Fabric Copilot: Billing Information Now Available

Exciting news for data enthusiasts! Microsoft Fabric’s Copilot, which empowers you to transform, enrich, and analyze data faster, is now entering its billing phase starting March 1st, 2024. This means you can leverage the power of Copilot for Power BI, Data Factory, and Data Science & Data Engineering with a flexible pay-as-you-go model.

Understanding Your Costs:

Copilot uses Fabric Capacity Units (CUs) for processing, measured by “tokens”. The cost depends on both input and output tokens:

• Input Prompt: 400 CU seconds per 1,000 tokens
• Output Completion: 1,200 CU seconds per 1,000 tokens

For example, a request with 500 input tokens and 100 output tokens would cost (500*400 + 100*1,200) / 1,000 = 320 CU seconds.

Monitoring and Transparency:

• Track your Copilot usage easily through the Fabric Capacity Metrics App under “Copilot in Fabric” starting February 2024.
• Billing details for Copilot will be clearly identified as “Copilot in Fabric” within your Fabric invoicing records.

Important Notes:

• Fabric Copilot is classified as a “background job,” allowing for smoother capacity utilization during peak hours.
• Data processing might occur in a different region for optimal resource availability, prioritizing data residency whenever possible.
• Costs are based on your billing region, regardless of the processing region.

Ready to Get Started?

With transparent pricing and flexible usage options, Fabric Copilot offers a powerful and cost-effective way to unlock insights from your data. Dive into the public preview and experience the difference!

Additional Resources:

• Announcing Fabric Copilot Pricing: https://blog.fabric.microsoft.com/en-us/blog/announcing-fabric-copilot-pricing-2/
• Fabric Documentation: https://learn.microsoft.com/en-us/fabric/

Notes about terms used

In Microsoft Fabric billing, “CU seconds” stands for Capacity Unit seconds. It’s a unit used to measure and charge for the resources consumed by your applications and services running on the Fabric platform.

Here’s a breakdown of the concept:

• Capacity Unit (CU): This is a virtual unit representing the compute resources available on the Fabric platform. It includes CPU, memory, storage, and networking resources.
• Seconds: This represents the duration for which the resources are used.

Therefore, CU seconds essentially combine the amount of computational resources used (CU) with the duration of usage (seconds), resulting in a unit that reflects the total resource consumption.

In the context of Fabric Copilot specifically, CU seconds are used to measure the cost of processing data. Each operation you perform with Copilot consumes a certain number of CU seconds based on the number and type of tokens processed. This determines the overall cost of using Copilot.

Here are some key points to remember about CU seconds:

• The cost of CU seconds varies depending on your region and subscription plan.
• Fabric offers different pricing tiers depending on your expected usage.
• You can monitor your CU second usage in the Fabric Capacity Metrics App.
• Understanding CU seconds helps you estimate and manage your costs associated with Fabric services like Copilot.

Tokens represent units used to measure the amount of data processed by Fabric Copilot, a tool for transforming, enriching, and analyzing data. They are similar to words, with 1000 tokens roughly equating to 750 actual words.

Here’s how tokens work within Fabric Copilot billing:

• Input tokens: These represent the amount of data you provide to Copilot as a starting point for its analysis or actions. For example, a prompt or query you submit would be counted as input tokens.
• Output tokens: These represent the amount of data generated by Copilot as a result of its processing. This could be the completed text, code, or analysis based on your input.

Billing is based on both input and output tokens, but at different rates:

• Input prompt: Charged at 400 CU seconds per 1,000 tokens
• Output completion: Charged at 1,200 CU seconds per 1,000 tokens

This difference reflects the computational resources required for different steps in the process. Analyzing and generating new data generally demands more resources than simply reading it.

Key points to remember about tokens in Fabric Copilot billing:

• They are not directly related to your actual data size but rather represent a scaled-up measure for billing purposes.
• The number of tokens depends on the complexity and length of your input and output data.
• Understanding token usage helps you estimate and manage your costs associated with using Copilot.

Remember, tokens are specific to Fabric Copilot within Microsoft billing and might not apply to other services or platforms.