SAP's shift to AI Units and use-based pricing is the most consequential commercial change in its history for enterprise buyers. Pricing described by SAP itself as "neither transparent nor explainable" — combined with 12-month expiry traps and opaque consumption metering — creates systemic budget risk. This guide covers everything you need to understand, measure, and negotiate SAP AI licensing before July 2026.
SAP AI Units are the consumption-based commercial currency for SAP's AI portfolio. Introduced as part of SAP's strategic shift to use-based pricing, AI Units are purchased in blocks and consumed when users or automated processes interact with AI features across SAP's product suite — including S/4HANA, Joule, SAP Analytics Cloud, and applications built on SAP Business Technology Platform (BTP).
Think of AI Units as a prepaid credit system. You purchase a volume of units — typically in blocks of thousands or tens of thousands — and those units are drawn down as AI features are used. Each AI feature type consumes a defined number of units per interaction, though SAP does not publish a comprehensive consumption rate table, and rates can change as SAP updates its AI product portfolio.
The critical commercial features of AI Units are: they expire after 12 months from purchase regardless of consumption, they cannot be rolled over to the next year, and unused units are forfeited with no credit or compensation. This expiry structure is not incidental — it is central to SAP's revenue model for AI, creating a reliable income stream from enterprise underutilisation.
In a notable admission in SAP's 2025 enterprise AI commercial framework documentation, SAP acknowledged that its AI Unit pricing model is "neither transparent nor explainable" to enterprise buyers without specialist support. This is the pricing model SAP is asking enterprises to commit to for multi-year terms. If you're evaluating SAP AI commitments, our SAP AI Licensing Advisory service provides the independent analysis SAP's own documentation acknowledges most buyers cannot perform alone.
SAP's Business AI portfolio is structured into two tiers — Base and Premium — with fundamentally different commercial treatment. Understanding what sits in each tier is the starting point for any AI licensing assessment, and it's more complicated than SAP's marketing materials suggest.
The boundary between Base and Premium is not contractually fixed by default in SAP's standard terms. SAP regularly reclassifies features between tiers as its AI product evolves — a capability included in Base at contract signing may be reclassified to Premium in a subsequent product update, without a separate contract amendment. Enterprises should negotiate a contractually fixed feature schedule at signing that requires mutual consent before any reclassification.
Joule is SAP's generative AI assistant — the flagship commercial manifestation of SAP's AI strategy. It operates across S/4HANA, SuccessFactors, Concur, Ariba, and SAP Analytics Cloud, providing natural-language interfaces, process automation suggestions, and — at the Premium tier — agentic task execution that can initiate and complete multi-step SAP transactions autonomously.
Joule's commercial model distinguishes between two interaction types. Joule as a copilot — where a user asks Joule a question and receives a response or recommendation — is consumed at a relatively low AI Unit rate. Joule Agents — where Joule autonomously executes a defined business task, potentially spanning multiple SAP modules — consume significantly more AI Units per execution and are charged at a rate that scales with task complexity and execution steps.
The metering gap is significant. SAP's standard tooling does not provide per-task, per-user, or per-feature AI Unit consumption reporting at the granularity most enterprise finance and IT teams need for budget management. Enterprises deploying Joule at scale have limited ability to correlate consumption with business value or to identify which departments or use cases are driving the highest unit consumption. This opacity is commercially convenient for SAP and structurally disadvantageous for buyers.
SAP Joule for Developers is licensed separately from the business-user Joule copilot. Developer Joule — which provides AI-assisted ABAP Cloud coding, test generation, and documentation — is typically accessed through the Business Application Studio or SAP AI Foundation and may require separate AI Unit allocation or a developer-specific licensing SKU. Enterprises building custom extensions on BTP should assess their Joule for Developers consumption separately from their business-user AI licensing position. See our article on SAP Joule for Developers licensing for the full picture.
