Egress Fees Explained: The Hidden Cost of Leaving Your Cloud

Here's a scenario I've watched play out more than once. A mid-sized fintech in Bengaluru builds their entire platform on AWS over three years. Everything runs beautifully. Then their board pushes for a multi-cloud strategy, or a better deal on GCP comes along, and someone finally sits down to calculate what it costs to move roughly 40 TB of accumulated data out. The number comes back at nearly ₹6.5 lakh in egress charges alone, before they've moved a single application. The room goes quiet.

That silence is the sound of a lesson learned too late. Cloud providers make it almost free to bring data in and surprisingly expensive to take it out. This asymmetry is by design, and it's the single most underestimated line item in an Indian SMB's cloud budget. In this post I'll break down how cloud egress fees India businesses face actually work across AWS, Azure and GCP, show you a real cost breakdown, and give you a practical checklist to avoid getting locked in or blindsided by an exit bill.

I've helped companies from 15-person logistics firms in Gurgaon to 200-seat SaaS teams in Pune plan cloud moves. The pattern is always the same: nobody reads the egress pricing until it hurts. Let's fix that.

Key Takeaways

• Egress (data transfer out) is billed per GB and can quietly become 10-20% of your total cloud bill once you serve real traffic.
• Ingress (data coming in) is almost always free. This is the trap that makes migration feel cheap and exit expensive.
• AWS, Azure and GCP all offer the first 100 GB/month free, then charge roughly ₹7-10 per GB for internet egress from Indian regions.
• Inter-region and inter-AZ transfers cost money too, and they hide inside "normal" architecture nobody flagged.
• The EU and India both now have "free egress on exit" rules brewing, but don't architect around promises that aren't in your contract.
• You can cut egress 40-70% with CDN caching, VPC endpoints, data compression, and picking the right region up front.

What exactly are cloud egress fees and why do they exist?

Egress simply means data leaving the cloud provider's network. Every time a user downloads a file from your app, an API returns a response to a mobile client, or you copy a database backup to another provider, that outbound traffic gets metered per gigabyte.

Ingress, the data flowing in, is free on all three major providers. That asymmetry isn't accidental. It lowers the barrier to move onto the platform and raises the cost of leaving. Once your data, your applications and your team's muscle memory live inside one ecosystem, the switching cost becomes a moat. Regulators in the EU flagged this under the Data Act, and Indian discussions around cloud portability under the digital competition framework are heading the same direction. But that's future policy. Your bill is present-day reality.

There are actually several flavours of "egress" you'll be charged for:

• Internet egress — data going out to the public internet (your users). The big one.
• Inter-region transfer — moving data between, say, Mumbai and Singapore regions.
• Inter-availability-zone transfer — traffic between zones within the same region, which surprises people constantly.
• Cross-cloud egress — the exit bill when you leave for a competitor.

Most teams budget for compute and storage. Almost nobody models data transfer properly. That's the gap I want to close here, and it's exactly the kind of thing we work through during a cloud IT consulting engagement.

How much do egress fees actually cost on AWS, Azure and GCP in India?

Pricing shifts, so always confirm against the provider's live calculator, but the ranges below reflect Indian region pricing (Mumbai for AWS/Azure, Delhi/Mumbai for GCP) as of the current cycle. All three give you the first 100 GB per month of internet egress free, then bill in tiered slabs. Prices below are converted to rupees and are indicative, before GST at 18%.

Criteria	AWS (Mumbai)	Azure (Central India)	GCP (Mumbai/Delhi)
Free tier (internet egress/month)	100 GB	100 GB	100 GB
Internet egress, first paid TB (approx ₹/GB)	₹8.9	₹7.2	₹9.8
Inter-region transfer (₹/GB, approx)	₹1.7	₹1.5	₹1.7
Inter-AZ within region	Charged (~₹0.85/GB)	Often free intra-region	Charged (~₹0.85/GB)
CDN egress (via CloudFront/Front Door/Cloud CDN)	Lower, cache-dependent	Lower, cache-dependent	Lower, cache-dependent
Exit/migration data transfer	Free if fully leaving (with request)	Free if fully leaving (with request)	Free if fully leaving (with request)

That last row matters. Following the EU Data Act pressure, all three now offer free egress when you're completely leaving the provider, but you have to formally request the credit, the process takes weeks, and it applies only to a full exit, not a partial multi-cloud split. Do not assume it's automatic.

To put the per-GB number in perspective: a media-heavy app serving 5 TB of downloads a month past the free tier will run you roughly ₹40,000-49,000 in AWS internet egress alone, monthly, before GST. That's a full-time junior developer's salary going to data transfer nobody costed.

A real worked example: a Gurgaon logistics firm's migration math

Let me walk through a case that's representative of what I see. A 15-person logistics and last-mile delivery company in Gurgaon was running two on-prem servers plus a rented rack, spending about ₹45,000/month on hardware amortisation, power, AMC and a part-time sysadmin. They wanted to move to AWS to support a new driver-tracking mobile app and real-time dashboards.

