Microsoft Fabric F-SKU Pricing Explained (F2-F2048)
By Jonathan Flach · Published 2026-06-20 · Reviewed 2026-06-20
The Microsoft Fabric F-SKU ladder runs from F2 at $262.80/month to F2048 at $269,107.20/month — but the price table is only half the story. Several tiers cross platform thresholds that change what you can do: free Power BI viewers appear at F64, Fabric Data Agents require F64 or higher, and free mirroring storage scales at exactly 1 TB per CU. The canonical price table lives in the Microsoft Fabric pricing and capacity planning guide; this article delivers something that guide deliberately left out: the F-SKU capability ladder — every rung annotated with what unlocks, what the reserved discount looks like at that tier, and which tiers most buyers skip and why.
All prices are USD, as of June 2026, computed from the published rate of $0.18 per CU-hour (Azure pricing, Microsoft, checked June 2026) across 730 hours/month. Reserved figures use the 0.5949 factor (approximately 59% of PAYG, a 40.51% discount) from the subscription pricing documentation (Buy a Microsoft Fabric subscription, Microsoft Learn, checked June 2026).
The F-SKU tier capability ladder (F2-F2048)
This is the table that exists nowhere else: prices at every tier plus the platform capabilities that unlock — or are locked — at each rung. Use it to find the lowest tier that clears every threshold you need.
| F-SKU | CUs | PAYG/mo | Reserved 1-yr/mo | Free PBI viewers? | Copilot? | Data Agents? | Free mirroring storage |
|---|---|---|---|---|---|---|---|
| F2 | 2 | $262.80 | $156.34 | No — Pro/PPU required | Yes (F2+) | No (F64 min) | 2 TB |
| F4 | 4 | $525.60 | $312.68 | No — Pro/PPU required | Yes | No | 4 TB |
| F8 | 8 | $1,051.20 | $625.36 | No — Pro/PPU required | Yes | No | 8 TB |
| F16 | 16 | $2,102.40 | $1,250.72 | No — Pro/PPU required | Yes | No | 16 TB |
| F32 | 32 | $4,204.80 | $2,501.44 | No — Pro/PPU required | Yes | No | 32 TB |
| F64 | 64 | $8,409.60 | $5,002.87 | Yes — free Fabric license | Yes | Yes (F64+) | 64 TB |
| F128 | 128 | $16,819.20 | $10,005.74 | Yes | Yes | Yes | 128 TB |
| F256 | 256 | $33,638.40 | $20,011.48 | Yes | Yes | Yes | 256 TB |
| F512 | 512 | $67,276.80 | $40,022.97 | Yes | Yes | Yes | 512 TB |
| F1024 | 1,024 | $134,553.60 | $80,045.94 | Yes | Yes | Yes | 1,024 TB |
| F2048 | 2,048 | $269,107.20 | $160,091.87 | Yes | Yes | Yes | 2,048 TB |
Reserved figures are estimates rounded from the 0.5949 factor; verify the Azure portal for your region and subscription before committing (reservations are subscription- and region-bound). Free mirroring storage scales at exactly 1 TB per CU purchased (Mirroring overview, Microsoft Learn, checked June 2026): an F64 includes 64 free TB of mirrored database storage. Mirrored storage becomes billable standard OneLake storage if the capacity is paused — another reason the pause decision deserves care.
The Copilot row needs a footnote: Copilot on all paid SKUs from F2 upward was announced in March 2025 (What's new archive, Microsoft Learn, checked June 2026). Trial SKUs are excluded — Copilot is not supported on trial capacity (Copilot in Fabric glossary, Microsoft Learn, checked June 2026). The Data Agents column reflects a harder platform constraint: Fabric Data Agents are not supported on workspaces with a capacity SKU smaller than F64 (Fabric Copilot capacity, Microsoft Learn, checked June 2026).
What the trial gives you — and what it withholds
Before spending anything, most teams start with the 60-day Fabric trial. The trial is configured as either an F4 capacity (4 CUs) or an F64 capacity (64 CUs) depending on your tenant — check Admin portal > Capacity settings > Trial to confirm yours (Fabric trial capacity, Microsoft Learn, checked June 2026; Understand Microsoft Fabric Licenses, Microsoft Learn, checked June 2026). Both configurations include 1 TB of storage and access to all Fabric experiences. If you receive an F64 trial, that clears the free Power BI viewer threshold; an F4 trial does not — every viewer still needs a Pro or PPU license. In either case, Copilot, AI features, and Data Agents are not available on trial SKUs (Copilot in Fabric glossary, Microsoft Learn, checked June 2026). If AI features are part of your evaluation, you need a paid capacity. A short F64 PAYG month costs $8,409.60 — treat it as the price of a real AI pilot versus a constrained trial.
