Cloud Cost Optimisation for Indian Startups: Cut Your AWS Bill

Open your AWS billing console on the first of the month and you will probably feel a small jolt. The number is bigger than last month, again, and nobody on the team can quite explain why. This is the most common pain we hear from founders across Delhi, Bengaluru, and Pune: the cloud bill keeps creeping up while revenue stays flat. The good news is that most of that spend is recoverable without touching your architecture, and this guide on cloud cost optimisation India walks through exactly where the money leaks and how to plug it.

We will go line by line, the way you would actually read an invoice, and give you steps you can act on this week.

Why does my AWS or GCP bill keep going up?

Cloud bills grow for boring reasons, not glamorous ones. It is rarely a viral traffic spike. It is usually accumulated neglect: instances nobody turned off, storage nobody cleaned, and pricing models nobody revisited after the free credits ran out.

Here are the usual suspects we find when we audit an Indian startup's account:

• Oversized compute. A founder picked a large instance "to be safe" during a demo, and it has been running at 8% CPU ever since.
• Idle resources. Dev and staging environments left running 24x7 when the team only works 10 hours a day.
• Forgotten storage. Old EBS volumes, unattached IPs, snapshots from 2023, and S3 buckets full of logs nobody reads.
• Data transfer charges. The silent killer. Moving data between regions or out to the internet costs real money, and it never shows up where you expect.
• On-demand everything. Paying full retail price for workloads you run every single day.

Notice that none of these require re-architecting. They require attention and discipline. Let us turn each into action.

How do I right-size my cloud instances without breaking anything?

Right-sizing means matching the resource to the actual workload. Most startups over-provision because their first instance was a guess, and guesses tend to round up.

Step one: look at the real numbers

Open CloudWatch (AWS), Cloud Monitoring (GCP), or Azure Monitor and pull two weeks of metrics for every running instance. Look at average and peak CPU, memory, and network. If a machine sits below 20% CPU with no real peaks, it is a candidate to shrink one or two tiers.

AWS even gives you free advice through Compute Optimizer and the Cost Explorer "rightsizing recommendations" panel. GCP has "recommendations" in the console, and Azure has Advisor. Switch these on first.

Step two: shrink in small steps

Do not jump from a large to a small in one move. Drop one tier, watch for a few days, then drop again if metrics stay healthy. Move during low-traffic windows, and keep a snapshot so you can roll back in minutes.

Step three: switch to current-generation and ARM instances

Newer instance families almost always give more performance per rupee. On AWS, moving from an older Intel family to Graviton (ARM) instances can cut compute cost by roughly 20% for workloads that support ARM, which most modern Node, Python, Java, and Go apps do. Test in staging first.

What is the fastest way to cut my cloud bill this week?

If you have a board meeting on Friday and need a visible win, do these in order. They are low risk and quick.

1. Schedule non-production environments. Shut down dev and staging outside working hours. A simple Lambda or instance schedule running them only 12 hours a day on weekdays cuts that spend by around 65%.
2. Delete orphaned resources. Hunt for unattached EBS volumes, old snapshots, idle load balancers, and Elastic IPs you are paying for but not using.
3. Set S3 lifecycle rules. Move logs and backups older than 30 days to infrequent-access storage, and older than 90 days to Glacier. This is a five-minute config change.
4. Kill zombie instances. Cross-check every running instance against a person who can explain why it exists. The unexplained ones usually can go.
5. Turn on a billing alert. Set a budget threshold so you get an email before the bill surprises you, not after.

We have seen these five steps alone trim 25 to 40% off a typical startup invoice, with zero impact on production.

Should Indian startups use Reserved Instances or Savings Plans?

Once you have right-sized and cleaned up, the next big lever is committed pricing. On-demand is the most expensive way to pay. If you know a workload will run for the next year, you are leaving money on the table by paying retail.

The simple decision rule

• Steady, predictable baseline? Cover it with a 1-year Savings Plan (AWS) or Committed Use Discount (GCP). Discounts often land in the 30 to 50% range.
• Unpredictable or experimental? Keep it on-demand.
• Fault-tolerant batch jobs, CI runners, or rendering? Use Spot instances and save up to 70 to 90%, accepting that they can be reclaimed.

A practical mix for most Indian SaaS startups: commit to roughly 60 to 70% of your steady baseline with a 1-year plan, run variable load on-demand, and push batch work to Spot. Start with a 1-year, no-upfront commitment so you keep cash flexibility, which matters when you are managing GST outflows and a tight runway.

One caution: do not commit until your usage has been stable for at least a month after right-sizing. Locking in a commitment for an oversized fleet just locks in waste.

