SAP AI Budget & Forecasting: The Complete Enterprise Guide for 2026

What SAP AI Actually Is in 2026

SAP's AI portfolio has expanded rapidly and is now marketed under the umbrella term SAP Business AI. Before you can budget for it, you need to understand what you're actually buying — because SAP's marketing language is deliberately imprecise about which capabilities are included versus which require additional spend.

The core components of SAP's AI landscape in 2026 are:

• SAP Joule — SAP's generative AI copilot, embedded across S/4HANA Cloud, SuccessFactors, and other SaaS products. Basic Joule scenarios (simple queries, guided workflows) are included in RISE and GROW subscriptions. However, the enterprise-grade Joule scenarios — custom extensions, cross-system orchestration, high-volume usage — require dedicated BTP AI Core capacity.
• SAP AI Core (on BTP) — The underlying infrastructure layer for training and running AI models. Charged via BTP consumption credits. Any custom AI development, fine-tuning, or model hosting runs through AI Core and is billed against your BTP credit pool.
• SAP AI Launchpad — The management interface for AI operations. Typically bundled at low usage levels but triggers overage charges at scale.
• Embedded AI features in S/4HANA Cloud — Predictive accounting, intelligent cash management, automated invoice matching. These are contractually included in RISE subscriptions but consume BTP service units at a rate SAP rarely discloses in the base contract.
• SAP Analytics Cloud — Planning with AI — Predictive forecasting and smart grouping capabilities within SAC. Subject to its own story-point and user-count licensing model, with AI features requiring higher-tier SAC licences or capacity add-ons. See our detailed guide to SAP Analytics Cloud planning licensing.

The critical distinction SAP sales teams exploit: there is a significant difference between AI features being technically available in your subscription and being commercially included at enterprise scale. SAP calls this "embedded" AI, but the embedding is often shallow — enabling the user interface while metering the underlying compute separately.

How SAP Charges for AI: The Consumption Trap

SAP AI billing operates across three parallel cost dimensions, and most enterprises only discover all three after they've signed. SAP licence optimisation requires understanding how these dimensions interact before you can build an accurate forecast.

Dimension 1: BTP Credit Consumption

SAP Business Technology Platform is the infrastructure backbone for all AI features. AI Core, AI Launchpad, and custom AI services are all metered BTP services — they consume credits from your BTP entitlement pool. If you signed RISE before 2024 and haven't revisited your BTP entitlement, your pool was almost certainly sized for integration and extension use cases, not AI workloads. AI workloads consume 3–8x more BTP credits per user-hour than typical extension scenarios.

Dimension 2: Cloud Credit Units (CCUs)

Some SAP AI services — particularly those embedded in S/4HANA Cloud — are priced using Cloud Credit Units rather than standard BTP credits. CCUs are a separate currency that SAP introduced in 2023 and which many enterprise contracts did not anticipate. If your RISE contract is 3+ years old, you likely have no CCU allocation, meaning any AI usage that draws on CCU-based services triggers unbundled consumption charges at SAP's list prices.

Dimension 3: Capacity SKUs and Seat Upgrades

For enterprise-scale Joule deployments, SAP sells capacity-based SKUs that provide a fixed volume of AI interactions per month or year. The pricing for these SKUs is not published and varies by industry, deal size, and — crucially — how much negotiating leverage the enterprise brings to the table. Enterprises that accept SAP's initial AI capacity proposal typically overpay by 40–60% compared to negotiated rates.

SAP AI Budget Planning Framework

Effective SAP AI budget planning requires a structured framework that accounts for all three charging dimensions. We recommend a five-layer approach that enterprises can apply whether they are in early evaluation, mid-implementation, or post-go-live stabilisation.

Layer 1: Baseline Entitlement Audit

Before forecasting AI costs, document exactly what you are contractually entitled to. Pull your current BTP contract and identify: total credit allocation, allocated CCUs (if any), AI-specific service entitlements (AI Core capacity units, Launchpad access), and any included SAP Business AI scenarios listed in the commercial schedule. This audit typically reveals 2–4 contract ambiguities that require SAP clarification before budgeting is reliable.

Layer 2: Scenario Mapping

Map your intended AI use cases to specific SAP service SKUs. For each use case, document: which SAP AI service delivers it (Joule, AI Core, embedded S/4HANA feature), whether it is BTP-credit, CCU, or capacity-SKU billed, estimated monthly volume (queries, documents, users), and SAP's published or quoted consumption rate. This mapping is where most budget failures occur — enterprises model the use case without confirming the billing mechanism.

Layer 3: Consumption Rate Validation

SAP provides consumption estimators, but these tools are optimistic. Our benchmarks from 25+ enterprise AI deployments show that real-world AI consumption runs 20–40% above SAP's estimator outputs for the first 12 months, primarily because:

• User adoption ramps faster than planned once AI features are available
• Development and testing in pre-production environments consumes the same BTP credits as production
• Iterative model tuning in AI Core generates consumption that no estimator accounts for
• Background AI processes (automated matching, predictive runs) operate on schedules that aren't visible in front-end usage reports

Apply a 30% uplift buffer to any SAP-provided consumption estimate.