The three thresholds that actually matter
The doubling ladder has 11 rungs, but three platform thresholds create real decision points. Everything else is a smooth scaling question.
Threshold 1 — F64: free Power BI viewers. At F64 and above, any user with a free Fabric license can view Power BI reports hosted on that capacity with a viewer role on the workspace. Below F64, every viewer needs a Pro ($14/user/month) or PPU ($24/user/month) license (Understand Microsoft Fabric Licenses, Microsoft Learn, checked June 2026). This creates the licensing cliff covered in depth in the F64 licensing threshold analysis: at some viewer count, buying a bigger capacity beats buying individual licenses.
The arithmetic at June-2026 prices: against an F32 baseline ($4,204.80/month), you're also paying for viewers at $14/Pro. F64 ($8,409.60/month) breaks even with F32 + Pro at roughly ⌈(8,409.60 - 4,204.80) / 14⌉ = 301 Pro viewers. More than ~301 viewers on an F32 and the F64 is actually cheaper — plus you get twice the compute.
Threshold 2 — F64: Data Agents. Fabric Data Agents, the conversational AI layer over your data, require a workspace on F64 or larger. This makes F64 the practical floor for enterprise AI-on-data scenarios, separate from the viewer math. If Data Agents are on your roadmap, the F64 is not optional.
Threshold 3 — F2: Copilot baseline. Copilot availability starts at F2 (any paid SKU). The decision here is not "which tier unlocks Copilot" but "do we need a paid capacity at all?" — the trial blocks Copilot, a $262.80/month F2 enables it.
PAYG vs reserved: what the discount ladder actually looks like
The 40.5% reserved discount applies uniformly across all F-SKUs — there is no steeper deal for bigger tiers, and the ladder is a flat multiplier. That arithmetic means the absolute dollar saving scales with the SKU:
| F-SKU | PAYG/mo | Reserved/mo | Monthly saving | Annual saving |
|---|---|---|---|---|
| F2 | $262.80 | $156.34 | $106.46 | $1,277.52 |
| F8 | $1,051.20 | $625.36 | $425.84 | $5,110.08 |
| F32 | $4,204.80 | $2,501.44 | $1,703.36 | $20,440.32 |
| F64 | $8,409.60 | $5,002.87 | $3,406.73 | $40,880.76 |
| F128 | $16,819.20 | $10,005.74 | $6,813.46 | $81,761.52 |
| F256 | $33,638.40 | $20,011.48 | $13,626.92 | $163,523.04 |
Estimated figures from the 0.5949 factor, as of June 2026. The saving at F64 is $40,880.76 per year — but only if the capacity runs. A reservation is a billing commitment that runs whether or not the capacity is active. The decision rule is utilization: if your capacity runs above roughly 60% of available hours, a one-year reservation wins. Below that, PAYG with scheduled pausing captures more of the idle hours. The full break-even model is in reserved vs PAYG for Fabric.
One trap to name here: the three-year reservation. In most Azure services, three years buys a meaningfully steeper discount than one year. In Fabric, the difference is marginal — the extra lock-in rarely earns its keep. Default to the one-year for steady baselines; use PAYG for spiky or windowed loads.
Which tiers most buyers skip — and why
The doubling structure means every step doubles cost. In practice:
- F4 and F16 are the most commonly skipped tiers. F4 is close enough to F2 in absolute cost that teams size F8 instead for headroom. F16 sits between F8 and the F32 that many BI-only teams choose as a production floor. Neither F4 nor F16 crosses a platform threshold.
- F32 is the largest pre-F64 tier. It is the common choice for teams with heavy workloads that do not yet have enough Power BI viewers to justify the F64 jump. If you are on F32 and your viewer count is climbing toward ~301, model the all-in cost with licenses before the next planning cycle.
- F128, F256, F512 are enterprise scale. Above F64, the capability ladder is flat — no new unlocks until enterprise-specific feature agreements. These tiers are pure throughput decisions.
- F1024 and F2048 are hyperscale. Monthly PAYG costs of $134,553 and $269,107 respectively. At this scale, you almost certainly have an enterprise agreement that changes the list prices — the figures in this article are published list rates, not EA prices.
The throttling blast-radius: the enemy this tier decision shapes
The most expensive mistake on the F-SKU ladder is not picking the wrong tier — it is picking the right tier and then letting one workload eat it for everyone else. Fabric CUs are pooled across all workloads on a capacity. Without configuration, a single workspace can consume all capacity resources. Workspace-level surge protection (Preview since January 2026) lets admins set per-workspace CU percentage caps, but this requires explicit configuration and does not provide hard isolation — the capacity-level throttling stages still apply if the total is exceeded (Surge protection, Microsoft Learn, checked June 2026). A runaway query, an unoptimized Dataflow Gen2, or a Power Query mistake in the report editor can consume the capacity's smoothed budget and push the whole tenant into throttling.