Where do hidden data transfer and managed-service costs hide?

This is the part of the bill most teams never read closely. A few things to check:

• Cross-AZ and cross-region traffic. Chatty microservices talking across availability zones rack up transfer fees. Keep tightly coupled services in the same zone where it makes sense. If you are still deciding on architecture, our take on monolith versus microservices for an Indian startup is worth a read before you spread services everywhere.
• NAT Gateway charges. NAT gateways bill per hour and per GB. High-traffic apps can run up surprising NAT bills. Consider VPC endpoints for traffic to S3 and DynamoDB so it never touches the NAT.
• Overprovisioned managed databases. RDS and Cloud SQL instances are often sized for a launch that hasn't scaled yet. Right-size these too, and consider whether your engine choice fits your load. Our Postgres vs MySQL vs MariaDB decision guide covers the tradeoffs.
• CloudFront and CDN egress. Serving large assets to users in India? A CDN with India edge locations often costs less than serving directly and is faster too.

Do I need a third-party tool, or can I do this in-house?

For a startup under roughly ₹5 lakh a month in cloud spend, the native tools are enough: Cost Explorer, billing alerts, and the built-in optimisation recommendations. You do not need to pay for a fancy FinOps platform yet.

What you do need is ownership. Pick one person, usually the IT lead or a senior engineer, to review the bill on the first working day of every month against a simple checklist:

• Did any new untagged resource appear?
• Are dev and staging still on a schedule?
• Any instance below 20% utilisation for two weeks?
• Are commitments covering the right baseline?
• Any spike in data transfer line items?

Tagging matters here. Tag every resource by environment, team, and project from day one. Without tags you cannot tell whether a cost belongs to a paying customer or a forgotten experiment.

If your team is stretched thin, this is exactly the kind of work eDarpan's cloud migration and managed services team handles for startups in Dwarka and across NCR. We can run a one-time audit or manage the account ongoing so your engineers stay focused on the product. A short IT consulting engagement often pays for itself in the first month's savings.

Putting it together: a cloud cost optimisation India playbook

Cloud cost optimisation India done well is not a one-off cleanup. It is a habit. Here is the order we recommend for a sustainable result:

1. Week 1: Quick wins. Schedule non-prod, delete orphans, set lifecycle rules and billing alerts.
2. Week 2: Right-size compute and managed databases using real metrics.
3. Week 3: Review data transfer, NAT, and CDN line items.
4. Week 4: Once usage is stable, buy Savings Plans or committed-use discounts for your baseline.
5. Ongoing: Monthly review by one accountable owner, with tagging enforced.

Before you scale, it is also worth running through a production readiness checklist so reliability and cost stay in balance. And if you are still choosing a platform, our practical guide to picking a cloud provider in India compares the options for 2026.

Frequently asked questions

How much can a startup realistically save on its cloud bill?

Most startups that have never optimised can cut 25 to 40% in the first month through right-sizing, scheduling, and cleanup alone. Adding committed pricing on top can push total savings past 50% for steady workloads. The exact figure depends on how much waste exists today.

Will reducing cloud costs slow down my application?

Not if you do it with data. Right-sizing based on real CPU and memory metrics keeps headroom for peaks. The mistake is cutting blindly. Always test changes in staging, move in small steps, and keep snapshots so you can roll back fast.

Do these tips apply to AWS, GCP, and Azure equally?

Yes. The names differ, Savings Plans on AWS, Committed Use Discounts on GCP, Reservations on Azure, but the principles are identical: right-size, clean up, commit to the baseline, and watch data transfer. Every major provider has native cost tools you should enable.

We are an MSME on a tight runway. Where should we start?

Start with the free wins in Week 1 of the playbook above. They cost nothing and need no commitment, which protects your cash. Only buy reserved capacity once your usage is stable. If you want a second pair of eyes, eDarpan offers a one-time cloud cost audit for Indian businesses.

Can eDarpan manage our cloud account so we do not have to?

Yes. Beyond cost work, we help with custom software development, mobile apps, and licensing for Google Workspace and Microsoft 365. You can contact our team to discuss an audit or ongoing management. Have a look at our full services overview to see how we support startups end to end.

Cutting your cloud bill is one of the few engineering tasks where the payoff is immediate and measurable. Spend a focused week on it, assign one owner, and treat cloud cost optimisation India as a monthly discipline rather than a fire drill. Your runway will thank you, and so will your next board meeting.

Image credit: Sad Cartoon versus Technology by Sean Loyless via flickr (BY 2.0), sourced through Openverse.