Cloud Market Hits $2T: What Indian SMBs Should Check Before Picking AWS, GCP or Azure

Your cloud bill last month probably looked fine. Your cloud bill next month, after your developer spun up a staging environment in Singapore, forgot to shut it down, and enabled a managed database with cross-region replication "just in case," probably won't. This is the pattern we see most often with Indian SMBs who move to the cloud: the migration itself goes smoothly, and then the monthly invoice quietly becomes the third-largest line item in the business, behind salaries and rent, with nobody able to explain why.

The global public cloud market has crossed roughly $2 trillion in annual spend according to recent analyst estimates, and that number gets quoted everywhere. It tells you the market is enormous. It tells you nothing about whether a 25-person textile exporter in Tiruppur should run its ERP on AWS Mumbai, Google Cloud Delhi, or a single Azure VM under a desk. Those decisions depend on local pricing, rupee exposure, data residency rules, and how much your team can realistically operate.

This post walks through the checklist we use with Indian SMB clients before they sign anything. You will get a worked migration example with real rupee figures, a comparison of the three hyperscalers on the criteria that matter at your size, the Indian compliance points that tend to get missed, and the hidden cost categories that turn a ₹18,000 estimate into a ₹52,000 invoice. If you are a founder or the person who ends up owning IT by default, this is meant to be something you can hand to your vendor or work through yourself.

Key Takeaways

• Choose your cloud region first, then your provider. Latency to your customers and data residency obligations narrow the field before pricing does.
• Price in egress, support plans, backups, and idle resources from day one. These typically add 25 to 60 percent to a raw compute estimate.
• Model your bill in rupees with a realistic exchange rate band, since most hyperscaler pricing is denominated in dollars.
• Check which of your data (especially personal data and payment-related records) must stay in India, and document where each workload actually runs.
• Budget for one person's time to operate the environment. A cloud account without an owner becomes an expensive liability within six months.
• Run a 60-day pilot with tagging, budgets, and alerts switched on before committing to a reserved-capacity contract.

What does cloud computing for Indian SMBs actually cost once you move off on-premise servers?

The honest answer is that it depends heavily on what you are moving, but the migration economics are more favourable than most founders expect when the workload is steady and the team is small. The cost you will underestimate is not the server. It is the surrounding services and the people time.

Consider a worked example. A 15-person logistics company in Gurgaon was running a file server, a small SQL database for its dispatch and billing system, a Windows application server, and a backup NAS in a rack in its office. Their annual hardware refresh was coming due, and the local vendor had quoted roughly ₹4.2 lakh for replacement servers plus three years of warranty. Their monthly cost of running the existing setup, including power at commercial rates, a managed-services contract for patching, the internet line dedicated to the rack, and a part-time admin, came to about ₹45,000 per month.

They moved to AWS in the Mumbai region (ap-south-1). The target architecture was deliberately modest: two small EC2 instances for the application and file tier, a managed RDS database instance, S3 for backups with lifecycle rules moving older snapshots to cheaper storage classes, and a site-to-site VPN back to the office for the handful of legacy desktop clients that still needed local network access. They bought a one-year compute savings plan after two months of observed usage, not on day one.

The post-migration steady state settled at roughly ₹18,000 to ₹19,000 per month in cloud charges, which included the AWS Business support plan at around 10 percent of monthly spend. Their part-time admin cost did not disappear, but the managed-services contract was dropped, and the hardware refresh was avoided entirely. Net, the monthly run-rate fell by more than half.

Notice what made that work. The workload was predictable, the data volume was modest, and the company had someone who could own the environment. If any of those three things were absent, the numbers would look different.