Layer 4: Growth Projection

Model AI consumption growth separately from ERP user growth. AI workloads scale with data volume and process automation depth, not just headcount. A common error is applying a flat 5% annual growth assumption — the same rate used for traditional SAP user licences. AI consumption in live deployments typically grows 30–60% year-on-year as use cases proliferate and automation depth increases. Use a tiered growth model: conservative (20%), base (40%), aggressive (60%).

Layer 5: Overage Risk Modelling

Model what happens when you exceed your entitlement. SAP's standard overage pricing for BTP services is 2–3x the contracted rate. For AI Core specifically, overage during peak processing periods (month-end close, year-end runs) can be severe. Build an overage risk reserve of 15–20% of your annual AI budget for the first two years of deployment.

Forecasting SAP AI Consumption: A 5-Step Model

Accurate SAP AI budget forecasting is a distinct discipline from traditional SAP licence forecasting. The consumption model introduces variability that seat-based licensing does not have. Here is the methodology our advisory team uses with enterprise clients.

Step 1: Identify AI Service Catalogue Entries

List every SAP AI service your project will consume. Reference SAP's BTP service catalogue for exact service names and billing metrics. Common entries for a full RISE AI deployment include: SAP Joule Foundation Capacity, SAP AI Core (Standard/Extended), SAP AI Launchpad, Document Information Extraction Service, and Cash Application (embedded). Each service has a different billing unit: some charge per API call, some per document page, some per model-hour, some per user-month.

Step 2: Define Usage Volume Inputs

For each service, define the primary volume driver. For Document Information Extraction, this is pages processed per month. For Joule, this is queries per active user per day. For AI Core model training, this is compute hours per training run. Gather these inputs from your process owners and IT teams — do not rely on vendor estimates. SAP sales teams typically use "best-in-class" benchmarks that overestimate AI adoption in year 1 to justify the pricing, then underestimate ongoing consumption to minimise the apparent budget impact.

Step 3: Apply Consumption Rate Coefficients

Map volumes to credit consumption using SAP's published coefficients (available in the BTP service catalogue) plus your 30% buffer. Build a monthly model, not an annual one. AI consumption is highly seasonal — many enterprises see 2–3x normal consumption during period-close processes, and SAP does not offer burst capacity pricing by default. Understanding your monthly peak versus average is essential for sizing entitlements correctly.

Step 4: Model Three Scenarios

Build conservative, base, and aggressive scenarios. The conservative scenario assumes slower-than-expected adoption and straightforward use cases. The aggressive scenario models full adoption of all planned use cases plus organic expansion of AI usage by business users. Present all three to your finance team and use the base case for budget planning with the conservative-to-aggressive spread as your risk range. SAP 5-year cost forecasting should incorporate this AI consumption model as a separate line item.

Step 5: Reconcile with Contract Entitlements

Compare your consumption model against your contract entitlements. Identify: where you are at risk of under-provisioning (overage exposure), where you have over-provisioned (cost reduction opportunity), and what additional purchases or contract amendments are required before go-live. This reconciliation step is where independent advisors add the most value — most enterprises lack the contractual expertise to map service consumption to specific commercial terms without assistance.

AI Licensing in RISE and GROW with SAP

RISE with SAP and GROW with SAP include a base layer of AI capability, but the boundary between "included" and "additional" is commercially designed to be ambiguous. Understanding this boundary is the difference between an on-budget AI deployment and an unexpected €1–5M invoice in year two.

In RISE with SAP, the included AI layer covers: basic Joule assistance in standard S/4HANA workflows, a limited BTP credit allocation (typically 10,000–50,000 credits depending on deal size), and a handful of embedded predictive features (cash flow predictions, smart payment scheduling). What it does not cover: custom AI model development, high-volume Joule usage across enterprise user populations, document processing at scale, and any AI use case requiring more than the base BTP credit allocation.

In GROW with SAP (the midmarket cloud ERP), the AI inclusion is even more limited. GROW pricing is driven by a user-count FUE (Full Use Equivalent) model, and the BTP credit pool is proportionally smaller. Any AI ambition beyond basic chatbot interactions requires a separate commercial negotiation. See our RISE with SAP advisory service for detailed guidance on navigating these negotiations.

