ALIBABA CLOUD

Aliyun

Say goodbye to pricing confusion and unexpected bandwidth billing. This comprehensive guide breaks down Alibaba Cloud Simple Application Server (SWAS), enterprise-grade ECS instances, Bailian Model Studio Token Plans, and Qoder CN agent tools—helping you build a highly reliable, evolvable cloud stack at optimal cost.

New customer offer

10% off for new customers · pay 90% of eligible price

Claim the offer through our referral link. Eligible products, account requirements and compatibility with other discounts are determined by Aliyun at checkout.

Claim new customer offer

Products and plans / 12 7 products

Last checked · Beijing time, UTC+8
Recent directory changes
  • Removed · MuleRun Enterprise
  • Removed · MuleRun Team
  • Added · Personal credit package
  • Added · Personal Subscription
  • Added · AI Agent advanced trial bundle · AI Agent advanced bundle
  • Added · OpenClaw AI assistant · AI Agent extended-use bundle
  • Added · Simple Application Server + Model Studio Bundle
  • Added · OpenClaw AI assistant
Pricing and update notes

Listings are checked roughly every 30 minutes. Prices refresh in batches, usually over a six-hour cycle. Quotes apply to the shown configuration; packages have specified validity periods and bundle prices total the listed components.
Prices do not deduct any additional new-customer offer. Confirm eligibility and renewal pricing with the provider. A previous quote should be reconfirmed before buying.

Aliyun configurations, regions, billing and prices
Product / use caseSpecifications / allowanceRegionBilling / durationReference priceUpdated · UTC+8Offer termsDetails
GPU and data computing

Function Compute GPU Resource Package

Package type
GPU Resource package
Quote details and terms
  • Resource package type: GPU Resource Package
  • GPU resource package: 500,000 GB*seconds
  • Region: Global
  • Quantity :1
  • Validity / commitment period: Year

Price for a resource package or credit allowance; usage beyond the allowance is charged separately.

Region
Global
Resource package
Validity / commitment
1 years
Quantity: 1
¥52/ package
See checkoutPlan details

A Serverless GPU deduction package designed for lightweight model inference and modern AI application serving. It eliminates the steep costs and idle waste associated with keeping persistent GPU servers running around the clock for LLMs, image synthesis, or speech processing. Ideal for independent developers, early-stage startups, and services with fluctuating traffic patterns, it leverages Function Compute's pay-per-request and scale-to-zero capabilities to drastically reduce inference costs with zero server maintenance overhead.

GPU and data computing

PAI-EAS Model Inference Resource Package

Committed spend (CNY)
200
Quote details and terms
  • Spending commitment period: None
  • Savings plan type: Starter Plan
  • Committed spend (yuan): 200
  • Installment payment type: All upfront
  • Quantity :1
  • Validity / commitment period: Year

Limit 1; Limit per verified identity ; Cannot combine with coupons ; Resource package 或抵扣 Allowance Price ,超出 Allowance 另计。

Promotion Ends: 2027/6/30 00:00:00 (Beijing Time)

—Resource package
Validity / commitment
1 years
Quantity: 1
¥59/ package
  • Limit per verified identity
  • Cannot combine with coupons
  • Limit 1
Plan details

An introductory discount resource pack for deploying model serving endpoints on PAI-EAS (Elastic Algorithm Service). It resolves the engineering hurdles of converting trained machine learning models into reliable, low-latency, and auto-scaling production APIs. Perfect for algorithm teams and engineers validating prototypes or launching commercial features, this discounted package offers an affordable path to enterprise-grade inference infrastructure without the burden of building custom serving gateways from scratch.

GPU and data computing

PAI-DLC Model Training Resource Package

Compute-hour allowance
100 compute-hours
Quote details and terms
  • Resource package edition: Starter
  • Compute-hour capacity: 100 compute-hours
  • Package type: DLC Mainland China Public Cloud General Purpose
  • Quantity :1
  • Validity / commitment :month

Price for a resource package or credit allowance; usage beyond the allowance is charged separately.

—Resource package
Validity / commitment
1 month
Quantity: 1
¥200/ package
See checkoutPlan details

A cloud container compute package purpose-built for model fine-tuning and deep learning training jobs on PAI-DLC. It helps algorithm engineers and researchers run domain adaptation, parameter tuning, and custom training workloads without the overhead of maintaining GPU drivers, cluster orchestration, or distributed training setups. Suitable for agile teams conducting periodic experiments, it lets you dispatch containerized training tasks on demand and release resources upon completion to maximize budget efficiency.

