Cloud compute and storage pricing: vCPU, RAM, disk tiers, IOPS and object storage explained

Compute and storage are the two lines you see first on any hosting bill. Compute is priced by the size of the server (vCPU and RAM) and its family or generation, per month on a VPS and per second or hour on public cloud. Storage is priced per GB per month, and the price climbs with performance: fast SSD with guaranteed IOPS costs more than standard SSD, which costs more than object storage, which costs more than archive. Getting the size and tier right is the single largest saving most small businesses can make. This guide explains how both are priced, with illustrative figures, as part of the hosting pricing overview.

How compute is priced

vCPU and RAM

A vCPU is usually one hardware thread of a physical core, not a whole core. Providers sell fixed combinations of vCPU and RAM, grouped into families:

Family (generic name)	vCPU : RAM ratio	Typical use
General purpose	about 1 : 4 GB	Web servers, small databases, business apps
Compute optimised	about 1 : 2 GB	CPU-heavy processing, build servers
Memory optimised	about 1 : 8 GB or more	Larger databases, caches, analytics
Burstable	Low baseline CPU with credits	Low-traffic sites, dev and test

On public cloud the price roughly doubles each time you double the size within a family. Newer generations often offer more performance for a similar price, and processors based on Arm designs are often priced below comparable x86 instances; check that your software runs on them.

Burstable instances

Burstable servers give you a baseline share of CPU and let you "burst" above it while you have credits. They are cheap for websites that are idle most of the day. If CPU stays high for long periods, you either run out of credits and slow down, or (on some plans) pay for extra CPU time.

VPS plans

A VPS plan bundles vCPU, RAM, disk and a bandwidth allowance at one monthly price. Two plans with the same numbers can perform differently, because hosts allocate host CPU differently. Test before committing to a long prepaid term.

Right-sizing: the cheapest change you can make

Most small servers are over-provisioned. A simple rule: look at CPU and memory use over two to four weeks, including your busiest days, and size for the peak plus a margin rather than for a guess.

Illustrative example. A cloud VM with 4 vCPU / 16 GB costs an illustrative ₹12,000 a month on demand. Monitoring shows peak CPU of 25% and memory of 6 GB. Moving to 2 vCPU / 8 GB at an illustrative ₹6,000 halves the compute line: a saving of ₹6,000 a month, or ₹72,000 a year before GST (₹84,960 including 18% GST).

How storage is priced

Block storage (the server's disks)

Tier (generic)	How it is priced	Fits
Standard HDD	Lowest per GB; low IOPS	Archives, cold logs on a server
Standard SSD	Per GB; baseline IOPS tied to size	Boot disks, most web apps
Provisioned IOPS SSD	Per GB plus per provisioned IOPS or throughput	Busy databases
Local NVMe	Included with some instance types; lost if the VM stops	Temporary data, caches

IOPS (input/output operations per second) and throughput decide how fast a disk responds. On some clouds performance grows with disk size, so teams buy larger disks just for speed; on others you pay for IOPS separately. Read the price page for your region carefully.

Illustrative example. A database needs 200 GB. On standard SSD at an illustrative ₹8 per GB-month: ₹1,600. On provisioned IOPS SSD at an illustrative ₹12 per GB-month plus ₹2,000 for extra IOPS: ₹4,400. If the database only needs baseline performance, standard SSD saves ₹2,800 a month.

Object storage

Object storage holds files (images, backups, exports) and is priced per GB-month, plus requests (per thousand reads or writes) and data transfer out. It is usually much cheaper per GB than block storage.

Archive tiers

Archive storage is the cheapest per GB but charges for retrieval and may take hours to return data. Many tiers also have a minimum storage duration; deleting early still bills the minimum. Use it for backups you hope never to read.

Storage class (generic)	Price per GB	Retrieval	Minimum duration
Hot / standard	Highest of the three	Free or low	Usually none
Infrequent access	Lower	Per-GB retrieval fee	Often 30 days
Archive	Lowest	Higher fee, slow	Often 90–180 days

Disk costs that keep running

Unattached disks. Deleting a VM does not always delete its disks. They keep billing.
Over-sized boot disks. A 100 GB boot disk on every server adds up across ten servers.
Snapshots. Each snapshot is billed by stored data. See backups and DR cost.
Stopped VMs. A stopped cloud VM usually stops compute charges but disks and reserved IPs still bill.

Sizing checklist before you request quotes

Record peak CPU, memory and disk use over a busy period.
Pick a family that matches the ratio you need.
Choose the cheapest disk tier that meets measured IOPS.
Move files and backups to object storage, and old backups to archive.
Plan a review after the first month of real use.

Then price the result with the server cost estimator and add the network and support extras and data transfer.

FAQ

Is 1 vCPU the same as 1 core?

Usually not. On most clouds and VPS platforms a vCPU is one thread, and a physical core has two threads. Software licensed per core may count differently; check the licence terms.

Do I pay for a cloud VM when it is stopped?

Usually not for compute, but disks, snapshots and reserved IPs continue to bill.

Why is the same instance cheaper in another region?

Prices are set per region. A cheaper region may add latency for Indian users and change data rules. See Indian regions and data rules.

How much disk headroom should I keep?

Enough for several months of growth plus space for updates and logs. On cloud, disks can usually be enlarged later, so start smaller.

Sources

AWS EC2 on-demand pricing — example of a per-region price page structure
Google Cloud regions and zones — where machine types are offered
CBIC GST — GST on IT infrastructure services
CERT-In — log retention that affects storage needs