Cloud provider discount & payment / Tencent Cloud

Cloud Virtual Machine (CVM)

Tencent Cloud compute — families, regions and billing

CVM is the most basic thing an overseas business buys from Tencent Cloud International, and usually the first: a virtual machine you log into, install your own OS on, and pay for by the hour or by the month. Choosing one looks like a "how many cores, how much RAM" question, but four things actually decide both the bill and the experience: the instance family, the CPU-to-memory ratio, how you pay for public bandwidth, and the region. This guide works through all four, then explains account opening and payment settlement.

Specs & pricing

InstancevCPUMemoryPublic bandwidthNetwork billingBest for
Standard S5.MEDIUM24GB1–5MbpsBandwidth or trafficWebsites, APIs, small backends
Standard S5.LARGE48GB5–10MbpsBandwidthMid-size production systems
Compute C5.LARGE816GB5–10MbpsBandwidthHigh-concurrency services, transcoding
Memory M5.2XLARGE832GB5MbpsBandwidthDatabases, Redis, in-memory queues
Standard SA5.LARGE48GB5MbpsBandwidthAMD platform, budget-sensitive work
Lighthouse1–42–8GB3–8MbpsBundled planSites, staging, light workloads

What CVM is, and how it differs from Lighthouse

CVM is a full virtual machine: pick an image, define your own security groups, attach CBS cloud disks, front it with CLB load balancing or AS autoscaling, and place it in a VPC alongside your database and object storage. Against Lighthouse, you get a wider network and instance range and the ability to join cluster architectures — at the cost of more knobs, and first-time buyers routinely over-provision bandwidth and disk.

How CVM is billed across four levers

  • Compute: on-demand bills hourly and suits development and load tests. Monthly and annual plans are prepaid, cheaper per hour, and suit steady workloads.
  • Spot instances: for interruptible batch work and rendering. Far cheaper, but reclaimable — never for a primary service.
  • Public bandwidth: bandwidth-based is a fixed, forecastable monthly cost; traffic-based bills actual egress and wins when usage is spiky.
  • Disks and resource packs: disks bill separately, and prepaid storage and bandwidth packs undercut on-demand pricing when usage is predictable.
  • Powered off is not free: compute stops billing, but disks and public egress can still cost. Unit prices follow the live official rate.

Choosing a region: start with where users are

RegionAudienceWhyWatch out
Hong KongMainland China, HK, Macau, TaiwanShortest path, lowest latencyHigher bandwidth unit price
SingaporeSoutheast and South AsiaRegional hub, stable networkingHigher latency to Europe and North America
TokyoJapan and KoreaLow latency for local audiencesInstance availability differs
Silicon ValleyNorth America west coastClose to North American usersHigher latency to Asia

Instance families: by workload shape, not price

FamilyCharacterGood forPoor fit
Standard S5Intel platform, balancedWebsites, APIs, microservicesExtreme compute or memory pressure
Standard SA5AMD platform, close to S5Budget-sensitive, general workSoftware tuned to one instruction set
Compute C5CPU-heavy ratioConcurrency, builds, transcodingMemory-bound work
Memory M5Memory-heavy ratioDatabases, Redis, queuesPure compute jobs
High I/O I5Local high-speed diskSmall-file heavy I/OCases needing bulk low-speed storage only

Tencent Cloud International vs the China site

AspectInternational siteChina site
Site and registrationintl.cloud.tencent.com, email plus company detailscloud.tencent.com, mainland identity verification
RegionsHong Kong, Singapore, Tokyo, Seoul, Silicon Valley, Frankfurt and othersMainland regions such as Guangzhou and Shanghai
PaymentInternational methods and corporate settlementMainland methods, settled in CNY
DocsEnglish-first, Chinese on some servicesComplete Chinese documentation
ICP filingOverseas regions generally avoid itMainland nodes serving the public need one

A selection order you can reuse

  • Region first, specs second. No CPU headroom compensates for the wrong region.
  • Start at 1:2 to 1:4 CPU-to-memory: memory for databases, CPU for builds and transcoding.
  • Start on traffic-based bandwidth. Read one month of billing before switching.
  • Start small. Most CVMs resize without a reboot and refund proportionally.

How account opening and payment settlement work

  1. Tell us the user region, application type and rough concurrency, and we will recommend a family, region and bandwidth.
  2. We walk you through official registration and provisioning. No personal credit card binding and no personal identity documents on your side — the account is yours once it is live.
  3. Once the account works you order and pay in the official console; if you prefer we settle payment, we pay at the discounted rate and issue a statement.
  4. When the instance is up, check security groups, remote login, firewall rules and disk attachment.
  5. If the region or size turns out wrong, resize and migration are both self-service and the bill follows.

The three mistakes almost every first-timer makes

1. Ticking the public IP and default bandwidth out of habit — a staging box that never stops quietly accruing egress charges. 2. Oversized system disks — 100GB of SSD for a machine that only runs Nginx costs more than the compute. 3. One config and one billing mode for dev and prod — on-demand in production is the most common cause of a surprising invoice. Set auto-shutdown on test instances, or destroy them.

Open a Tencent Cloud account with us

We handle official provisioning and payment verification — no personal card, no identity documents. Order directly in the vendor console once live.

Top up now