GPU and data computing

PAI-DSW Model Development Resource Package

Capacity
100 compute-hours
Quote details and terms
  • Resource package edition: Starter
  • Capacity: 100 compute-hours
  • Package type: DSW Mainland China Public Cloud General Purpose
  • Quantity :1
  • Validity / commitment period: Year

Limit per verified identity ; Cannot combine with coupons ;Limit 1; Resource package 或抵扣 Allowance Price ,超出 Allowance 另计。

Promotion Ends: 2027/6/30 00:00:00 (Beijing Time)

—Resource package
Validity / commitment
1 years
Quantity: 1
¥59/ package
  • Limit per verified identity
  • Cannot combine with coupons
  • Limit 1
Plan details

An entry-level resource pack for interactive AI development on PAI-DSW (Data Science Workshop), built for data scientists, algorithm engineers, and researchers. It eliminates the limitations of local workstations, tangled package dependencies, and costly idle equipment by delivering ready-to-code cloud JupyterLab and VS Code environments with on-demand GPU acceleration. With deep-discount credits, it offers an accessible starting point for data preparation, paper replication, and rapid model prototyping in a standardized runtime.

GPU and data computing

GPU Cloud Servers gn6i · T4

Buying advice →
CPU
4 vCPU
RAM
15 GiB
System disk
Ultra Disk / 20GiB
Bandwidth
Fixed bandwidth / 0Mbps
Quote details and terms
  • Region : North China 2( Beijing )
  • Availability zone: Randomly assigned
  • Network type: Virtual Private Cloud
  • Instance specification: GPU compute-optimized gn6i, 4vCPU 15GiB (ecs.gn6i-c4g1.xlarge)
  • System disk: cloud_efficiency / 20GiB
  • Bandwidth: Fixed bandwidth / 0Mbps
  • Quantity :1
  • Billing period :month

Limit per verified identity ;Limit 1。

Promotion Ends: 2027/4/1 00:00:00 (Beijing Time)

Region
North China 2( Beijing )
Recurring billing
Duration
1 month
Quantity: 1
¥1,681/ month
  • Limit per verified identity
  • Limit 1
Plan details

An economical GPU cloud instance optimized for online inference and graphics/video workloads, powered by NVIDIA T4 accelerators to overcome high serving latency, prohibitive deployment expenses of training-grade GPUs, and heavy video transcoding demands. It is tailored for teams running computer vision, speech recognition, recommendation engines, and cloud rendering services in production. Leveraging Turing Tensor Cores with exceptional INT8/FP16 mixed-precision performance and energy efficiency, it delivers responsive inference and smooth transcoding while keeping ongoing computing budgets predictable and low.

GPU and data computing

GPU Cloud Servers gn7i · A10

Buying advice →
CPU
32 vCPU
RAM
188 GiB
System disk
ESSD / 20GiB
Bandwidth
Fixed bandwidth / 0Mbps
Quote details and terms
  • Region: China South 2 (Heyuan)
  • Availability zone: Randomly assigned
  • System disk performance level: PL0 (maximum 10,000 IOPS per disk)
  • Network type: Virtual Private Cloud
  • Instance specification: GPU compute-optimized gn7i, 32vCPU 188GiB (ecs.gn7i-c32g1.8xlarge)
  • System disk: cloud_essd / 20GiB
  • Bandwidth: Fixed bandwidth / 0Mbps
  • Quantity :1
  • Billing period :month

Limit per verified identity ;Limit 1。

Promotion Ends: 2027/4/1 00:00:00 (Beijing Time)

Region
South China 2( Heyuan )
Recurring billing
Duration
1 month
Quantity: 1
¥2,884.59/ month
  • Limit per verified identity
  • Limit 1
Plan details

An enterprise-grade high-performance GPU cloud instance powered by NVIDIA A10 Tensor Core GPUs, designed as a dependable dedicated foundation for medium-to-heavy AI workloads. It is optimized for sustained production tasks including concurrent LLM inference, moderate model fine-tuning, AI video transcoding, and 3D cloud rendering. For growing businesses and AI platforms with steady traffic and strict latency SLAs, this monthly dedicated instance protects critical workloads from resource contention while delivering guaranteed availability and predictable budgeting.

GPU and data computing

GPU Cloud Servers gn6v · V100

Buying advice →
CPU
8 vCPU
RAM
32 GiB
System disk
Ultra Disk / 20GiB
Bandwidth
Fixed bandwidth / 0Mbps
Quote details and terms
  • Region : North China 2( Beijing )
  • Availability zone: Randomly assigned
  • Network type: Virtual Private Cloud
  • Instance specification: GPU compute-optimized gn6v, 8vCPU 32GiB (ecs.gn6v-c8g1.2xlarge)
  • System disk: cloud_efficiency / 20GiB
  • Bandwidth: Fixed bandwidth / 0Mbps
  • Quantity :1
  • Billing period :month

Limit per verified identity ;Limit 1。

Promotion Ends: 2027/4/1 00:00:00 (Beijing Time)

Region
North China 2( Beijing )
Recurring billing
Duration
1 month
Quantity: 1
¥3,817/ month
  • Limit per verified identity
  • Limit 1
Plan details

Designed for high-performance computing scenarios, this GPU cloud server is powered by NVIDIA V100 accelerators with high-bandwidth memory to tackle deep learning model training, large-scale matrix operations, and complex scientific simulations that overwhelm traditional CPUs. It is ideal for academic researchers, enterprise AI labs, and engineers working on high-precision HPC workloads such as molecular dynamics and seismic processing. By eliminating the heavy capital expenses and maintenance overhead of on-premises clusters, it delivers an out-of-the-box cloud environment with broad framework compatibility, robust double-precision float (FP64) performance, and proven tensor computing capabilities for stable, efficient model iteration.

