Power BI Embedded vs Fabric F SKUs: A1 at $1.0081 an hour vs F8 at $1.44

For a software company that shows Power BI reports to its own customers inside a web app, the cheapest capacity on Microsoft's retail price list is still the Power BI Embedded A SKU. In the East US region, an A1 node lists at $1.0081 per hour, while the Fabric F8 capacity that Microsoft maps to the same compute lists at $1.44 per hour on pay-as-you-go, about 43% more. The picture flips once the capacity runs around the clock: a one-year Fabric reservation brings F8 down to $625.33 a month, below the $735.91 an always-on A1 costs over a 730-hour month.
Both SKUs are bought in Azure, both support the "embed for your customers" scenario in which end users need no Power BI license, and both can be scaled. They differ in what else they run, in which embedding options they allow, in whether Copilot can be used, and in how billing works when the capacity is paused. The prices below come from Microsoft's public Azure retail price list for East US in USD, checked on November 19, 2026; the Power BI Embedded and Fabric pricing pages themselves only show prices after a region and currency are selected.
··········
POWER BI EMBEDDED A SKU AND FABRIC F SKU HOURLY PRICES.
Microsoft's embedded capacity documentation (updated May 12, 2026) maps each A SKU to an F SKU with the same number of Power BI v-cores: A1 to F8, A2 to F16, A3 to F32, A4 to F64, A5 to F128 and A6 to F256. The Fabric licensing page adds that the table compares compute capacity and is not a statement of functional or licensing equivalence. Fabric capacity is priced per capacity unit (CU), and the East US retail meter for Power BI capacity usage is $0.18 per CU hour, so an F8 costs 8 × $0.18 = $1.44 an hour.
The one-year Fabric reservation lists at $938 per CU per year. Microsoft's Power BI pricing page describes the reservation as an annual purchase that saves 40.5% over pay-as-you-go, which matches the arithmetic: $938 divided by 8,760 hours at $0.18 is a 40.5% discount. Monthly figures below use a 730-hour month (8,760 hours divided by 12).
........
A SKU | A price per hour | Fabric SKU | F pay-as-you-go per hour | F reserved per month | Max model memory |
A1 | $1.0081 | F8 | $1.44 | $625.33 | 3 GB |
A2 | $2.0081 | F16 | $2.88 | $1,250.67 | 5 GB |
A3 | $4.0242 | F32 | $5.76 | $2,501.33 | 10 GB |
A4 | $8.0565 | F64 | $11.52 | $5,002.67 | 25 GB |
A5 | $16.121 | F128 | $23.04 | $10,005.33 | 50 GB |
A6 | $32.2506 | F256 | $46.08 | $20,010.67 | 100 GB |
........
The memory column follows the Power BI Embedded pricing page for the A nodes (3 GB RAM for A1 up to 100 GB for A6) and the Fabric semantic model limits table for the matching F SKUs, which list the same values. Microsoft applies that limit to a single item such as one semantic model, so an F64 or A4 caps one dataset at 25 GB rather than capping the total. The A7 and A8 nodes appear on the pricing page but had no East US retail price on the day of verification, and F2 and F4 have no A equivalent at all. Readers sizing a model for these tiers can compare with the full Fabric limits in the Power BI semantic model size limits by license.
··········
WHAT A POWER BI EMBEDDED A SKU INCLUDES COMPARED WITH A FABRIC F SKU.
The Power BI Premium to Fabric migration FAQ (updated September 11, 2026) states that Power BI Embedded is included with every F SKU and that customers do not need to keep separate EM or A SKUs for embedded scenarios after moving to Fabric. The FAQ announces no retirement of A SKUs; the retirement it describes applies only to the per-capacity P1 to P5 SKUs.
The practical differences come from Microsoft's embedded capacity scenario table and the Copilot requirements page.
........
Question | A SKU (Power BI Embedded) | F SKU (Microsoft Fabric) |
Purchased in | Azure | Azure |
Embed for your customers (app owns data) | Yes | Yes |
Embed for your organization (user owns data) | Yes | Yes |
Teams, SharePoint, PowerPoint embedding | Yes | Yes |
Secure URL embedding from the Power BI service | No | Yes |
Free users view content in the Power BI service | No | F64 and larger only |
Copilot in Power BI | Not listed | F2 or larger, paid |
Other Fabric workloads | No | Yes |
Billing | Hourly, usage computed to the second | Per second, one-minute minimum |
........
The Copilot introduction for Power BI, updated November 17, 2026, lists a paid Fabric capacity of F2 or higher, or Power BI Premium P1 or higher, as the requirement, and the page does not mention A SKUs. A product team that plans to offer Copilot inside an embedded report therefore has to budget for an F SKU. The full capacity, license and region rules are in the Copilot in Power BI requirements reference.
The "other Fabric workloads" row covers lakehouses, warehouses, pipelines, notebooks and the rest of the Fabric items that run on the same capacity units. Microsoft sells the A SKU as a Power BI Embedded resource for Power BI items, so a company that also wants to land and transform its data in Fabric would need a second capacity next to the A node.
··········
LICENSES NEEDED FOR CUSTOMER-FACING POWER BI EMBEDDED REPORTS.
In the "embed for your customers" scenario, also called app owns data, Microsoft's documentation says there are no licensing requirements for the end users who view the reports in the vendor's application. This holds for A SKUs and for every F SKU size, including F8.