The hidden cost categories that inflate the first invoice

Most pricing calculators show compute, storage, and a rough data transfer figure. They underweight the following, and in our experience these are where the surprises come from:

• Data egress. Moving data out of the cloud to your customers, your office, or another provider is billed. A reporting job that pulls 400 GB a month to an on-prem analytics tool is a real line item, often ₹3,000 to ₹8,000 monthly before anyone notices.
• Support plans. AWS, Google Cloud, and Azure all offer tiered support. Basic support is free but gives you no meaningful response times. Paid tiers are typically a percentage of spend with a minimum, and that minimum matters for small accounts.
• Backups and snapshots. Retention policies quietly accumulate. Daily snapshots kept for 90 days on large volumes can match the cost of the primary storage.
• Idle and forgotten resources. Test environments, unattached disks, old load balancers, and public IP addresses (which now carry a charge) accumulate without an owner.
• Cross-region and cross-zone traffic. Architectures that span regions for resilience are excellent engineering and expensive operations. Be deliberate about which workloads actually need it.
• Licensing. Windows Server and SQL Server licences on cloud VMs often cost more than the compute. Bring-your-own-licence options exist but have eligibility conditions that vary by provider and version.

Pro Tip: Before you sign, ask your provider or partner for the last 90 days of usage in a comparable account, or run a two-week pilot and export the cost-and-usage report. Look specifically at the line items labelled "data transfer out," "support," and "NAT gateway." The NAT gateway charge alone, which is billed hourly plus per gigabyte processed, has caught more Indian SMB teams than any other single service.

How should Indian SMBs compare AWS, GCP and Azure on the criteria that matter at their size?

The hyperscalers are functionally similar for the workloads most SMBs run. The differences that matter tend to be pricing structure, the quality of Indian regional presence, licensing alignment with what you already own, and how much hand-holding your team needs. Here is how we typically frame it.

Criterion	AWS	Google Cloud (GCP)	Microsoft Azure	Practical note for SMBs
Indian regions	Mumbai and Hyderabad	Mumbai and Delhi	Central India (Pune), South India (Chennai), West India (Mumbai)	All three offer in-country hosting. Confirm your specific services are available in the region you pick, since not every managed service launches everywhere at once.
Ecosystem fit	Broadest third-party marketplace and largest partner pool in India	Strong for data and analytics workloads, Workspace integration	Strongest if you already run Microsoft 365, Active Directory, or Windows Server	If your staff live in Outlook and Teams, Azure often reduces friction and licence complexity.
Pricing model	Granular, many small line items, savings plans and reserved instances	Sustained-use discounts applied automatically, simpler on-demand pricing	Reserved capacity and hybrid benefit for existing Windows licences	GCP's automatic sustained-use discounts can suit steady workloads without a commitment decision.
Billing in INR	Available via Indian reseller or partner billing	Available via Indian reseller or partner billing	Available via Indian reseller or partner billing	Direct USD billing exposes you to exchange rate movement. Partner billing in rupees with GST invoicing is usually the cleaner option for an Indian entity.
Learning curve for a small team	Steeper, the most options to get wrong	Moderate, opinionated defaults help	Moderate, especially if the team knows Microsoft tooling	Your existing skills usually outweigh raw feature differences at SMB scale.

A pattern we keep seeing: the provider that is cheapest on paper is rarely cheapest in practice, because the difference is made up in engineering hours. A team that already knows Windows administration will be faster and cheaper on Azure even if a GCP calculator shows a lower figure. Conversely, a product team building on containers and managed databases may find GCP or AWS less friction.

If you are still deciding, our broader guide to choosing a cloud provider for Indian SMBs goes deeper on the decision logic, and the pricing and hidden-cost breakdown for AWS, Azure and GCP covers the line-item detail behind the table above.

What are the data localisation and compliance points Indian businesses must check?

Data localisation in India is not one law. It is a patchwork of sector rules and proposed frameworks, and the safe approach is to document where each category of data lives rather than assume a single rule covers everything. Your obligations depend on your sector, your customers, and what kind of personal data you process.

Start with the Digital Personal Data Protection Act, 2023, and its rules. The Act sets out obligations for data fiduciaries processing personal data of Indian residents, including consent, purpose limitation, breach notification, and the rights of data principals. The Act permits cross-border transfer to countries not restricted by the government, but sector regulators can impose stricter localisation requirements, and the rules that operationalise the Act are being phased in, so track the timelines on the official gazette rather than relying on secondhand summaries.

