The email arrived on a Tuesday morning: "Your cloud bill for last month: $2.4 million."
The CFO's response was immediate: "That's 3x our budget. What the hell are we running?"
The answer? Nothing special. Just a standard data analytics workload that happened to cross availability zones. A lot.
Turns out, 80% of that bill—nearly $2 million—was data egress fees. Not compute. Not storage. Just the privilege of moving their own data between regions to serve global customers.
Welcome to 2025, where yet another of the cloud's secrets are finally out: The meter is JUST running on your compute. It's running on your data movement and storage. And the house always wins.
The Hotel California Pricing Model
Here's what nobody tells you when you're signing that enterprise cloud agreement: You can check your data in anytime you like, but it costs a fortune to leave.
The numbers are staggering. According to Wasabi's 2025 Global Cloud Storage Index, 56% of organizations are experiencing operational delays due to egress fees, and 62% of them paid overages. Not performance issues. Not technical problems. Just the fear of the bill that comes from accessing their own data.
It is not at all uncommon to hear stories of companies paying 10x more than budgeted, and it's not all big companies! I personally chatted with someone whose bill went from $200 to $3,500 monthly in eight months. They hadn't changed their architecture. Folks just started actually using the product. Every API call, every report download, every backup replication; the meter starts and it just keeps running.
The real kicker? It hurts even inside a SINGLE CLOUD. A Fortune 500 company discovered they were paying $220,000 per week—per WEEK in unplanned egress fees from cross-region replication they didn't even know was happening. No alerts. No warnings. Just a seven-figure surprise at the end of the quarter.
Why 37signals Is Deleting Their AWS Account
This isn't theoretical. 37signals just announced they're completely exiting AWS, projecting $10 million in savings over five years. Not from optimization. Not from reserved instances. From simply buying servers and running their own infrastructure.
And while not every organization will do something quite as extreme, nearly EVERY organization is at least evaluating what makes sens.
And not just the new tech elite! GEICO - the Oracle of Omaha's insurance company who famously said he never understood technology - spent a decade moving to the cloud, only to see their cloud costs balloon to over $300 million annually. Their solution? Building their own OpenStack-based private cloud and bringing workloads back home.
The pattern is everywhere. Node4's latest research shows 86% of enterprises plan to repatriate workloads. Not some. Not a few. Eighty-six percent.
But here's the trap: The cost of leaving often is just a trade off of rented datacenters for hired engineers because platforms are not designed to run workloads and data processing in distributed environments.
The Multi-Cloud Fantasy Meets Data Gravity Reality
"We'll go multi-cloud to avoid vendor lock-in!"
Sure. Let me know how that works when you realize moving 10TB between clouds costs more than storing it for a year.
The math is brutal:
- AWS to Azure transfer: $0.09 per GB after your generous 100GB free tier
- 10TB migration: $920 just to move it once
- Daily synchronization: $27,600 per month
- Disaster recovery test: $920 every time you want to verify your backups actually work
One enterprise discovered their disaster recovery test - JUST THE TEST - cost $1,200 in egress fees. They now test quarterly instead of monthly. So much for "best practices."
The Hidden Tax on Everything You Do
Here's what makes this particularly insidious: Egress fees tax innovation.
- Want to run analytics across regions? That'll cost you.
- Need to sync data for machine learning? Pay up.
- Building a global application? Hope you budgeted for data transfer.
- Implementing proper backup strategies? Every copy costs extra.
A marketing analytics company reported that 60% of their cloud bill was egress fees. Not because they were doing anything wrong. They were doing everything right—automated exports, API integrations, real-time dashboards. The cloud providers built the tools to make this easy. They just didn't mention the meter was running the whole time.
The Path Forward: Intelligence at the Edge
This is why intelligent data pipelines aren't just nice to have; they're economic survival. The future isn't about moving all your data to the cloud or pulling it all back on-premises. It's about being smart about what moves where.
At Expanso, we're building pipelines that understand the economics of data movement, not just the mechanics. Instead of blindly replicating everything everywhere, our intelligent pipelines:
- Process data where it lives, not where it's expensive
- Move insights, not raw data
- Cache strategically based on access patterns and costs
- Route requests through the most economical path
Think of it as Google Maps for your data—but instead of avoiding traffic, you're avoiding tolls that charge by the megabyte.
The old model was simple: centralize everything, figure it out later. The new model requires intelligence: process at the edge, aggregate intelligently, move only what matters.
The $100 Billion Question
Flexera's 2025 State of the Cloud Report found that organizations waste 27% of their cloud spend on average. Apply that to the projected $723 billion in cloud spending this year, and we're burning nearly $200 billion on inefficiency.
But here's the thing: It's not inefficiency. It's architecture designed for a world where data transfer was free and bandwidth was infinite. That world doesn't exist.
The real question isn't "cloud or on-premises?" It's "how do we build systems that don't bankrupt us when they succeed?"
Every time your application scales, every time you add a region, every time you implement a best practice—the bill grows exponentially. Not linearly. Exponentially.
Your Next Steps (Before Your Next Bill)
- Run the egress audit: Pull your last three cloud bills. What percentage is data transfer? If it's over 20%, you have a problem.
- Map your data flows: Where is data actually moving? Most companies have no idea until the bill arrives.
- Calculate repatriation costs: Sometimes it's cheaper to pay the ransom once than keep paying protection money forever.
- Implement intelligent routing: Stop moving raw data. Process locally, move results.
- Challenge the architecture: That beautiful microservices architecture might be beautiful for your cloud provider's revenue.
The cloud isn't evil. It's brilliant for elastic workloads, innovation, and scaling. But the economics only work if you're smart about data movement.
Because right now, while you're reading this, your data is probably crossing a region boundary somewhere. And the meter is running.
What percentage of your cloud bill is data transfer? Have you calculated what it would cost to leave your current provider? Let's talk about the real economics of cloud.
In addition to being CEO of Expanso, I'm currently writing a book based on what I have seen about the real-world challenges of data preparation for machine learning, focusing on operational, compliance, and cost. I'd love to hear your thoughts!
Originally published at 'My Data Transfer Bill Cost What?': When Cloud Economics Go Wrong.