BTP Credits and AI: What the Contract Doesn't Tell You

BTP credits are the currency of SAP AI consumption, and most enterprise contracts are silent on the details that matter most. The issues we encounter most frequently in SAP licence compliance reviews include:

• Credit expiry: BTP credits typically have a 12-month use-it-or-lose-it window. Credits allocated for year 1 that go unspent due to delayed implementation do not roll over. This creates a perverse incentive to over-consume at year-end to avoid losing credits — exactly the wrong time to be running unplanned AI workloads.
• Credit pooling restrictions: Not all BTP credits are interchangeable. SAP sometimes bundles credits with use restrictions (e.g., "for CF [Cloud Foundry] environment only" or "for integration scenarios only"). AI Core credits are not always fungible with other BTP service credits. Verify that your credit pool is unrestricted before designing AI architecture around it.
• Credit burn rate visibility: SAP's standard BTP cockpit provides consumption dashboards, but the granularity and latency of these reports is insufficient for proactive budget management. By the time a monthly overage shows up in the dashboard, you're already committed to the cost. See our guide on SAP AI consumption tracking for tools and processes to monitor burn rate in real time.
• Capacity versus consumption model: Some AI services can be procured on either a capacity basis (fixed monthly fee for a defined throughput) or a consumption basis (pay per unit). SAP defaults to consumption because it generates more revenue. Capacity pricing is almost always cheaper for predictable workloads and is available but requires negotiation to access.

Governance and Cost Controls for SAP AI

Without governance, SAP AI deployments become cost sinkholes. The consumption model creates a dynamic where every additional use case feels free at the pilot stage — the costs only become visible at scale. Implementing governance controls before go-live is essential.

Budget Governance Structure

Establish a cross-functional SAP AI governance board that includes IT, Finance, and Business stakeholders. This board should own the AI budget, approve new use cases before development begins, monitor monthly consumption against forecast, and escalate any projected overage within 48 hours of detection. The governance structure should have explicit authority to pause or throttle AI workloads if consumption trends indicate overage risk.

Technical Cost Controls

Implement technical controls at the BTP level: set up consumption alerts at 70%, 85%, and 95% of monthly budgets; establish dedicated BTP subaccounts for AI workloads to isolate consumption tracking; configure AI Core resource groups with explicit quota limits per use case; and implement automated scaling controls that cap AI Core compute at defined thresholds during off-hours.

Commercial Controls

Negotiate contractual cost controls as part of your SAP contract negotiation: overage caps that limit your liability if consumption exceeds entitlement; quarterly consumption reviews where SAP must provide 90-day forward projections; the right to purchase additional credits at contracted rates rather than list prices; and a consumption credit carryover provision that allows up to 20% of unused credits to roll into the following year.

Negotiation Leverage for SAP AI

SAP AI pricing is not fixed — it is highly negotiable, particularly at enterprise deal sizes. The problem is that most enterprise buyers negotiate their RISE contract first and treat AI as an add-on, surrendering their negotiating leverage before the AI conversation begins. The right approach is to negotiate AI and RISE as a single commercial package.

The most effective negotiation levers our clients use include:

• Competitive pressure: Microsoft Azure AI, AWS AI services, and Google Gemini are credible alternatives for many AI use cases that SAP delivers. Position your AI architecture as cloud-agnostic and demonstrate willingness to build AI capabilities outside SAP. This immediately shifts the commercial conversation.
• Volume commitment: Offer a 3-year forward consumption commitment in exchange for a capped per-unit rate and an overage multiplier cap. SAP values predictability and will offer significant discounts for committed AI consumption volumes.
• RISE renewal leverage: Use your RISE renewal as a negotiation anchor for AI pricing. SAP's biggest fear in RISE negotiations is an on-premise repatriation risk — use that leverage to extract AI credit inclusions, consumption guarantees, and overage protections in the same commercial conversation. Our detailed guide on the SAP AI negotiation approach covers the full tactical playbook.

See also our broader guide to SAP licence optimisation for the full commercial strategy framework.

Case Study: €2.8M AI Budget Overrun Prevented

A European manufacturing enterprise with 14,000 SAP users signed a 5-year RISE with SAP agreement in 2023 that included what SAP marketed as "comprehensive Business AI capabilities." Eighteen months into the contract, the enterprise began deploying Joule across their finance and supply chain organisations and activated Document Information Extraction for their accounts payable automation programme.

Six months after activation, BTP consumption was running at 340% of the contracted allocation. SAP's account team sent an overage notification projecting an additional €2.8M in BTP credit charges for the remaining contract term based on current consumption trajectories — none of which had been budgeted.

We were engaged to conduct an independent BTP consumption audit and contract review. Our findings: the original contract contained no CCU allocation and an inadequate BTP credit pool for the AI use cases being deployed. SAP had sold the RISE contract knowing the included credits were insufficient for the described AI ambitions. The enterprise had no contractual overage cap.

Our intervention: we renegotiated the AI commercial terms at mid-contract, securing a retrospective credit grant for prior overages, a 3-year forward AI credit allocation at a 52% discount to SAP's original overage pricing, and a contractual overage cap of 15% per year. Total value of the intervention: €3.1M against an advisory fee of €85,000.