Other points that commonly catch SMBs:

• RBI payment data. The RBI's 2018 directive requires that payment system data be stored only in India. If you are a payment aggregator, a fintech, or process card and UPI-related data as part of your product, this applies directly.
• CERT-In incident reporting. CERT-In's directions require certain cyber incidents to be reported within six hours of noticing them, and they require logs to be maintained within India for a defined period. Your cloud logging configuration must reflect this, including where CloudTrail, Cloud Audit Logs, or Azure Monitor logs are stored and for how long.
• GST records. Your GST invoices and books must be accessible for audit and retained for the statutory period. This is not strictly a localisation rule, but if your accounting system is in a foreign region, make sure access and retention are still compliant.
• Customer contracts. Large enterprise and government clients often specify in-country hosting in the contract, sometimes with specific audit rights. Read the data clauses before you assume your region choice is enough.

Common Mistake: Teams often choose an Indian region for the primary database and forget the backups, logs, analytics exports, and staging copies. We have reviewed accounts where production data sat in Mumbai but nightly backups replicated to a Singapore bucket created during a disaster-recovery experiment two years earlier. Map every place data is copied, not just where it is born.

For a fuller treatment of the cost side of sovereign and in-country hosting, see our practical decision framework for sovereign cloud costs. If you need help producing a data-flow map or reviewing your current setup against these obligations, our IT consulting team does this as a fixed-scope engagement.

How do you run a cloud migration step by step without getting locked into a bad contract?

The biggest mistake we see is treating migration as a single project with a go-live date. The teams that do it well treat it as a sequence of small, reversible decisions. Here is the walkthrough we use.

1. Inventory what you actually run. List every server, database, scheduled job, SaaS integration, and person who touches them. Include the things nobody remembers, like the printer driver server and the cron job on an old desktop. Most inventories miss about a fifth of the real estate.
2. Classify by data sensitivity and residency. Tag each workload as public, internal, personal, or regulated. This drives your region and encryption choices, and it tells you which workloads need the strictest review.
3. Pick one low-risk workload for a pilot. A file share, an internal intranet, or a reporting database is ideal. Avoid starting with the billing system.
4. Set up the account governance before the first resource. Create separate accounts or projects for production, staging, and sandbox. Turn on multi-factor authentication for every human user. Apply a tagging standard (at minimum: owner, environment, cost centre). Set budget alerts at 50, 80 and 100 percent of your expected monthly spend.
5. Build the target architecture in code. Use Terraform, CloudFormation, Bicep, or the provider's native equivalent. Infrastructure you cannot rebuild from a repository is infrastructure you cannot safely change.
6. Migrate and run in parallel for at least two weeks. Keep the old system authoritative and sync data to the new one. Test restores, not just backups.
7. Cut over with a rollback plan written down. Decide in advance what conditions trigger rollback, who makes that call, and how long reverting takes.
8. Measure for 60 days before committing. Watch the real bill, not the forecast. Only then consider savings plans or reserved capacity, which typically offer meaningful discounts in exchange for a one- or three-year commitment.
9. Write an exit plan. Document how you would move data out if you changed provider. Export formats, database dump procedures, and DNS ownership should all be recorded. This is your protection against lock-in and your leverage in renewal negotiations.

On contracts specifically: negotiate for clarity on INR billing and GST invoicing if you are buying through a partner, get support response times in writing, and confirm what happens to your data and access if the reseller relationship ends. Many small businesses discover their reseller has changed hands only when invoices start arriving from an unfamiliar name.

If your migration includes rebuilding applications rather than lifting them as-is, our team handles that under cloud migration and managed services, and where the application itself needs changes we often work alongside the custom software development side of the engagement.

What should a budget for cloud computing look like for a 20-person Indian company?

A practical rule that has held up across our client base: budget the cloud bill at roughly 1.3 to 1.5 times the raw infrastructure estimate in the first year, and expect it to settle lower once you have tuned the environment. The multiplier covers support, backups, egress, and the inevitable first-quarter learning. Add a separate line for someone's time, because the cloud does not run itself.

