This topic describes the capabilities, billing methods, limits, and purchase options for Hologres AI nodes (GPU). Prices are subject to change. For the latest pricing, see the Hologres Management Console.
AI Node Overview
Hologres AI nodes are GPU-powered compute nodes optimized for AI workloads. They retain Hologres’ high-concurrency, low-latency analytics capabilities while adding GPU acceleration for large language model (LLM) inference, vector search, and intelligent analytics. With AI nodes, you can run AI services on GPU infrastructure without moving data. You obtain dedicated GPU resources. Each AI node includes built-in popular LLMs. You invoke these models using AI Functions. This enables end-to-end workflows—from data query to model inference—without data movement. It reduces data forwarding costs and accelerates real-time intelligent analytics and application development. AI nodes provide a unified foundation for recommendation, search, AI chat, and retrieval-augmented generation (RAG) scenarios.
Pricing Details
Hologres AI resources use GPU instances. You pay based on the AI node specifications. Two billing methods are available:
Subscription: Pay upfront for the AI resources you purchase.
Pay-as-you-go: Pay after use. Billing occurs hourly. You are charged whether or not the resources are in use.
When you purchase AI resources, a 200 GB system disk is automatically included at no cost.
Supported Regions
This feature is supported only in the China (Hangzhou) and China (Beijing) regions.
Limits
You must run Hologres V3.2 or later to attach AI resources. Each instance supports only one AI resource.
You can purchase only one AI node specification per instance. To change the specification, first unsubscribe from your current AI resource and then purchase a new one.
Backup and restore operations do not include AI resources. If you restore an instance as a new instance, you must repurchase AI resources to use them.
For primary/secondary instances, all AI resource operations—including purchase, upgrade or downgrade, unsubscribe, release, and model deployment—must be performed on the primary instance. Secondary instances can only query the AI resources and deployed models attached to the primary instance. Secondary instances can also invoke those AI resources using AI Functions.
Purchase AI Resources
Go to the Hologres Instance List. Click Create Instance. On the purchase page, set the following parameters. For more information, see Purchase a Hologres Instance.
Parameter name | Description |
Product Type | Select AI Resource (Subscription) or AI Resource (Pay-as-you-go). |
Region | The region where your AI resource runs. You can select only regions where AI resources are already available. Note You do not need to select a zone. The system routes and assigns zones automatically. |
Instance ID to Bind | Select the instance to bind. After binding, the AI resource becomes part of that instance’s configuration. AI Functions use it automatically. |
AI Node Specification | Select the GPU instance type that fits your workload. |
AI Node | Enter the number of AI nodes you need. |
Duration | Select how long you want to use the AI resource. Note This parameter applies only to AI Resource (Subscription). |
AI Resource Billing
Hologres AI resources use GPU instances. You pay based on the AI node specifications. Two billing methods are available: pay-as-you-go and subscription. For detailed pricing, see AI Resource Billing.
Subscription: Pay upfront for the AI resources you purchase.
Pay-as-you-go: Pay after use. Billing occurs hourly. You are charged whether or not the resources are in use.
When you purchase AI resources, a 200 GB system disk is automatically included at no cost.