1. Functional Features of the Service
1.1. Nebius Serverless AI is a consumption-based compute platform that enables Customers to deploy AI workloads (Jobs, Endpoints, Devlabs) without provisioning or managing underlying Compute Cloud infrastructure. Workloads run on Nebius Compute Cloud resources provisioned and managed by the Service in Customer Account. 1.2. As part of the Service, the following resources are available: (a) Jobs: single-execution workloads that automatically provision compute resources, run to completion or failure, stream execution logs, and automatically release resources upon completion or cancellation. Customer may cancel a Job at any time; (b) Endpoints: persistent workloads that automatically provision Compute Cloud resources, accept incoming network requests and run continuously until Customer explicitly stops or deletes them. (c) Devlabs: interactive development environment workloads that automatically provision Compute Cloud resources, provide access to the containerized application via a managed HTTPS URL, and run until Customer stops or deletes them. Customer may stop and start a Devlab at any time; a persistent workspace is preserved while the Devlab is stopped and is deleted together with the Devlab. 1.3. Customer may: (a) submit containerized workloads with Compute Cloud resources, specified by the Customer in Console or CLI. (b) deploy and manage Endpoints and Devlabs. (c) access Virtual Machines via SSH using key pairs, subject to the limitations set forth in Section 3.6. (d) mount Object storage buckets and Shared filesystems as Container Volumes. (e) manage workload lifecycle. (f) use other functional features of the Service, which Nebius makes available to Customers. Full functional features of the Service are described in Documentation. (g) expose HTTP ports of Endpoints and Devlabs via managed HTTPS URLs and configure token-based authentication for access to them. 1.4. The Service is in Preview stage. 1.5. There is no Service Level provided for the Service.2. Pricing
2.1. The use of the Service is chargeable for the consumption of underlying Nebius Compute Cloud resources. 2.2. Service Fee is determined as per Service Rates and billing units specified on Site. Nebius may change Service Rates as stipulated by Agreement. 2.3. Service Fees for Virtual Machine usage depend on the allocated computing resources, operating system, and usage time. Service Fees are calculated for the time of using the VM in accordance with the rates set forth at https://docs.nebius.com/compute/resources/pricing, from the moment it is started until it is completely stopped. Time when the Virtual Machine is stopped is not charged. 2.4. Service Fee for Disk usage depends on usage time (from Disk creation to Disk deletion) and Disk space. The usage of Disk is charged regardless of whether a Virtual Machine is running. Pricing is based on Network SSD disk https://docs.nebius.com/compute/resources/pricing#volumes 2.5. When Jobs are cancelled by the Customer or completed, the Service automatically deletes the associated Virtual Machine and Disk. 2.6. For Endpoints, when stopped, the Service automatically deletes the associated Virtual Machine and Disk. 2.7. As part of the Service, the volume of Outgoing Traffic passing through Customer’s Virtual Machine or Disk, and other costs related to integrated Services are charged separately according to their respective Service Terms. 2.8. For Devlabs, when stopped, the Service automatically deletes the associated Virtual Machine. The Disk storing the persistent workspace is preserved and remains chargeable under Section 2.4 until the Devlab is deleted.3. Limitations
3.1. The Service is designed for single-node workloads only. Multi-node distributed workloads are not supported. 3.2. The Service does not provide automatic retry, recovery, or redundancy mechanisms. Customer is responsible for implementing application-level error handling if recovery across failures is required. 3.3. The Service does not provide Service Level Agreement uptime guarantees. Infrastructure failures may result in workload failure with no automatic recovery. 3.4. Endpoint availability depends on the underlying Virtual Machine. Customer must implement application-level health checks and redundancy if production uptime is required. 3.5. Where the Service makes a workload available through a managed HTTPS or TLS URL, the Service provides TLS termination for connections to that URL. Customer is responsible for configuring authentication where authentication is optional or customer-configurable and for implementing appropriate encryption in transit for any other inbound or outbound connections made to or initiated by Customer’s workloads. To the extent permitted by applicable law, Nebius shall not be liable for any security breach or data exposure to the extent caused by Customer’s failure to comply with this Section 3.6. Virtual Machines created by the Service are managed by the Service and are not available for direct management as Compute Cloud resources. Customer may access them via SSH using key pairs. Manual modification of these Virtual Machines may result in unexpected behavior of associated Jobs, Endpoints, or Devlabs. 3.7. Customer is responsible for persisting any results to Container Volumes. 3.8. Customer shall monitor and ensure the security and relevance of software used in containers, timely update software versions, or make configuration changes as per the Documentation or requirements published. 3.9. Customer shall counteract installing malware on Disks; software specifically designed to gain unauthorized access to information; software, databases, and other Content in violation of the intellectual property rights of third parties. 3.10. If any violations of information protection requirements threatening the operation of Platform or third-party resources are revealed on the part of Customer, Nebius may interrupt the operation of Virtual Machines associated with Jobs, Endpoints, or Devlabs that violate security requirements.4. Definitions
Endpoint means a long-running containerized workload deployed via the Service serving HTTP requests. Job means a containerized workload submitted via the Service that automatically terminates and releases resources upon completion. Devlab means an interactive containerized development environment deployed via the Service, accessible via a managed HTTPS URL, that runs until stopped or deleted by Customer. Container Volumes are storage resources that can be added to Jobs, Endpoints and Devlabs. Disk means virtual disk space used to place and store operating systems and Content. Defined in Compute Cloud https://docs.nebius.com/legal/specific-terms/compute#4-definitions Outgoing Traffic means an amount of information outgoing to external information systems and passing through the virtual computing resources of Platform provided to Customer. Information on consumed outgoing traffic is provided to Customer as Statistics aggregated from all Services. Defined in Compute Cloud https://docs.nebius.com/legal/specific-terms/compute#4-definitions Virtual Machine means virtual computing resource of the Platform, emulating hardware and used by Customer to execute containerized workloads. Virtual Machines created by the Service are provisioned in Customer’s account and are managed by the Service. Defined in Compute Cloud https://docs.nebius.com/legal/specific-terms/compute#4-definitionsWeb address: https://docs.nebius.com/legal/specific-terms/serverless-ai Publication date: September 15, 2026 Effective date: September 28, 2026 Previous version of the document: https://docs.nebius.com/legal/archive/specific-terms/serverless-ai-20260203