GPU and data computing

Elasticsearch Serverless Resource Package

Package type
Elasticsearch Serverless compute Resource package
Compute package capacity
100 CU* hours
Quote details and terms
  • Resource package type: Elasticsearch Serverless Compute Resource Package
  • Supported application types: All types
  • Compute resource package capacity: 100 CU*hours
  • Quantity :1
  • Validity / commitment period: Year

Price for a resource package or credit allowance; usage beyond the allowance is charged separately.

—Resource package
Validity / commitment
1 years
Quantity: 1
¥38/ package
See checkoutPlan details

Designed for maintenance-free search and log analytics, this serverless computing deduction pack eliminates the hassle of node capacity planning, shard sizing, and paying for idle cluster resources inherent in traditional Elasticsearch deployments. It is ideal for teams and developers with fluctuating search traffic or bursty log indexing workloads who want the full power of Elasticsearch without managing JVM heaps or cluster scaling. The pack directly offsets consumed Compute Units (CUs), providing native search capabilities and true pay-as-you-go economics with predictable cost savings.

GPU and data computing

OpenSearch · Intelligent Q&A Resource Package

Compute package specification
100CU
Quote details and terms
  • Region: China East 2 (Shanghai)
  • Package type: Basic Package
  • Compute resource package specification: 100CU
  • Quantity :1
  • Validity / commitment period: Year

Price for a resource package or credit allowance; usage beyond the allowance is charged separately.

Region
East China 2( Shanghai )
Resource package
Validity / commitment
1 years
Quantity: 1
¥52.5/ package
See checkoutPlan details

An integrated RAG solution pack for intelligent knowledge retrieval and generative QA, addressing the pain points of keyword search failing to grasp natural conversational intent and generic LLMs hallucinating without private domain grounding. It is tailored for enterprise knowledge managers, customer service operations, and SaaS developers looking to launch accurate internal assistants or document search bots. By combining enterprise-grade indexing, semantic recall, and reranking with LLM summarization, it eliminates the complexity of assembling custom RAG pipelines, letting you deploy reliable, source-cited conversational search at predictable usage costs.

GPU and data computing

MaxCompute Resource Package Bundle

Component 1
500 CU·H Compute Savings Package
Component 2
100 GB Storage Savings Package
Quote details and terms
  • MaxCompute Compute Savings Package
  • Compute package type: General Purpose Mainland China Public Cloud
  • Compute package specification: 500 units
  • Compute package capacity: 500
  • Quantity :1
  • Purchase term: Year
  • MaxCompute Storage Package
  • Storage package type: General Purpose Mainland China Public Cloud
  • Storage package specification: 8 units
  • Storage package capacity: 8
  • Quantity :1
  • Purchase term: Year
—Bundle
Duration
1 years
Quantity: 1See quote for each component's duration
¥474/ bundle ( Bundle total )
See checkoutPlan details

Designed for enterprise-grade data warehousing and batch analytics, this combined compute and storage pack eliminates the steep infrastructure costs and operational barriers of self-hosting massive Hadoop clusters, solving compute and storage bottlenecks in large-scale data ingestion, ETL processing, and recurring BI reporting. It is suited for data engineers, BI analysts, and teams processing large volumes of structured data. By bundling computing CU hours and storage capacity, it provides an economical, predictable baseline to leverage Alibaba Cloud's serverless exabyte-scale data warehouse engine with instant elastic scalability.

GPU and data computing

EMR Serverless StarRocks · Starter Edition

Detail 1
StarRocks compatible
Quote details and terms
  • EMR Serverless StarRocks Starter Edition

Promotional list price; eligible specifications and purchase requirements are subject to the official website.

—Listed unit

Billed per month

¥59/ month
See checkoutPlan details

A fully managed, vectorized OLAP engine designed to eliminate slow real-time dashboards, sluggish multi-table joins, and delayed analytical reporting. It is built for business analysts, data warehouse engineers, and growth teams requiring sub-second interactive queries and responsive BI dashboards. This starter edition offers an accessible, low-friction entry point to StarRocks' powerful Cost-Based Optimizer and vectorized execution without the burden of manual cluster management, enabling agile data exploration and high-concurrency ad-hoc analytics at an economical price point.

