As large organizations rapidly scale their generative AI initiatives, managing the underlying infrastructure becomes a top priority. In the SAP ecosystem, SAP BTP AI Core is the engine powering these advanced capabilities. However, with great AI power comes the critical responsibility of cost management, governance, and resource tracking.
As usage grows across different departments, lines of business, and use cases, keeping a tight grip on resource allocation is essential. To address these specific tracking challenges, the SAP BTP AI Core Consumption Monitor offers an open-source solution. The dashboard provides clear visibility into token data and credit consumption, allowing organizations to manage their AI infrastructure in BTP effectively.
You can check out the source code, architecture, and deployment instructions directly on GitHub: SAP-samples/btp-ai-core-consumption-monitor.
Credit Where Credit Is Due: The FinOps Inspiration
The inspiration for this tool came from an excellent community contribution by fellow SAP experts. If you haven't read it yet, I highly recommend checking out the blog post BTP FinOps: Keeping track of your credits by @Willem_Pardaens .
That post showcases a brilliant application for overall SAP BTP FinOps, helping organizations track their global credit consumption. It got me thinking: as Generative AI explodes, we need an equally robust, highly specialized “double-click” into the specific, token-heavy workloads running inside SAP BTP AI Core. The AI Core Consumption Monitor was built to fill that exact operational niche.
The Challenge: Visibility in a Generative AI World
When utilizing Large Language Models (LLMs) and other foundation models via SAP BTP AI Core, costs are primarily driven by token consumption. For enterprise architecture and finance teams, tracking raw API usage isn't enough. They need a clear mapping of how input text, output generations, and advanced features—such as grounding data for RAG pipelines—translate into billing units.
Without granular tracking, organizations risk unexpected cost spikes from data-heavy prompts, making it difficult to allocate budgets accurately across multiple lines of business or calculate the true ROI of internal AI implementations.
Introducing the AI Core Consumption Monitor
The SAP BTP AI Core Consumption Monitor is an analytical dashboard crafted to give you 360-degree visibility into your AI ecosystem. It acts as a centralized “single pane of glass” for administrators to observe, analyze, and optimize their AI consumption.
Here are the core capabilities available in the application today:
1. Granular Token Tracking
The application breaks down your consumption into granular metrics. Instead of viewing a single consolidated figure, you get a clear bifurcation of input and output tokens for each Model . This detail is crucial because foundation models often charge different rates for the prompt (input) versus the generation (output).
2. Capacity Unit (CU) Correlation
SAP BTP services consume credits via Capacity Units (CUs). The dashboard maps your token usage directly to the corresponding Capacity Unit consumption. This bridges the gap between technical metrics (tokens) and financial impact (CUs), giving procurement and IT leaders exact data for budgeting and resource optimization.
3. Proactive Notifications
To ensure you never get an unexpected bill at the end of the month, the application allows you to set up notifications. You can establish thresholds based on your expected consumption patterns and receive alerts before usage gets out of hand.
Technical Architecture
The application is built keeping SAP best practices in mind, ensuring it securely communicates with your SAP BTP environment while providing a responsive, analytical user experience.
Here is a quick look at the tech stack powering this monitor:
-
Frontend: [Vite React] for a clean, filterable dashboard interface.
-
Backend: Built using the [ SAP Cloud Application Programming Model (CAP) / Node.js] to handle data aggregation, orchestrate API requests, and execute token-to-CU calculations.
-
Database: [SAP HANA Cloud ] to securely store historical consumption data for long-term trend analysis.
-
Integration: Utilizes standard Usage Data Management Service (UAS ) API to securely fetch usage details.
What’s Next? The Roadmap
This is just the beginning. As SAP BTP AI Core and enterprise FinOps requirements evolve, this consumption monitor will evolve alongside them. I am actively working on expanding its capabilities to make it even more robust for complex enterprise environments.
Future improvements include:
| CMS Integration | Integration with Cloud Management Service (CMS) for fetching sub account information through API — enabling automatic discovery and onboarding of subaccounts |
| Robust Notification Systems | Upgrading the alert mechanism to support complex threshold rules and direct integration with enterprise communication and automation channels. |
Get Started Today!
The project is fully open-source and hosted as an official SAP Sample. I invite you to deploy it in your landscape, test it out, and share your feedback.
How is your organization currently managing and tracking its AI Core consumption? Have you started implementing FinOps frameworks for your cloud credit management?