SAP's shift from named-user, per-seat pricing to use-based consumption pricing for AI is presented as a customer-friendly innovation — you only pay for what you use. In practice, the use-based model introduces a set of budget risks that are new to enterprise SAP licensing and that most internal finance, procurement, and IT teams are not yet equipped to manage.
AI Units expire at the end of the contract year. SAP provides no standard notification before expiry, offers no credit for unused units, and has no obligation to roll over unconsumed balances. Enterprises that purchase AI Units based on SAP's projected consumption estimates — which our experience shows are typically 25–40% higher than actual consumption in year one — will forfeit significant value at year end. The renewal conversation begins immediately after, with SAP framing low remaining units as evidence you need a larger purchase next year.
SAP's account teams provide consumption estimates to support AI Unit purchasing decisions. These estimates are based on modelling that considers your user count, planned feature adoption, and industry benchmarks from SAP's own data set. They are not independent forecasts. In our experience reviewing SAP AI proposals, SAP's consumption estimates routinely overstate expected usage — particularly in year one when Joule adoption is still ramping. Build your own consumption model before purchasing AI Units, based on your planned rollout schedule and comparable independent benchmarks.
AI adoption in enterprise organisations is slower than SAP's projections typically assume. Joule deployment requires change management, user training, process redesign, and integration testing — factors that SAP's consumption models often undercalibrate. Enterprises that purchase AI Units sized for aggressive adoption scenarios and then experience typical enterprise adoption rates will systematically underutilise their purchases.
As SAP expands its AI portfolio, new features consume AI Units. Capabilities that didn't exist at contract signing may be deployed in your landscape — consuming units without explicit approval — as part of SAP product updates. Without consumption monitoring and explicit feature activation controls, enterprises can find their AI Unit balance depleting for features they didn't specifically plan to use.
Before purchasing AI Units, get an independent consumption model. Our SAP AI Licensing Advisory service builds consumption projections from your actual system data — not SAP's commercial team's estimates.
July 2025 marked the most significant restructuring of SAP's AI and data licensing commercial model since AI Units were introduced. The changes affected how SAP packages Business AI across its cloud portfolio and introduced new consumption structures that affect enterprises already under contract as well as those evaluating new agreements.
The headline changes were: the formal introduction of Business AI Base as an explicitly included entitlement in RISE with SAP (Cloud ERP Private) and S/4HANA Cloud subscriptions; the repositioning of Business AI Premium as a separately purchasable add-on requiring AI Units; the introduction of new AI Unit SKUs specifically for Joule Agent task execution (previously these were handled under a more general AI consumption model); and changes to how BTP AI Core and AI Foundation credits are bundled with cloud subscriptions.
For enterprises with existing RISE or S/4HANA Cloud contracts signed before July 2025, the implications of these changes depend on the specific terms of your contract. SAP has positioned the July 2025 changes as beneficial — more AI included at no extra charge in Base — but the repositioning of previously ambiguous Premium features into a clearly metered model may represent a commercial deterioration for enterprises that were accessing those features without explicit charging. Pre-existing contracts should be reviewed against the July 2025 framework to identify whether any features you were consuming are now reclassified to the paid Premium tier.
SAP AI licensing is increasingly bundled into RISE and Cloud ERP Private contracts, but the bundling is more complex than the marketing suggests. Understanding exactly what AI entitlements are included in your RISE or Cloud ERP Private subscription — and what requires additional AI Unit purchases — requires contract-level analysis, not just SAP's product marketing materials.
Business AI Base is formally included in Cloud ERP Private subscriptions as of the July 2025 commercial framework. This includes the Joule copilot for standard S/4HANA interactions, embedded AI recommendations in core business processes, and a defined set of SAP AI Foundation capabilities. It does not include unlimited consumption of all AI features — specifically, it excludes Joule Agents at scale, custom AI scenario development, and any AI capabilities that SAP designates as Premium.