GPU and data computing

Hologres Compute Savings Package

Compute package capacity
15,000 CU*H(units)
Quote details and terms
  • Compute package type: Mainland China - General Purpose Compute Package
  • Compute package capacity: 15,000 CU*H (units)
  • Quantity :1
  • Validity / commitment period: Year

Price for a resource package or credit allowance; usage beyond the allowance is charged separately.

—Resource package
Validity / commitment
1 years
Quantity: 1
¥5,313/ package
See checkoutPlan details

A dedicated compute deduction pack for real-time warehousing and high-concurrency analytical data services (HSAP), eliminating the pain of maintaining disparate systems for real-time stream ingestion, OLAP reporting, and Key-Value point lookups. It is ideal for data engineers and system architects building real-time dashboards, user profiling engines, recommendation feature stores, or high-QPS live data APIs. With native PostgreSQL compatibility and direct acceleration for MaxCompute datasets, this package lowers the cost of continuous high-throughput stream processing and sub-second interactive analytics.

GPU and data computing

DataWorks

Detail 1
Data development and governance
Quote details and terms
  • Selected plan: Professional Edition
  • Region: China South 1 (Shenzhen)
  • Quantity :1
  • Billing period :month
Region
South China 1( Shenzhen )
Recurring billing
Duration
1 month
Quantity: 1
¥5,000/ month
See checkoutPlan details

Alibaba Cloud's unified big data development and governance platform, addressing the chaos of siloed data sources, brittle script-based scheduling, untracked data lineage, and compliance risks across heterogeneous environments. It is designed for enterprise data platform architects, ETL developers, and data governance managers orchestrating complex workflows across engines like MaxCompute, EMR, and Hologres. By unifying multi-source data integration, visual pipeline scheduling, collaborative development, data quality inspection, and granular security controls, it transforms fragmented scripts into an audited, reliable production data factory.

GPU and data computing

DataWorks Dedicated Resources

Detail 1
Subscription
Quote details and terms
  • Region: China East 2 (Shanghai)
  • Dedicated resource type: Dedicated data integration resources
  • Dedicated data integration resource group type: Standard
  • Dedicated data integration resources: 4 vCPU 8 GiB
  • Resource quantity: 1
  • Quantity :1
  • Billing period :month
Region
East China 2( Shanghai )
Recurring billing
Duration
1 month
Quantity: 1
¥492.5/ month
See checkoutPlan details

A dedicated execution environment designed to resolve complex network connectivity hurdles and ensure production-grade job stability. When organizations need to bridge data sync across isolated VPCs, on-premises IDCs, or dedicated lease lines—or prevent mission-critical ETL jobs from queuing up in shared multitenant pools—DataWorks Dedicated Resources provide strictly isolated compute and memory. It is ideal for enterprise data teams requiring rigorous network security, private line interoperability, and guaranteed pipeline SLAs.

GPU and data computing

DataWorks Resource Group Savings Package

Transfer package capacity
300CUH
Quote details and terms
  • Data transfer package capacity: 300CUH
  • Eligible region: Mainland China public cloud
  • Quantity :1
  • Validity / commitment :3 month

Price for a resource package or credit allowance; usage beyond the allowance is charged separately.

—Resource package
Validity / commitment
3 month
Quantity: 1
¥12.5/ package
See checkoutPlan details

A prepaid allowance package designed to lock in volume discounts and offset compute consumption (CUH) generated by DataWorks general-purpose resource groups across data synchronization, workflow scheduling, and DataService APIs. It eliminates unpredictable bill spikes and higher unit prices associated with pure on-demand pay-as-you-go billing by automatically deducting runtime charges against the balance. It is ideal for teams running regular data pipelines and batch syncs who want to lower ongoing operational spend.

GPU and data computing

Alibaba Cloud Elasticsearch

Detail 1
Search and analytics
Quote details and terms
  • Instance type: General Business Edition
  • Elasticsearch version: 8.9 (New)
  • Initial scenario configuration: General purpose
  • Deployment mode: New cloud-native control plane
  • Region : North China 2( Beijing )
  • Availability zone: Beijing Zone K
  • Availability zones: Single zone
  • Cold data node: No
  • Data node: Yes
  • Data node specification family: Next-generation cloud disk
  • Data node specification: 2 cores, 4G developer configuration
  • Data node storage type: Ultra disk
  • Cloud disk encryption: No
  • Storage per data node: 20 GB
  • Number of data nodes: 3
  • Kibana node: Yes
  • Kibana specification: 1 core, 2G
  • Dedicated master node: No
  • Coordinating node: No
  • Elastic data node: No
  • Network type: Virtual Private Cloud
  • Virtual Private Cloud: Select a VPC
  • vSwitch: Select a vSwitch
  • Login name: elastic
  • Quantity :1
  • Purchase term: Year

Promotion Ends: 2029/4/1 00:00:00 (Beijing Time)

Region
North China 2( Beijing )
Recurring billing
Duration
1 years
Quantity: 1
¥3,482.28/ year
See checkoutPlan details

A fully managed distributed search and real-time analytics service that removes the heavy operational burdens of self-hosted clusters, such as fault recovery, shard rebalancing, hot-warm data tiering, and rolling upgrades. Pre-configured with native Kibana, commercial security features, and cloud-optimized kernel enhancements, it empowers engineering and DevOps teams to rapidly deliver millisecond-level full-text search, centralized log monitoring, distributed tracing, and operational dashboards with zero infrastructure maintenance.