The naive migration plan looked cheap. Compute and storage came to around ₹18,000/month on right-sized instances. Everyone celebrated. Then we modelled the data flows.

Their driver app pushed GPS pings and photos into the cloud (ingress, free) but their operations dashboard, customer tracking links, and API integrations with partner marketplaces pushed a lot of data out. We estimated 3.5 TB/month of internet egress once fully live. That's roughly ₹30,000/month in transfer fees on top of the ₹18,000 compute. Suddenly the "₹18K cloud bill" was closer to ₹48,000, worse than on-prem.

Here's what we actually did to bring it back down:

1. Put CloudFront in front of everything — customer tracking pages and photos got cached at the edge, cutting origin egress by around 60% since the same delivery-tracking link gets opened many times.
2. Compressed API payloads — switching partner integrations from verbose JSON to gzipped responses roughly halved the bytes on the wire.
3. Kept internal traffic inside one AZ where possible — the app servers and database were pinned to the same availability zone to avoid inter-AZ charges on chatty queries.
4. Used S3 lifecycle rules to push old delivery photos to cheaper storage and stop them being repeatedly pulled out.

Final steady-state bill: about ₹26,000/month including GST, comfortably below their old ₹45,000 on-prem cost, with far better reliability and a proper mobile app backend feeding real-time data. The lesson: egress isn't a reason to avoid cloud. It's a reason to architect for it. This kind of modelling is core to our cloud migration and managed services work.

Common Mistake: Teams model egress based on their current traffic, then get shocked when a successful launch triples downloads. Always model egress at 2-3x your projected steady-state, because success is the most expensive scenario. A viral marketing campaign that 10x's your traffic will 10x your egress bill on the same day.

Where do hidden egress charges sneak into your architecture?

Internet egress is the obvious one. The sneaky charges are the internal ones that look like normal engineering decisions.

Inter-availability-zone chatter

On AWS and GCP, sending data between availability zones in the same region is billed both ways. If your database is in AZ-A and your busy application servers are spread across AZ-B and AZ-C for high availability, every query and result crosses a billed boundary. A chatty microservices setup can rack up tens of thousands of rupees a month purely on intra-region transfer. The fix is careful placement and read-replicas positioned close to the workloads that hammer them.

NAT gateway data processing

On AWS, traffic flowing through a NAT gateway is charged per GB processed, on top of egress. Private instances pulling large OS updates or Docker images through NAT can generate a bill nobody expects. Using VPC endpoints for S3 and other AWS services bypasses NAT entirely and is often the single highest-ROI change we make.

Cross-region backups and replication

Replicating a database to a DR region in Singapore for resilience is smart. But every byte of that replication is inter-region egress. Budget it explicitly, and consider whether daily snapshots suffice instead of continuous replication for non-critical data.

Logging and monitoring pipelines

Shipping verbose logs to a third-party observability tool outside your cloud counts as internet egress. High-volume debug logging in production has quietly cost clients ₹15,000+/month. Sample your logs, don't fire-hose them.

How do you reduce cloud egress fees without crippling performance?

You don't need to accept egress as fixed. Here's the playbook I actually deploy, roughly in order of impact per rupee of effort.

1. Cache aggressively with a CDN. CloudFront, Azure Front Door or Cloud CDN serve repeated content from the edge at lower per-GB rates and reduce origin egress. For any app serving images, static assets or repeatable API responses, this is the biggest single win.
2. Compress everything. Enable gzip/Brotli on responses. Compress backups. Use efficient formats (WebP over PNG, Parquet over CSV). You pay per byte, so every byte you don't send is money saved.
3. Use private endpoints and VPC gateway endpoints. Keep service-to-service traffic on the provider's private backbone instead of routing through NAT or the public internet.
4. Co-locate chatty workloads. Put your database and the services that talk to it most in the same AZ. Reserve multi-AZ spread for genuine HA needs.
5. Pick the right region on day one. Serving mostly Indian users? Stay in Mumbai/Delhi regions so you're not paying inter-region hops to reach them. Region choice is nearly impossible to reverse cheaply later, thanks to egress.
6. Batch and schedule bulk transfers. If you must move large datasets, some providers offer discounted or committed-use data transfer tiers. Negotiate these into your enterprise agreement.

Pro Tip: If you're moving more than 50 TB, don't push it over the internet at all. AWS Snowball, Azure Data Box and Google's Transfer Appliance physically ship you a device, you load your data, and courier it back. For a one-time migration this can be dramatically cheaper and faster than saturating your ISP link for a week, and it sidesteps a chunk of transfer billing.

For deeper cost surgery beyond egress alone, our detailed breakdown on cloud cost optimisation for Indian startups covers reserved instances, spot fleets and right-sizing.