For a 20-person company with a typical mix of an accounting or ERP system, a file and collaboration layer, a customer-facing website, and some internal reporting, a sensible first-year envelope often looks like this in rupee terms:

• Compute and managed databases: ₹9,000 to ₹16,000 per month, depending on workload steadiness
• Storage and backups: ₹2,000 to ₹5,000 per month
• Data transfer and networking: ₹1,500 to ₹5,000 per month
• Support plan: 3 to 10 percent of spend, with a sensible minimum
• Security tooling and monitoring: ₹2,000 to ₹6,000 per month
• Internal or partner operational time: budget this separately, usually a fraction of an engineer's month

These ranges are illustrative and will move with your architecture and your region. The point is to have a line for each category so that when the bill surprises you, you can see which category moved. Keep the exchange rate in the model as a band rather than a single number, and revisit it quarterly if you are paying in dollars.

Where eDarpan fits into this is mostly in the parts that are easy to underbudget: the initial architecture review, the governance setup, and the ongoing managed operation for teams who do not want to hire a cloud engineer yet. If your team also uses Microsoft or Google productivity tools, our Microsoft 365 licensing and Google Workspace licensing services can be bundled into the same planning exercise so you are not optimising the server bill while overpaying for seats.

Is a cloud computing setup worth it for a business that is still small?

Often yes, but not for the reason vendors usually give. The cloud is not automatically cheaper. It is more flexible. You can scale down when a seasonal business quietens, you can add a region when a new client demands it, and you can avoid a capital expenditure cycle that ties up cash you may need elsewhere. Those benefits are real, but they only pay off if someone is paying attention.

If your workload is small, your data is not sensitive, and nobody on your team has the time to own an environment, a well-managed SaaS stack may serve you better than infrastructure you have to operate. Our comparison of Zoho, Google Workspace and Microsoft 365 for Indian SMBs covers when a SaaS-first approach makes more sense.

If you are building a product, handling payment or health-adjacent data, or expect to grow quickly, the cloud usually earns its place early. The question then shifts from whether to move, to how to move without creating a new kind of mess.

Frequently asked questions

Which cloud provider is best for a small business in India?

There is no single best provider. Choose based on where your team already has skills, whether you use Microsoft or Google productivity tools, and which Indian regions support the services you need. For most small teams, the deciding factor is the one your people can operate confidently rather than the one with the lowest headline price.

How much does cloud computing cost for a small business in India per month?

For a typical small business running an accounting system, file sharing, a website and some reporting, costs often fall between ₹15,000 and ₹35,000 per month once the environment is tuned. Heavier workloads or regulated data can push this higher. Always add support, backups, and data transfer to the raw compute estimate before comparing with on-premise costs.

Do Indian businesses have to store all their data in India?

Not necessarily all of it, but some categories have localisation requirements. RBI payment data must be stored in India, and sector regulators or client contracts may impose further restrictions. Map your data by category and check the applicable rules rather than assuming one answer covers everything.

Is AWS Mumbai region enough for Indian data residency?

For many businesses, hosting primary workloads in an Indian region is the core requirement. However, backups, logs, replicas and analytics exports can also hold regulated data, so you must check where every copy lives. The region alone does not settle your compliance position.

What is the biggest hidden cost in cloud computing?

For most small teams, it is a combination of data transfer out and idle or forgotten resources. Egress charges are easy to overlook in calculators, and unattached storage, old snapshots and test environments accumulate silently. Tagging every resource with an owner and reviewing monthly spend by tag catches most of these.

Should I pay for a reserved instance or savings plan on day one?

Usually not. Run the workload for at least 60 days, observe the real usage pattern, and then commit to savings plans or reserved capacity for the steady portion. Committing too early locks in a size you may not need, and the discount is only valuable if the capacity is actually used.

Can a non-technical founder manage a cloud environment?

A founder can set the policy and approve the budget, but someone with operational skills needs to own the environment. Many small companies use a managed services partner for this, which can be cheaper than a full-time hire at their current scale. Whatever you choose, make sure there is a named owner, documented access, and a tested recovery plan.

Where to go from here

Cloud computing for Indian SMBs is not a single decision. It is a series of choices about region, provider, governance, budget and ownership, and each one is easier to get right when made deliberately. Start with the inventory and the data map. Pick a pilot. Measure the real bill before you commit. Keep an exit plan in a drawer you can open.

If you want a second opinion on your current setup or a fixed-scope review before you sign a contract, eDarpan works with Indian SMBs on exactly this kind of planning. You can explore our full services overview, learn more about eDarpan and how we work, or get in touch with our team to talk through your workloads.