An enterprise-grade, fully managed stream processing platform deeply integrated with Apache Flink, built to tackle ultra-low latency, massive state management, and end-to-end Exactly-Once consistency across massive data streams. With turnkey streaming SQL authoring, dynamic auto-scaling, and cluster-free management, it is designed for backend and data engineering teams building real-time data warehouses, live executive dashboards, financial fraud detection systems, and IoT anomaly alerts.

GPU and data computing

OpenSearch

Storage capacity
1GB
Quote details and terms
  • Region: China South 1 (Shenzhen)
  • Application type: Advanced Edition
  • Edition type: General Edition
  • Purchased configuration: Shared general purpose
  • Storage capacity: 1GB
  • Compute resources: 20LCU
  • Quantity :1
  • Purchase term: Year

Promotion Ends: 2029/4/1 00:00:00 (Beijing Time)

Region
South China 1( Shenzhen )
Recurring billing
Duration
1 years
Quantity: 1
¥3,151.7/ year
See checkoutPlan details

A turnkey, search-as-a-service platform that empowers developers to integrate intelligent search into their applications without building search infrastructure or training NLP models from scratch. It comes pre-loaded with advanced capabilities including domain-specific tokenization, typo tolerance, synonym expansion, query intent recognition, and multimodal vector search with customizable ranking algorithms. It is tailored for e-commerce platforms, content aggregators, and enterprise portals aiming to boost search relevance and user discovery.

GPU and data computing

Elastic Algorithm Service EAS

System disk
200 GB
Quote details and terms
  • Region: China East 1 (Hangzhou)
  • Availability zone: Randomly assigned
  • Machine specification: 8 cores, 32G + 1 V100 (16G)
  • System disk: 200 GB
  • Purchase method: Create a resource group
  • Quantity :1
  • Billing period :month

Promotion Ends: 2026/12/30 00:00:00 (Beijing Time)

Region
East China 1( Hangzhou )
Recurring billing
Duration
1 month
Quantity: 1
¥5,238.75/ month
See checkoutPlan details

A production-grade online model serving and inference hosting platform built to bridge the engineering gap between experimental models and high-concurrency production APIs. It orchestrates heterogeneous GPU/CPU resources, automated dynamic scaling, seamless blue-green rollouts, and canary traffic routing while providing built-in kernel-level acceleration for large language models and deep neural networks. It enables AI developers and algorithm teams to deploy models as robust, low-latency, scalable APIs without managing cluster infrastructure.

GPU and data computing

MaxCompute

Detail 1
Cloud-native data processing
Quote details and terms
  • Region: East China 2 (Shanghai)
  • Specification type: Standard compute resources
  • Compute resource units: 50CU
  • Non-reserved compute resources: 0CU
  • Duration: 1 month
  • Quantity :1
  • Billing period :month
Region
East China 2( Shanghai )
Recurring billing
Duration
1 month
Quantity: 1
¥7,500/ month
See checkoutPlan details

A cloud-native, serverless enterprise data warehouse and batch computing platform that eliminates the idle hardware overhead, capacity limits, and operational toil of self-managed big data clusters. Featuring decoupled storage and compute, it scales seamlessly to process datasets from gigabytes to exabytes on demand with ANSI SQL support and multi-tenant security isolation. It is the core foundation for data architects and enterprise analytics teams orchestrating massive offline ETL pipelines, unified lakehouses, and cross-domain business intelligence.

Alibaba Cloud Complete Buyer's Guide: From SWAS and ECS to Bailian AI Infrastructure

In today's cloud landscape—deeply infused with generative AI and agentic automation—buying a virtual server is no longer just about tallying cores, gigabytes, and megabits of bandwidth. Budget VPS providers frequently penalize users with aggressive CPU throttling during spikes, packet loss exceeding 30% during peak evening hours, and hidden overage penalties when traffic surges. Hosting your production workloads or personal innovations on a premier global hyper-scaler ensures you acquire not just raw compute, but a comprehensive SLA-backed ecosystem backed by proprietary hardware acceleration, multi-AZ resilience, and native AI integration.