How do egress fees affect vendor lock-in and multi-cloud strategy?

This is where egress stops being a line item and becomes a strategic risk. The whole point of the free-in, pay-out model is to make leaving expensive. If your 40 TB of data costs ₹6+ lakh to extract, the finance team will veto the move on cost grounds alone, even when a competitor is technically better.

A few practical principles for staying free:

• Keep data portable. Store in open formats (Parquet, standard SQL dumps) rather than proprietary managed-service formats that only make sense inside one cloud.
• Document your data gravity. Know exactly how much data lives where, and re-estimate your exit cost annually. If exit cost is climbing faster than your business value, that's a lock-in warning light.
• Use the free-exit provision deliberately. If you genuinely commit to leaving a provider fully, file the free-egress request early. It's real, but it's paperwork-heavy and only covers complete departure.
• Be sceptical of "free" multi-cloud. Running active-active across two clouds means constant cross-cloud egress between them. That ongoing bill can dwarf any resilience benefit for an SMB.

If you're still choosing a provider, our comparison of AWS, Azure and GCP for Indian SMBs weighs these trade-offs including regional pricing and support quality.

What about compliance, GST and the paperwork side?

Two things Indian businesses forget. First, egress charges attract 18% GST like any other cloud service, and if your provider bills from an overseas entity you may deal with reverse-charge mechanism and input tax credit reconciliation. Loop your CA in before signing an enterprise agreement.

Second, data localisation. Sectors under RBI, SEBI or the DPDP Act framework may be required to keep certain data within India. That's a compliance win for egress too, because keeping data in Mumbai/Delhi regions serving Indian users avoids inter-region transfer. Compliance and cost occasionally pull in the same direction here.

If you're a young company that hasn't sorted its registered address for GST and vendor agreements yet, our virtual office address for GST and company registration gets that foundation in place cleanly.

Frequently asked questions about cloud egress fees

Is data transfer into the cloud free on AWS, Azure and GCP?

Yes. All three major providers offer free ingress, meaning data coming into their network costs nothing. This is deliberate, as it lowers the cost of migrating in while making egress out expensive. Budget for egress, not ingress.

How much is AWS egress cost per GB in India?

From the Mumbai region, internet egress runs roughly ₹8-9 per GB in the first paid tier after the free 100 GB/month, before 18% GST, with volume discounts at higher tiers. Inter-region and inter-AZ transfers are cheaper but still add up. Always verify against the live AWS pricing calculator as rates change.

Can I move data out of my cloud provider for free?

Only if you are fully leaving the provider. Following EU Data Act pressure, AWS, Azure and GCP now offer free egress credits for complete migrations away, but you must formally request it, it takes weeks to process, and partial or multi-cloud moves don't qualify.

Why is my cloud bill higher than the compute cost I planned for?

The usual culprits are egress and data transfer charges that weren't modelled, including internet egress to users, inter-availability-zone traffic, NAT gateway processing, and cross-region replication. These commonly add 10-25% on top of compute and storage. A cost audit will pinpoint exactly where the leakage is.

Does using a CDN reduce egress fees?

Yes, significantly for cacheable content. A CDN serves repeated requests from edge locations at lower per-GB rates and reduces the load pulled from your origin, cutting origin egress. For image-heavy or high-traffic apps this is usually the single biggest saving available.

Do egress fees attract GST in India?

Yes, cloud services including data transfer charges attract 18% GST. If billed from an overseas provider entity, reverse-charge mechanism and input tax credit handling may apply, so involve your chartered accountant when structuring the agreement.

Should an SMB go multi-cloud to avoid lock-in?

For most Indian SMBs, no. Active multi-cloud generates constant cross-cloud egress between providers, and the ongoing cost usually outweighs the lock-in benefit. Instead, keep your data in portable open formats and single-provider so you preserve the option to move cheaply later.

The bottom line on cloud egress fees for Indian businesses

Egress isn't a scam and it isn't a reason to fear the cloud. It's a predictable, modellable cost that punishes teams who ignore it and rewards teams who architect around it. The Gurgaon logistics firm didn't win by avoiding AWS. They won by understanding that cloud egress fees India businesses pay are shaped by architecture decisions made on day one, from region choice to CDN caching to where the database sits.

Model your data flows before you migrate. Assume success will multiply your egress. Keep your data portable so you're never held hostage by an exit bill. And re-audit your transfer costs every quarter, because they creep.

If you'd like a straight, no-jargon assessment of your current cloud spend or a migration plan that costs egress properly from the start, that's exactly what we do. Explore our full services, or get in touch with the eDarpan team and we'll help you avoid the surprise-bill silence I described at the top. You can also read up on the swadeshi productivity-suite math and, if custom infrastructure is on your roadmap, our custom software development team can build it cost-aware from the ground up.

Image credit: Vibrant Technologies by vibrant.com via flickr (BY 2.0), sourced through Openverse.