GPU Marketplace
Estimates only — rates are collected automatically from public provider pages and may be out of date. Prices vary by region, commitment term, and availability, and typically exclude storage, egress, and tax. Confirm current pricing with the provider before purchasing. Last collected .
Company Profile
Provider TypeIaaS
Legal EntityTencent Cloud
Infrastructure
GPU FleetNVIDIA Tesla T4, NVIDIA Tesla V100 NVLink 32GB, V100, A100
deep learning training, scientific computing, graphics and image processing, video encoding and decoding, and other highly intensive workloads
Compute & Deployment
On-DemandYou can go to the CVM purchase page to purchase on demand.
Spot / InterruptibleSpot Instance
VM-BasedHyper Computing Cluster takes high-performance CVMs as nodes
Container-BasedqGPU-based container-level resource splitting and virtualization technologies
GPU Hardware
Latest GenNVIDIA Tesla T4, NVIDIA Tesla V100 NVLink 32GB, V100 and A100
Multi-GPU Nodesmulti-instance multi-card distributed GPU training
Max GPUs/Node8
NVLinkNVIDIA Tesla V100 NVLink 32GB
Pricing Model
SubscriptionGPU instances are pay-as-you-go, where the instances are billed by second and settled hourly.
Public Pricinghttps://www.tencentcloud.com/products/gpu
Egress ChargesGPU instances in the same region can communicate over the private network, while those in different regions can only communicate over the public network, which incurs fees.
Pay-as-you-goGPU instances are pay-as-you-go, where the instances are billed by second and settled hourly.
Performance & Scaling
Multi-Node Trainingmulti-instance multi-card distributed GPU training
Elastic ScalingCloud GPU Service supports elastic scaling, so you can quickly adjust the service scale based on your current business needs.
Perf IsolationqGPU virtualization sharing technology featuring GPU computing power and VRAM isolation
Noisy NeighborqGPU virtualization sharing technology featuring GPU computing power and VRAM isolation
Developer Experience
OnboardingPurchase via the CVM console (create Hyper Computing Cluster under HPC Cluster > Cluster Resources). Instances are pay-as-you-go; billed by second and settled hourly. Must verify order and complete payment to create instances. You can purchase or release instances at any time.
TemplatesPreinstalled driver images and images with auto-installed GPU drivers, CUDA, and cuDNN.
DocumentationProvides Getting Started guides, Installing CUBA/Getting Started with NVIDIA, Computing Instance and Rendering Instance docs, and an FAQ covering Billing, Network, Storage, Regions and AZs, Security, Image. Also includes driver installation and other how-to documents (e.g., Installing NVIDIA Tesla Driver, Pricing Overview, Storage Overview, Image Types, Creating Custom Image, Creating Snapshot, Security Violation Handling Guide, Cloud Workload Protection).
Security & Compliance
Named customers: Ubitus, WeBank, WeChatPartnership with NVIDIA (CloudXR)High-performance technical capability: 100 GiB RDMA interconnection for HPC clusters
Data Center Locations
Coverage
CountriesSingapore, Thailand, China
CitiesShanghai
Latency Tiersultra-low latency
Compliance Regions
Datacenter Locations
Key Strengths
High-performance parallel computing for deep learning, scientific computing, graphics and video processing
Quick environment deployment with auto-installed GPU drivers, CUDA, cuDNN and preinstalled driver images
Provides a native acceleration engine 'TACO Kit' for distributed training and inference
Elastic scaling and qGPU-based container-level resource splitting and virtualization to increase GPU utilization
Provides a one-stop CloudXR solution in partnership with NVIDIA
High-performance HPC clusters with interconnection over a 100 GiB RDMA network and support for V100/A100 GPUs
Known Limitations
GI3X does not support instance configuration adjustment (within-family adjustment unsupported for GI3X)
An image can be shared with up to 50 users (shared image limit)
Local storage poses a risk of data loss and is only suitable for scenarios with high storage I/O performance requirements and a high-availability architecture
Several physical passthrough GPU models do not support the GRID driver; vGPU can use only the GRID driver on certain versions
Additional Information
Core Proposition
Cloud GPU Service is an elastic computing service that provides GPU computing power with high-performance parallel computing capabilities.
Notable Customers
Ubitus
WeBank
WeChat
Last updated August 2026. Information subject to change.