As a global cloud leader continuously recognized in the Gartner Magic Quadrant for Cloud Infrastructure and Platform Services, Alibaba Cloud in 2026 delivers an exceptionally compelling lineup: Simple Application Servers (SWAS) upgraded to default 200Mbps peak bandwidth, enterprise-grade ECS e-instances breaking into sub-100-RMB pricing, and the Bailian Model Studio (Qwen3.8-Max / Wan3.0) together with Qoder CN agent tools forging an end-to-end highway from silicon to autonomous software engineering.

Whether you are launching an author blog, deploying an overseas WooCommerce storefront, or building autonomous AI agents, this guide provides the concrete numbers, practical step-by-step setups, and frank trade-offs you need to allocate every dollar wisely.


The Crossroads: Simple Application Server (SWAS) vs. Elastic Compute Service (ECS)#

The most frequent question on the Alibaba Cloud console is straightforward: If Simple Application Servers cost under a hundred RMB a year and boast huge bandwidth, why does ECS exist, and which one should I purchase?

These two flagship computing services serve fundamentally different operational mindsets.

┌────────────────────────────────────────────────────────────────────────┐
│                     Alibaba Cloud Decision Blueprint                   │
└────────────────────────────────────────────────────────────────────────┘
                                    │
                    Is your workload a standalone app or a cluster?
                                    │
            ┌───────────────────────┴───────────────────────┐
            ▼                                               ▼
   【Standalone / Single-node Application】         【Enterprise Production / Scalable Services】
            │                                               │
   Want zero-maintenance, bundled billing?         Need custom VPC subnets, cross-AZ failover?
            │                                               │
            ▼                                               ▼
   【Simple Application Server (SWAS)】             【Cloud Server ECS (e / u1 Series)】
   • 200Mbps peak burst bandwidth                   • Custom CIPU silicon, 99.975% instance SLA
   • Pre-baked Docker / BT-Panel / OpenClaw         • Seamless VPC links to RDS, SLB, and OSS
   • Best for blogs, overseas stores, agents        • Best for high-concurrency APIs & microservices

1. Simple Application Server (SWAS): Peak Bandwidth & Predictable Budgeting#

SWAS is engineered specifically for developers, content creators, and small-to-medium businesses requiring out-of-the-box agility:

  • Predictable All-in-One Billing: Compute, NVMe SSD storage, and generous monthly egress allowances are bundled together into a single transparent price tag. You will never open your monthly statement to discover unexpected bandwidth surge charges.
  • Upgraded 200Mbps Peak Burst Bandwidth: General-tier SWAS instances now feature default 200Mbps peak public bandwidth. When visitors hit your portfolio or web service, pages render in milliseconds without the excruciating lag typical of traditional 1Mbps–3Mbps VPS throttles.
  • Specialized Workload Sub-families:
    • General Tier: 2C2G (from 68 RMB/year) and 2C4G configurations deliver exceptional value for general web stacks and staging environments.
    • International (Overseas Linux): Route traffic across optimized BGP_NCO international backbones without mainland ICP registration, offering friction-free worldwide speed.
    • Multi-IP Tier: Provides 2–3 dedicated IPv4 addresses per host, essential for multi-account management, proxy pools, and independent domain isolation.
    • Capacity Tier: Ships with 60GiB+ NVMe system disks, eliminating disk saturation anxieties for database and media-heavy setups.

2. Cloud Server ECS: Enterprise Resilience & Boundless Scalability#

For transactional commerce, high-concurrency APIs, or multi-tier corporate architectures, ECS is the true standard:

  • 99.975% Single-Instance SLA & In-house CIPU Acceleration: Powered by Alibaba Cloud's proprietary Cloud Infrastructure Processing Unit (CIPU), ECS pushes network throughput up to 400Gbps, handling up to 60 million PPS and 3.6 million IOPS with microsecond-level latency (down to 8µs). Transparent hot migration protects workloads against bare-metal hardware faults.
  • The e-Instance (Economic) & u1 Instance (Universal Compute): Breaking historical enterprise pricing barriers, the 2C2G 3Mbps e-instance starts at 99 RMB/year, granting developers uncompromised VPC isolation, enterprise security groups, and console access at mass-market rates.
  • Seamless Intranet Connectivity: ECS instances reside within customizable Virtual Private Clouds (VPCs). High-speed links to cloud databases (RDS), object storage (OSS), and load balancers (ALB) incur zero internal network egress charges.
Evaluation MetricSimple Application Server (SWAS)Cloud Server ECS (e / u1)Recommendation
Public Bandwidth200Mbps peak burst with bundled monthly quota1Mbps–100Mbps dedicated or pay-per-trafficChoose SWAS for bursty web traffic; choose ECS for 24/7 steady streaming
Network TopologyFixed single-NIC topology; minimal manual routingFull VPC SDN with custom subnets, route tables, and secondary ENIsChoose ECS whenever multi-tier enterprise network segregation is needed
Management BarrierVery low (1-click app templates, web firewall)Intermediate (full Linux/Windows server administration)Solo developers love SWAS; Ops/SRE teams require ECS
Pricing FloorFrom 68 RMB/year (maximum entry savings)From 99 RMB/year (enterprise baseline)For budget-sensitive prototyping pick SWAS; for commercial revenue pick ECS
Evolution Path1-click snapshot migration to standard ECSScale specs up/down, attach block disks, auto-scaleSmoothly upgrade from SWAS to ECS as your scale expands

Next-Gen AI Compute: Bailian Model Studio & Qoder CN Agent Ecosystem#

In 2026, a server is more than an engine for Nginx and MySQL—it acts as the operational nerve center executing and coordinating autonomous AI agents. Alibaba Cloud unites IaaS compute with a comprehensive AI application suite.

