The Celonis Platform is licensed by users and process scope, but its infrastructure is licensed by capacity. Three units dominate the Metrics Description for the Celonis Process Intelligence - AI Platform edition: the analytical processing capacity (APC), the Celonis Compute Unit (CCU) and the AI Output. This article explains how each is defined, how entitlements are derived, how extra capacity is bought, and how the earlier Celonis Infrastructure (OCPM Enabled) edition differed. Capacity units sit beside the user and module metrics described in Celonis users, process modules and packages.
Definitions
The Metrics Description defines the APC as “the amount of uncompressed raw data that the customer ingests into the Celonis Platform concurrently at any given time for processing”.[1] It is a stock measure, not a flow: it is the volume held at any one moment, not the sum of loads over the year. Footnote 2 clarifies that purchased capacity must cover the additional storage on Celonis infrastructure, “inclusive of the transformed data and data copies required for the customer’s operation on the Celonis platform”.[1] A licence manager sizing APC therefore needs to consider transformed tables and copies, not only raw extractions.
A CCU is the unit of measure for processing capacity. Each CCU represents the equivalent compute power of a four virtual CPU instance set up in the Celocore new ETL stack for a month, and each customer’s allocation is provided every month for the duration of the contract.[1]
An AI Output is a standardized unit of consumption for AI-powered features. One AI Output is one response generated by a Process Copilot to a user query, one individual record processed and annotated by the Annotation Builder, or one response generated by the Chat API. The definition excludes usage from customers who bring their own large language models, and usage by Insights Explorer and AI Assistants.[1]
Entitlements in the AI Platform edition
The infrastructure entitlement is bought with the AI Platform and gives one productive team environment, one standard sandbox team environment and unlimited Analysts and Business Users. Capacity is derived from the annual software licence fee in US dollars:[1]
| Entitlement | Quantity | Details |
|---|---|---|
| CCUs | 1 CCU per $50,000 for annual spend from $1 to $1,000,000; 1 CCU per $100,000 for spend over $1,000,000 | Rounded up to the nearest CCU; per contract year |
| APC | 1.5 TB per $50,000 of annual software spend | Rounded up to the nearest APC unit; per contract year |
| AI Outputs | 500,000 per $50,000 of annual software spend | Rounded up; unused AI Outputs do not roll over; quotas reset at the start of each annual term |
As an illustration, an annual software fee of $200,000 would produce 4 CCUs, 6 TB of APC and 2,000,000 AI Outputs, applying the stated ratios and rounding up. The document does not say whether the $100,000-per-CCU rate for spend above $1,000,000 applies to the whole fee or only to the portion above that threshold, so a customer near that boundary should confirm the calculation with its Order. The same “(4)” footnote ties all three entitlements to the Functional Currency rule: if the Order currency differs from the customer’s reporting currency, a fixed exchange rate applies for the whole year, being the closing rate on the last trading day of the previous calendar year as published by the Board of Governors of the Federal Reserve System.[1]
Buying more
The Metrics Description gives list rates for top-ups, each annual and expressed in US dollars or the equivalent amount in the Functional Currency:[1]
- additional CCUs, sold in increments of individual CCUs at $12,000 per CCU per year;
- additional APC, sold in increments of 10 TB of storage at $10,000 per year; and
- additional AI Outputs, sold in increments of 1,000,000 at $10,000 per year.
The MSA adds that use beyond the scope of an Order must be paid for at Celonis’s then-current price list,[2] so a customer who exceeds an entitlement before buying more may be charged at current prices rather than at the Order’s rates. The MSA also allows Celonis to meter consumption.[2]
Sandboxes
Celonis offers two sandbox types, Standard and Premium. Both are testing, evaluation and release management environments that operate independently of production and cannot be connected to non-sandbox systems; their scope is limited to testing licensed scope only.[1] The Standard Sandbox is provisioned by default; the Premium Sandbox is an optional add-on for an additional flat fee.[1]
| Feature | Standard Sandbox | Premium Sandbox |
|---|---|---|
| Analyst limit | 5 Analysts | Unlimited |
| Business User limit | None (0) | Unlimited |
| Table row limit | 100,000 rows | Unlimited |
| CCUs | 1% of productive team entitlement | Shared across the customer’s environments within the total infrastructure limit |
| Team-to-team copy | Not available | Included |
The table follows the Metrics Description.[1] A sandbox is not a free place to evaluate extra processes: the document says the scope of both sandboxes is limited to testing for licensed scope only.[1] Under the older OCPM Infrastructure edition the Standard Sandbox had a 20 GB APC limit and the Premium Sandbox the greater of production APC plus 10% or 40 GB, with 30 Analysts and no Business Users.[3]
Machine Learning Workbench
The AI Platform includes a Machine Learning Workbench x16 (16 CPUs, 128 GB RAM, 200 GB disk) with every production, standard sandbox and premium sandbox licence. Larger dedicated plans (x32, x64 and x128) can be bought and split into multiple workbenches in multiples of two. A dedicated add-on must name the environment (Production, Sandbox or Premium Sandbox) and its resources are provisioned for that environment only. Where no dedicated size is selected, the x16 configuration is the default, and an auto-shutdown stops each workbench after 12 hours of non-use, except for scheduled runs.[1]
Action Flow executions and operations
Executions and Operations are metered separately from CCUs. An Execution is an action that occurs in the customer’s operational system when an Action Flow runs, each time a record is created, updated or deleted. An Operation is each trigger run or action performed by Celonis when executing scenarios using Action Flows, and each Execution is also an Operation.[3] In the November 2025 OCPM Infrastructure description the infrastructure included 50,000 Executions and 500,000 Operations per contract year, with unused Executions not rolling over.[3] The AI Platform Metrics Description in this series lists CCUs, APC and AI Outputs in its infrastructure table and does not repeat an Action Flow allowance, so customers on that edition should read their Order for any Action Flow limits. The product description lists Action Flows as included in both editions.[4]
The OCPM Infrastructure edition
The Celonis Infrastructure (OCPM Enabled) edition, no longer available to new customers,[4] used the same CCU ratios but different data and AI quotas:[3]
- APC of 30 TB per $1,000,000 of annual software spend, prorated;
- AI Outputs capped at 500,000 in aggregate per customer, scaling at 100,000 per process module and 300,000 per domain module, with higher entitlements if the AI Innovation or AI Lighthouse SKU is bought; and
- additional CCUs and storage sold at the then-current market price plus a 15% management fee, with the document pointing to Amazon AWS and Microsoft Azure instance and storage pricing as references.
The edition also set a Center of Excellence data allowance: combined ingestion for three concurrent Process Modules was restricted to 10,000,000 Data Records (a Data Record is a row in a base table), plus 100,000 additional Data Records for each $10,000 associated with the Center of Excellence SKU on the Order.[3] Because the two documents use different bases, a customer migrating between editions should recalculate each entitlement rather than assume the ratios carry over.
Counting practice
Capacity is measured by the platform. A practical review compares the Order’s entitlements with the consumption the platform reports, checking APC at the high-water mark including transformed copies, CCU use against the monthly allocation, and AI Output use against the annual quota. Because unused AI Outputs and Executions do not roll over,[1] [3] an unused quota cannot offset a later overrun. Because the Order prevails over the Metrics Description in a conflict,[1] any negotiated figures on the Order, such as a different CCU ratio, replace the standard ones.
Out of scope
This article does not cover Make credits, the Celonis for Consulting credit model, or Workforce Productivity concurrent users; those are in Celonis AI, automation and add-on licensing and Celonis legacy editions and partner programs. It does not describe hosting regions, data residency or the technical limits of connectors.