SAP is moving its enterprise software pricing from subscription licenses to a model tied directly to AI consumption, Techzine Global reported, in one of the more concrete signs that the German software vendor expects customers to pay for intelligent workloads by the drink rather than by the seat.

What does the change actually mean?

The shift reframes what an SAP customer is buying. Under that contract, a buyer pays a fixed fee per user, per module, per year, with usage rights attached to seats. Use-based AI pricing replaces that flat rate with metering tied to inference calls, tokens processed, or specific AI features invoked.

The mechanic resembles how cloud infrastructure providers charge for compute time or API calls: predictable only for buyers who can forecast their own AI activity. Annual recurring revenue (ARR) — the number SAP, like every public SaaS company, reports against — becomes harder to forecast because consumption fluctuates.

What changes for buyer and vendor?

Three tradeoffs matter most:

• Cost transparency rises for the buyer. Customers see line-item charges for each AI workflow, which makes total cost of ownership (TCO) easier to model than buried seat costs, provided the metering is legible.
• Lock-in changes shape. Switching costs shift from contract penalties to the cost of retraining or porting accumulated data and prompts.
• Vendor revenue volatility rises. Forecast accuracy becomes a real-time exercise instead of a quarter-end one.

When does usage-based AI pricing fail?

The model fails when buyers cannot predict consumption or audit the meter. A workflow that issues a million inference calls where a legacy process would have been a single database query overruns a quarterly budget without notice. Seat-based pricing hides that exposure; usage-based pricing exposes it. SAP customers who onboard AI features without instrumenting call volume will see the same shock AWS customers experienced in the early 2010s when unmetered workloads moved into EC2.

When does it work?

It works when the AI capability genuinely augments an existing process and the buyer knows how often the augmentation fires. A finance team running automated invoice matching converts its known monthly volume into projected inference costs. A product team whose AI feature is customer-facing and growth-dependent cannot.

What should product managers ask on Monday?

First, what exactly gets metered. AI use-based pricing can mean tokens, API hits, agents invoked, transactions processed, or outcome-based pricing tied to a measurable result — a closed ticket, a matched invoice, a generated report. Each transfers risk differently.

Token-based pricing puts cost risk on the buyer when models change and consume more tokens per call. Outcome-based pricing pushes risk back to the vendor but is harder to define and audit. SAP has not disclosed which metering approach it will adopt for which product line, nor the migration path for current subscription customers. The announcement signals intent, not mechanics.

Subscriptions remain stable revenue while AI continues active deployment, and metering AI consumption separately lets vendors capture upside from customers who scale agentic workflows without forcing every buyer into the same contract shape.