1. Bailian Model Studio: Qwen3.8-Max & Wan3.0#

Alibaba Cloud Bailian (Model Studio) delivers end-to-end MaaS spanning infrastructure, model serving, and workflow orchestration:

  • Flagship Qwen3.8-Max (MoE Architecture): With 2.4 trillion parameters, Qwen3.8-Max excels in autonomous programming, deep visual reasoning, and multi-step plan execution. It processes massive multi-page PDF filings and multi-modal media with exceptional fidelity.
  • Wan3.0 Video Generation: Natively synthesizes up to 30-second high-definition clips with unified audiovisual scheduling, transforming e-commerce product promos, social media marketing, and dynamic content pipelines.
  • Token Plan Universal Savings (Up to 55% Off): Rather than paying fluctuating pay-as-you-go rates, Token Plan subscriptions provide shared credit pools valid across 150+ leading models (Qwen, Wan, DeepSeek, GLM). Standard and Pro tiers integrate 12 Agent Harness privileges, with excess calls discounted by 12%.
  • Bailian CLI for Vibe Coding: Installed via a single bash or npm script, the Bailian CLI equips local and remote terminal agents to invoke managed RAG vector indexes and registered Model Context Protocol (MCP) toolkits directly.

2. Qoder CN: Enterprise-Grade AI Coding Agent Workbench#

Moving beyond conventional line-by-line copilot autocompletion, Qoder CN introduces a comprehensive agentic workflow:

  • Flexible Deployment Interfaces: Available as a standalone Native AI IDE, plugins for JetBrains IDEs, and an interactive terminal CLI.
  • Autonomous Execution via Agent Harness: When assigned complex refactoring or bug fix tickets, Qoder CN autonomously indexes repository context, creates execution roadmaps, edits files, executes unit tests, and iterates against stack traces until the build succeeds.
  • Enterprise Security and Privacy Guarantees: Accredited with China Academy of Information and Communications Technology (CAICT) 4+ certification and recognized by Gartner, user source code is utilized solely for inference and is never retained for secondary training.

Step-by-Step Hands-On Guides#

Here are two proven, production-grade tutorials designed for immediate execution.

Practice 1: Deploying a 24/7 OpenClaw AI Agent / Docker Stack on SWAS in 5 Minutes#

Step 1: Provisioning the SWAS Instance#

  1. Navigate to the SWAS Product Portal.
  2. Select Region: If targeting Chinese domestic audiences, select regions physically closest to your users (e.g., East China, South China). For global projects or exempt workflows, choose Hong Kong, Singapore, or US West.
  3. Image Selection:
    • For an automated AI assistant with zero coding overhead, pick OpenClaw;
    • For visual multi-site administration, choose BT-Panel 9.2.0 Alibaba Cloud Edition;
    • For custom containers, choose Docker on Ubuntu/Debian.
  4. Select the 2C2G (200Mbps peak bandwidth) plan and complete provisioning. The instance will spin up within 30 to 60 seconds.

Step 2: Configuring Inbound Firewalls#

Inside the SWAS console under Security > Firewall:

  • Open the required inbound ports (e.g., port 8888 for BT-Panel, 80/443 for HTTP/HTTPS, or your custom webhook listener ports).
  • Security Best Practice: Restrict SSH port 22 to your dedicated static IP or VPN gateway to block brute-force scanners.

Step 3: Logging In and Hooking up Bailian Model Keys#

Log in via the web-based Workbench or local terminal SSH, then configure your environment:

bash
# 1. Verify Docker operational status
sudo systemctl status docker

# 2. Install the Alibaba Cloud Bailian CLI suite
curl -fsSL https://bailian.aliyun.com/install-cli.sh | bash

# 3. Export your Bailian API key (obtained from the Token Plan dashboard)
export DASHSCOPE_API_KEY="sk-your-bailian-api-key-here"

# 4. Run an initial test prompt with Qwen3.8-Max
bailian run qwen3.8-max --prompt "Summarize my cloud agent operational readiness in two sentences."

In less than five minutes, your virtual server is equipped with world-class frontier intelligence.

