric tier for enterprise Power BI economics because it unlocks "Fabric Free" per-user consumption for report readers. Below F64: every user reading published Power BI reports needs Power BI Pro or Power BI Premium Per User (PPU) license — ~$10-24 per user per month for Pro, ~$20/user/month for PPU. At F64+: report readers can consume with Fabric Free license (no per-user cost). Break-even math: F64 at Azure retail (approximately hourly rate) + Azure reservation discount for organizations with ~200-500+ Power BI Pro users typically justifies the F64 tier vs continuing Pro licensing. Additional F64+ benefits: full Fabric capacity workload access (Data Warehouse, Real-Time Intelligence, Data Engineering, Copilot for Fabric) + burst-and-smooth capacity + OneLake storage inclusion. Sizing exercise for 1,000-user enterprise: F64 with 1-year Azure reservation vs 1,000 Power BI Pro licenses annualized cost comparison typically favors F64. Above 5,000 users: F128 or higher becomes economically necessary for concurrency + workload isolation. Below 200 users: F32 or Power BI Premium remains cheaper.
Burst-and-smooth is Fabric F-SKU's architectural improvement over Power BI Premium P-SKU's hard-capped v-cores. Six mechanics: (1) Baseline capacity — each F-SKU tier has an assigned baseline capacity unit (CU) allocation. (2) Interactive workloads (Power BI queries, Q&A) can burst above baseline for short periods when needed. (3) Background workloads (dataset refresh, scheduled Data Factory pipelines, Warehouse batch jobs) queue and execute within baseline capacity. (4) Smoothing period — Fabric smooths capacity consumption over 24-hour rolling window; short bursts absorbed without throttling. (5) Throttling threshold — sustained overage beyond capacity for 10+ minutes triggers throttling; sustained overage 60+ minutes triggers rejection. (6) Capacity Metrics App — Microsoft-provided app in Fabric portal shows CU% utilization by workload + operation + user for capacity planning. Planning implications: F-SKU tier selection should consider peak concurrent workloads + burst tolerance + smoothing headroom + Copilot for Fabric metered consumption + throttling risk tolerance for interactive users. Enterprise pattern: over-provision +30-50% headroom on top of workload baseline sizing; enable Azure reservation for baseline; use PAYG bursting for peaks. EPC Group Fabric F-SKU right-sizing methodology delivers 25-40% cost reduction vs naive tier selection.
Copilot for Fabric consumes Fabric capacity units (CU) per invocation, adding to baseline workload consumption. Seven consumption dimensions: (1) Data Warehouse Q&A queries — moderate CU per natural language query (LLM + query generation + query execution). (2) Notebook Copilot — high CU per code generation + execution operation. (3) Data Factory Copilot — moderate CU per pipeline authoring + execution. (4) Power BI Copilot (Q&A + narrative + report generation) — moderate CU per interaction. (5) Real-Time Intelligence Copilot — moderate CU per KQL generation. (6) Data Science Copilot — high CU per model training + inference. (7) Fabric Copilot Studio agents — variable CU depending on agent complexity + data volume + LLM tokens. Enterprise Copilot for Fabric budget guidance: Copilot workloads typically consume 15-40% of Fabric capacity in first 6 months of adoption + can scale to 60-80% at mature deployment. Planning approach: (a) baseline non-Copilot workloads determine minimum F-SKU tier, (b) add 30-60% CU headroom for Copilot for Fabric adoption, (c) monitor via Fabric Capacity Metrics App + adjust tier during quarterly capacity reviews, (d) consider separate F-SKU capacity for Copilot workloads vs analytics workloads to isolate cost tracking + prevent noisy-neighbor throttling. EPC Group Copilot for Fabric governance methodology includes capacity budget + cost telemetry + user quota controls.
Six factors for Azure Reservation decision on Fabric F-SKU capacity: (1) Baseline predictability — reservations require 1-year or 3-year commitment; only apply reservations to baseline capacity you're certain to consume. (2) Discount magnitude — 1-year reservation ~30% savings, 3-year reservation ~41% savings on baseline capacity vs pay-as-you-go. (3) Pause/resume flexibility trade-off — reservations are billed regardless of pause/resume state; PAYG capacity can be paused for cost savings. Enterprise pattern: reserve baseline continuously-running capacity, use PAYG for burst + dev/test. (4) Right-sizing before reserving — commit to reservation only after 60-90 days of production usage validates the tier. (5) Multi-year discount stacking — 3-year reservation combined with Microsoft Enterprise Agreement + Azure Consumption Commitment can achieve 45-50% total savings. (6) Reservation exchange — Azure allows exchange within the same reservation type; enterprises can shift from F64 → F128 reservation as workloads grow. Enterprise pattern: reserve ~70% of baseline capacity, PAYG for ~30% burst + Copilot growth headroom, quarterly right-sizing review. EPC Group Fabric FinOps methodology delivers 25-45% year-over-year Fabric capacity cost optimization.
Fixed-fee scope covering four workstreams: (1) Discovery + Assessment (2-3 weeks) — Power BI Premium P-SKU current state, Azure Synapse dedicated pool current state, dataset + report inventory, refresh window analysis, concurrency analysis, business user count for PPU/Pro/Free license optimization. (2) F-SKU Sizing + Architecture (2 weeks) — tier selection using 7-step framework, workload isolation strategy, capacity separation for dev/test/production, Copilot for Fabric capacity budget, Azure reservation strategy. (3) Migration + Cutover (3-6 weeks) — Power BI Premium → F-SKU migration, Synapse dedicated pool → Fabric Warehouse migration, dataset semantic model migration to Direct Lake mode where appropriate, connectivity + gateway reconfiguration. (4) Sustainment (ongoing) — quarterly capacity right-sizing review, Copilot for Fabric governance, cost telemetry dashboard, reservation renewal advisory. Fixed-fee ranges: $85K-$185K for mid-market Fabric transition + $285K-$585K for large enterprise multi-workload consolidation. Anchored by Microsoft Solutions Partner (Data & AI + Modern Work + Security + Infrastructure) designations. Named senior architect with PL-300 + DP-500 + DP-600 credentials. Delivered under fixed-fee scope + capacity cost SLA commitment.
30-min qualification with senior Fabric architect. F-SKU sizing + Azure reservation strategy within 5 business days. Call (888) 381-9725.
Monday-Friday, 8 AM - 7 PM CT
We respond to all inquiries within one business day