Throttling stages are measured as future-capacity time windows, not utilization percentages:
- Up to 10 minutes of future usage consumed: overage protection absorbs it, no user impact
- 10-60 minutes of future usage consumed: ~20-second delay added to interactive requests
- 60 minutes to 24 hours consumed: interactive requests rejected
- More than 24 hours consumed: all requests including background jobs rejected
Source: Understand throttling in Microsoft Fabric, Microsoft Learn (checked June 2026).
A single uncontrolled workload at F64 can drag 250 report viewers from "working" to "rejected" without per-workspace surge protection in place. The architectural answer is either a multi-capacity setup (isolated blast radius per department, at higher total cost), enabling workspace-level surge protection, or tight monitoring of the Capacity Metrics app — which holds 14 days of compute detail on its Compute page — and a preview Item History page (available since August 2025) that extends that window to 30 days (Understand the metrics app item history page, Microsoft Learn, checked June 2026) — to catch repeating offenders before they own the next production incident. This is the throttling blast-radius enemy: one greedy workspace, no hard isolation by default.
What to do next
- Locate your threshold. Find the three thresholds that apply to your situation: viewer count (F64 cliff), AI/Data Agent requirements (F64 floor), Copilot on trial vs paid (F2+). These set your minimum viable tier, independent of load.
- Size to smoothed load, not peak. Run the 60-day trial at F64 or a short PAYG burst, read the 14-day Compute page (or the 30-day preview Item History page) from the Capacity Metrics app, and find your true 24-hour smoothed background peak. Pick the smallest SKU where that peak stays under 100%.
- Run the viewer math before locking in an F32. If you are considering F32, count your Power BI viewers and compute F32 + Pro all-in versus F64 alone. The crossover at ~301 viewers is a mechanical calculation, not a judgment call.
- Choose PAYG or reserved from utilization, not from the discount headline. The 40.5% discount is real but only beneficial above ~60% utilization. If you have spiky or windowed workloads, PAYG with scheduled pausing is the honest choice.
- Never pause to clear throttling. Pausing a capacity settles all smoothed and cumulative overages to your Azure bill immediately at full PAYG rates — including on reserved capacities, where you then pay the reservation on top of the PAYG overage. Size up or optimize the workload.
For a plain-math breakdown of the full billing model behind these tiers, subscribe to the SpendWeave Fabric cost teardowns — real capacity bills, dissected, on a regular cadence.
Frequently asked questions
What is an F-SKU in Microsoft Fabric? An F-SKU (Fabric SKU) is a capacity tier that determines how many Capacity Units (CUs) your Fabric environment gets per second. Each tier doubles the CUs of the one below: F2 gives 2 CUs, F4 gives 4, and so on up to F2048 with 2,048 CUs. More CUs mean more throughput for all workloads — pipelines, Spark, SQL, Power BI renders — that share the capacity.
How much does each Microsoft Fabric F-SKU cost per month? On pay-as-you-go at $0.18/CU-hour and 730 hours/month (as of June 2026): F2 $262.80, F4 $525.60, F8 $1,051.20, F16 $2,102.40, F32 $4,204.80, F64 $8,409.60, F128 $16,819.20, F256 $33,638.40, F512 $67,276.80, F1024 $134,553.60, F2048 $269,107.20. A one-year reservation costs approximately 59% of the PAYG rate (about a 40.5% discount).
Which F-SKU do I need for Copilot in Microsoft Fabric? Copilot and AI capabilities are available on all paid Fabric SKUs from F2 upward (as of March 2025), but Copilot is not supported on trial SKUs. However, Fabric Data Agents require at least F64 — they are not supported on capacities smaller than F64.
At which F-SKU do Power BI viewers not need Pro licenses? At F64 and above, users with a free Fabric license can view Power BI reports on that capacity without a Pro or PPU license. Below F64 (F2 through F32), every viewer must hold a Power BI Pro ($14/user/month) or Premium Per User ($24/user/month) license.
What is the difference between PAYG and reserved pricing for Fabric F-SKUs? Pay-as-you-go bills per second the capacity is active and lets you pause it when idle. A one-year reservation costs about 59% of the PAYG rate — roughly a 40.5% discount — but it is a billing commitment you pay regardless of whether the capacity runs. A three-year reservation barely beats the one-year rate and is rarely worth the added lock-in.
Researched with AI assistance, written and fact-checked by Jonathan Flach, verified against Microsoft Learn.