Practice 2: Building a Resilient Web Architecture with ECS + RDS MySQL + OSS (Zero-Egress Cost)#

Separating application compute, transactional databases, and static assets prevents single points of failure:

                                    Internet
                                       │
                                       ▼
                         ┌───────────────────────────┐
                         │   Elastic Public IP (EIP) │
                         └─────────────┬─────────────┘
                                       │
                                       ▼
   ┌───────────────────────────────────────────────────────────────────────┐
   │ Alibaba Cloud VPC Subnet (192.168.0.0/16)                             │
   │                                                                       │
   │   ┌──────────────────────────┐     Private Fast Backbone (Zero Egress)│
   │   │  Cloud Server ECS (e/u1) ├───────────────────┐                   │
   │   │  Web Runtime / Nginx     │                   │                   │
   │   └─────────────┬────────────┘                   ▼                   │
   │                 │                  ┌──────────────────────────┐       │
   │                 │                  │   ApsaraDB RDS MySQL     │       │
   │                 │                  │   HA Replica (From 99 RMB)│       │
   │                 │                  └──────────────────────────┘       │
   │                 ▼                                                     │
   │   ┌──────────────────────────┐                                        │
   │   │  Object Storage (OSS)    │                                        │
   │   │  Media, Backups & Assets │                                        │
   │   └──────────────────────────┘                                        │
   └───────────────────────────────────────────────────────────────────────┘
  1. Establish a Common VPC: When purchasing an ECS e-instance and RDS MySQL, always place them in the exact same Region and VPC ID.
  2. Private Network Security Whitelisting: In the RDS console under Data Security > Whitelist, input your ECS instance's private IP (e.g., 192.168.1.10) rather than its public IP. This keeps your database completely hidden from the public internet, slashing attack surfaces while avoiding data transit costs.
  3. Binding OSS via Internal Endpoints: Offload user uploads and media assets to OSS Object Storage. Point your application SDK to the OSS Internal Endpoint. In-VPC data reads and writes flow across Alibaba Cloud's private backplane at zero network transfer cost.

Buyer's Checklist: 4 Practical Rules to Avoid Pitfalls#

Avoid common onboarding errors with these four operational guidelines:

1. Differentiate Burst Peak Bandwidth from Constant Dedicated Pipe#

The 200Mbps spec on SWAS represents peak burst capability. It accelerates initial asset transfers in seconds and uses monthly quotas (hundreds of gigabytes to terabytes) to contain expenditures. For continuous 24/7 video ingest, high-throughput proxies, or heavy file relays, select ECS with dedicated bandwidth or shared bandwidth pools to prevent quota exhaustion.

2. Understand Mainland ICP Filing vs. Overseas Deployment#

  • Mainland China Nodes: Under local regulations, web services hosted on mainland nodes require statutory ICP registration. Alibaba Cloud provides an online filing workflow with real-time face verification. Complete this if your core audience is in Mainland China to gain access to premium low-latency mainland CDNs.
  • Exempt Workloads: If launching global Shopify/WooCommerce stores, international API webhooks, or rapid prototypes, select Hong Kong, Tokyo, Singapore, Frankfurt, or Silicon Valley regions for immediate deployment without filing requirements.

3. Decouple Critical Data from Local Storage#

Even with a 99.975% instance SLA, hardware exceptions, software bugs, and human error are inevitable. Always set up automated snapshots via Cloud Backup and offload primary data to managed RDS and OSS storage.

4. Optimize Introductory Promos and Subscription Windows#

Alibaba Cloud offers generous onboarding pricing:

  • SWAS 2C2G from 68 RMB/year;
  • ECS e-instance 2C2G 3Mbps from 99 RMB/year;
  • Bailian Token Plan AI packages starting from 10 RMB. Where available, lock in 1-year terms during promotional windows to guarantee uninterrupted runtime and avoid month-to-month price adjustments.

Final Recommendation Matrix#

User Persona & ScenarioRecommended StackKey Business Problem Solved
Indie Developers & HobbyistsSWAS 2C 2G (68 RMB/yr) + BT-Panel200Mbps instant page loads, visual GUI management, zero bill shock
Cross-border Commerce & Overseas TeamsSWAS Overseas Linux + DockerBGP_NCO global routing, immediate deployment without ICP filing
SMEs & Production PlatformsECS e-instance 2C 2G (99 RMB) + RDS MySQLProprietary CIPU architecture, 99.975% instance SLA, free internal VPC transit
AI Engineers & Agent BuildersBailian Token Plan + Qoder CN Agent SuitePooled discounts across 150+ models (Qwen3.8-Max/Wan3.0), autonomous code delivery
Remote Teams & Secure DevelopersWuying Cloud Desktop (4C8G from 59 RMB)Centralized DLP security, seamless multi-device access, zero hardware lock-in

Compute is the foundational lever that amplifies engineering speed. Explore the official links above, select the architecture tailored to your stage, and launch your next milestone on Alibaba Cloud.

Explore Alibaba Cloud Official Special Offers ↗