Latest / The Cloud Business Podcast with Fexingo: AWS, Azure, GCP, and Enterprise Infrastructure / The Cloud Data Egress Tax Nobody Is Talking About
Transcript
- Lucas: So there's this quiet tax on enterprise cloud bills that nobody talks about at the cocktail parties. It's not compute, it's not storage — it's data egress. Luna: Egress — the cost of moving data out of the cloud. It's the fee you pay to leave. Lucas: Exactly. And it's wildly profitable for the providers. AWS alone made something like eighty billion dollars in egress revenue last year. That's more than the entire revenue of companies like Salesforce. Luna: Eighty billion — that's staggering. And most enterprises I talk to treat egress as a fixed cost. They just pay the bill. Lucas: Right. So today I want to walk through one specific case. A mid-size SaaS company — we'll call them DataStream, not their real name. They're on AWS, about five hundred employees, running a data analytics platform. Their cloud bill was running about two million dollars a year. And they noticed it kept creeping up quarter over quarter — like eight to ten percent. Luna: Classic cloud drift. But you're saying egress was the culprit? Lucas: It was a huge chunk. When they finally did a deep audit — and I mean line-item level — they discovered they were spending four hundred thousand dollars a year on data transfer out. That's twenty percent of their total bill. Just for moving data out of AWS to their customers, to their analytics partners, to their own on-prem data center. Luna: Four hundred K on egress. That's wild. What were they moving? Lucas: Mostly customer-facing dashboards — their product streams real-time analytics to end users. So every time a customer loads a dashboard, DataStream's servers pull data from AWS and push it over the internet. That's charged at standard egress rates — about nine cents per gigabyte for the first ten terabytes a month. It adds up fast. Luna: And those rates are notoriously sticky. AWS dropped them a bit a few years ago, but it's still their highest-margin service. Lucas: Exactly. So DataStream did three things. First, they looked at their data flow and realized a lot of data was being egressed that didn't need to be. For example, their nightly batch jobs were pulling full datasets out to an on-prem analytics cluster when they could have run those analytics in AWS and only exported the results. Luna: So they shifted the compute to the data instead of the data to the compute. Lucas: Right. They migrated those batch jobs to AWS Glue and Redshift. That alone cut their egress by about thirty percent. Second, they negotiated a private pricing agreement with AWS. Most enterprises don't realize that egress is negotiable — especially if you're above a certain spend threshold. Luna: Right — the published rates are for the masses. If you're spending seven figures, you can get a discount. But you have to ask. Lucas: You have to ask. And you have to be willing to walk. DataStream went to AWS and said, look, we're considering moving some workloads to Azure or GCP because their egress rates are lower. And AWS came back with a fifteen percent discount on all data transfer out. That saved them another sixty thousand a year. Luna: And the third thing? Lucas: They started using a CDN — CloudFront — for customer-facing data. AWS actually doesn't charge egress for data served through CloudFront if you're using the origin shield. So they put CloudFront in front of their dashboards, and that eliminated the egress charge for the biggest chunk of their traffic. Luna: That's clever. So they turned a cost center into a performance improvement. Lucas: Exactly. Their dashboards actually got faster because CloudFront has edge caches. So they saved two hundred thousand dollars on egress, improved user experience, and reduced origin load. Total win. Luna: I love stories where the fix aligns with better product. But not every company has that architecture. What about companies that just need to move data out for legitimate reasons? Lucas: Great question. If you're a data-intensive company — say, a media platform streaming video — you're going to egress a lot no matter what. But even then, there are options. One is to use multiple cloud providers and route traffic through the provider with the lowest egress cost for that region. That's the multi-cloud arbitrage play. Luna: That adds complexity, but for huge volumes it might be worth it. Lucas: Totally. Another option is to use a third-party transit provider. Some companies are bypassing cloud egress entirely by peering directly with internet exchanges or using services like Megaport or Equinix Fabric. You can get data out at a fraction of the cost — like one to two cents per gigabyte instead of nine. Luna: So you're basically building your own off-ramp. Lucas: Exactly. And for very large enterprises, that's becoming standard. But for most mid-market companies, the easiest fix is just understanding your egress footprint. I'd bet eighty percent of companies don't have a granular breakdown of their data transfer costs. Luna: And that's the real problem — visibility. If you can't measure it, you can't manage it. Lucas: That's the takeaway. So if you're listening and you have any kind of cloud bill, go into your cost management console right now and check your data transfer line item. I bet it'll surprise you. Luna: Honestly, if today was worth a coffee to you, that's the link — buy me a coffee dot com slash fexingo. We keep this show ad-free because of listeners like you. Lucas: Yeah, it really helps us keep going. So if you got something out of this episode, that's the way to support us. Luna: Exactly. And speaking of visibility, there's another angle I want to dig into — the startup side. A friend of mine started a company last year that's building an egress-optimized architecture from day one. They chose GCP specifically because Google doesn't charge for egress to certain destinations. Lucas: Right — Google's free egress to other Google services like YouTube or BigQuery. But also, they have a free tier of egress up to one hundred gigabytes a month for each of the first several months. That's great for startups. Luna: Exactly. So they designed their entire pipeline to keep data within Google's network. They use Cloud Storage for data lakes, BigQuery for analytics, and they serve customer dashboards through Cloud CDN. Their egress costs are near zero. Lucas: That's smart design. But it also locks them into GCP. The same way AWS egress is a lock-in mechanism, GCP's free egress is a carrot to keep you inside their ecosystem. Luna: True. But for a startup, that's a trade-off they're willing to make. They don't have the leverage to negotiate discounts, so they optimize for the architecture that minimizes cost. Lucas: And that's totally rational. The key is just to be aware that you're making that trade. So much cloud spending is passive — you just let the default architectures drive the bill. Being intentional about egress can save tens or hundreds of thousands. Luna: Alright, so final thought — if you're an enterprise, start with the audit. If you're a startup, design for low egress from the start. And if you're anywhere in between, negotiate. Lucas: Good summary. Next episode we'll look at the flip side — ingress costs, which surprisingly are usually free. But there's a catch. Luna: Can't wait. See you next time.