BTP credits included with RISE and Cloud ERP Private can be used to power some AI workloads on SAP AI Core, but the relationship between BTP credits, AI Units, and SAP AI Foundation entitlements is complex. BTP credits are a separate commercial currency from AI Units, and not all AI Unit consumption can be offset by BTP credits. Before assuming your included BTP allocation covers your AI requirements, get a detailed analysis of which AI workloads consume BTP credits, which require AI Units, and which are included in your subscription entitlement.
SAP AI Unit pricing is negotiable — significantly so. Our experience across 50+ cloud and AI contract reviews shows consistent patterns in where SAP will and won't move, and the commercial levers that produce the best outcomes for enterprise buyers.
AI Unit pricing varies by 30–50% across comparable enterprise deals. SAP's initial pricing position will almost always be at or above the top of the market range. Benchmarking requires access to comparable deal data — SAP's commercial team will tell you their pricing is competitive, but without independent reference points, you can't evaluate that claim. We benchmark every AI Unit purchase against our client reference database before any negotiation begins.
SAP's default position is that AI Units expire at 12 months with no rollover. This is negotiable — particularly for large volume commitments, multi-year deals, and in SAP's current commercial environment where cloud targets are under pressure. Rollover provisions — even partial rollover of a defined percentage of unused units — significantly reduce the financial risk of the use-based model. Some clients achieve carry-forward rights of 20–30% of unused units into the next year.
At the point of purchase, your consumption forecast is inherently uncertain. Negotiate mid-year right-sizing provisions that allow you to adjust your AI Unit volume up or down based on actual consumption data — within defined bands. This is more achievable than rollover but still requires active negotiation against SAP's standard terms.
The Base/Premium feature boundary must be contractually fixed at signing. Without a defined, signed feature schedule, SAP's product evolution will move features between tiers without your consent. Negotiate a schedule that lists what is included in Base at contract signing, and requires written amendment before any feature is reclassified to Premium. This is the single most important contractual protection in any SAP AI agreement.
Negotiate explicit consumption monitoring and metering requirements into your contract. SAP should be obligated to provide per-feature, per-department AI Unit consumption reporting on a monthly basis — not just a total consumption figure at year end. This reporting is the operational foundation for governance and the evidentiary basis for any right-sizing or renewal negotiation.
Our SAP AI Licensing Advisory service has reviewed 50+ cloud and AI contracts. We know exactly where SAP will move — and where it won't. Book a free consultation before your deal closes.
The single most important operational capability for enterprises committed to SAP AI is a robust consumption governance framework. Without governance, the use-based model will produce unpredictable spend, units will expire unused or be consumed by unapproved use cases, and each renewal will be a reactive commercial conversation rather than a planned one.
Establish a technical monitoring layer that captures AI Unit consumption at the feature level. SAP's standard SAP for Me portal provides some consumption visibility, but typically at an aggregate level that is insufficient for departmental budgeting and cost allocation. Supplement SAP's native reporting with custom monitoring built on SAP BTP or your existing IT monitoring infrastructure to capture per-feature, per-department consumption data.
Treat AI Units as a shared corporate resource with departmental allocations, similar to how mature organisations manage cloud compute costs. Establish annual AI Unit budgets by business unit based on planned AI feature adoption. Implement monthly consumption reporting against these allocations, with escalation triggers when consumption is running ahead of pace toward expiry with significant unused balance remaining.
AI Unit renewal negotiations should begin 6 months before expiry, not when SAP's account team reaches out at month 10. Use your 12 months of consumption data to build an evidence-based renewal position — documenting actual consumption versus SAP's original projections, identifying which features drove value, and sizing the renewal based on data rather than SAP's forward projections. Enterprises that enter renewal with documented consumption data consistently achieve better terms than those negotiating on the basis of uncertainty.
Our SAP AI Licensing Advisory service provides independent consumption modelling, contract review, pricing benchmarking and negotiation support for enterprise AI commitments.
Audit defence, contract negotiation, licence optimisation — buyer-side only, zero SAP affiliation.
Explore All Services → Case StudiesSee how we've helped enterprises reduce SAP spend by 30–60% and win audit disputes.
Read Case Studies →