The people who build and publish the reports are a different case. Microsoft requires a Power BI Pro or Premium Per User account to publish content; the documented alternative is publishing through a service principal that calls the Post Import In Group REST API. Power BI Pro lists at $14.00 per user per month and Premium Per User at $24.00 per user per month, both paid yearly, on Microsoft's Power BI pricing page as checked today. Free embed trial tokens exist for development testing with a Pro license, but production traffic needs a capacity.
The license rule changes when the same reports are embedded for internal users, the "embed for your organization" or user-owns-data scenario, and in Teams or SharePoint. There, every viewer on an A SKU or on an F SKU smaller than F64 needs Pro, Premium Per User or an individual trial. Only P SKUs and F64 or larger let free users view the content. The trade-off between per-user licenses and a large capacity for internal audiences is worked through in Power BI Pro vs Premium Per User vs Fabric F64.
··········
POWER BI EMBEDDED COST EXAMPLE: A1 VS F8 FOR A SAAS PORTAL.
The following example is illustrative. A software vendor embeds customer dashboards in its portal with app-owns-data embedding. The largest semantic model is 2.5 GB, so it fits the 3 GB limit of A1 or F8. Two report developers hold Power BI Pro licenses, and customers view reports without any license. All figures were computed in Python from the East US list prices above.
CU_HOUR = 0.18 # Fabric pay-as-you-go, USD per CU hour, East US
RES_CU_YEAR = 938 # Fabric 1-year reservation, USD per CU per year
A1_HOUR = 1.0081 # Power BI Embedded A1, USD per hour, East US
HOURS_MONTH = 730
a1_always_on = A1_HOUR * HOURS_MONTH # 735.91
f8_payg_always_on = 8 * CU_HOUR * HOURS_MONTH # 1051.20
f8_reserved = 8 * RES_CU_YEAR / 12 # 625.33
business_hours = 12 * 22 # 264 hours
a1_business = A1_HOUR * business_hours # 266.14
f8_business = 8 * CU_HOUR * business_hours # 380.16
break_even = f8_reserved / A1_HOUR # 620.3 hours........
Scenario (illustrative) | A1 | F8 pay-as-you-go | F8 reserved 1 year |
Always on, per month | $735.91 | $1,051.20 | $625.33 |
Always on, per year | $8,830.96 | $12,614.40 | $7,504.00 |
12 hours x 22 days, per month | $266.14 | $380.16 | $625.33 |
Pro licenses for 2 developers, per month | $28.00 | $28.00 | $28.00 |
........
When the portal must answer at any hour, the reserved F8 is the cheapest of the three at $7,504 a year against $8,830.96 for an always-on A1, a difference of $1,326.96. When customers use the portal only during business hours and the vendor pauses the capacity outside them, A1 costs $266.14 a month against $380.16 for F8 on pay-as-you-go. The break-even point sits at about 620 running hours a month: below it, a paused-and-resumed A1 is cheaper than the F8 reservation; above it, the reservation wins. The ratio holds at every size, because each A SKU lists at about 70% of the matching F SKU hourly price and the reservation removes 40.5% from the F price.
Pausing behaves differently on the two products. The Power BI Embedded pricing page says that a paused service stops charging and that embedded content does not load while paused. On Fabric, pausing a capacity adds the remaining smoothed operations and carry-forward overages to the Azure bill at the moment of the pause, according to the Fabric pause-and-resume page, so a heavily used F SKU can produce a final charge when it is switched off. A reservation, being an annual purchase, gives no saving from pausing.
··········
COMMON MISREADINGS OF POWER BI EMBEDDED AND FABRIC PRICING.
The first is that "embedded" always means license-free viewers. Microsoft's documentation limits the no-license rule to the app-owns-data scenario. Internal embedding on an A SKU or on an F SKU below F64 still requires a paid license per viewer.
The second is that the A1 and F8 rows of the mapping table describe identical products. The Fabric licensing page warns against reading the table as functional or licensing equivalence, and the feature differences above (secure URL embedding, Copilot, Fabric workloads) are documented by Microsoft.
The third concerns trials. Fabric trial capacities are F4 or F64 and, according to the embedded capacity documentation, inherit the limits of the assigned SKU; free embed trial tokens are restricted to development testing. Neither replaces a paid capacity for a production portal. The trial rules and what happens when it expires are covered in the Microsoft Fabric free trial reference.
Government customers have one more rule. Microsoft states that F SKU capacities are not supported in the GCC environment, where only EM and P SKUs are available, while Azure embedded capacities are supported in GCC High and DoD.
··········
CHOOSING BETWEEN POWER BI EMBEDDED AND FABRIC CAPACITY FOR CUSTOMER REPORTS.
Microsoft confirms the hourly A SKU prices, the $0.18 per CU hour Fabric meter and the $938 per CU one-year reservation for East US on its retail price list, along with the SKU mapping, the scenario table and the license rules quoted above. Prices differ by region and currency, and enterprise agreements can change them, so the Azure pricing calculator signed in with the company's own agreement is the final check before buying.
For a customer-facing portal that only serves Power BI reports and can be paused outside business hours, an A SKU is the lower-cost option at every size. For a capacity that has to run more than about 620 hours a month, a one-year F SKU reservation costs less than the matching A SKU. Any plan that includes Copilot, secure URL embedding or other Fabric workloads requires an F SKU regardless of the hourly comparison, and any internal audience on a capacity smaller than F64 adds a Pro or Premium Per User license for each viewer.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org




