Google recently announced that it would eliminate data egress fees for those leaving the platform. Given our position on the cloud operating model and the lifecycle of the cloud, this appeared to be a major announcement. It is not.
You could understand our initial enthusiasm. Google stated that any "customers who wish to stop using Google Cloud and migrate their data to another cloud provider and/or on-premises can take advantage of free network data transfer to migrate their data." They went on to say that "eliminating data transfer fees for switching cloud providers will make it easier for customers to change their cloud provider." Google even addressed the inherent issue with access and egress fees outright: "The fundamental issue that prevents many customers from working with their preferred cloud provider in the first place [is] restrictive and unfair licensing practices”.
Pretty noble stuff. But the devil is in the details, and that is where these noble statements fall flat. Frankly, very little changes, except you do the work to help them assess the value of your patronage. There is, however, something here and we break it down as there are far-reaching implications – mostly from a regulatory perspective. Let’s dig in.
No More Data Hostages
We all know how the cloud works at the Big 3: it's free or inexpensive to move data in, but getting your data back comes at a much greater cost. For example, data egress costs $0.09/TB at AWS, $0.08/TB at Azure and $0.12/TB at GCP (prices are simplified from tiered pricing to get a single fee). Google's announcement acknowledges that customers shouldn't be held hostage by egress fees, especially when they serve as a barrier to exit.
However, egress fees exist to offset the well-defined Lifecycle of the Cloud. Conventional wisdom is that the cloud is a very expensive place but the value of provided tooling and services makes it worth it. The pattern is to start on the cloud, ramp up and build using all their great tooling and services – and then repatriate out of the cloud, run systems in a data center and stop paying a premium for open-source tooling and services.
Egress fees threaten the cost-effectiveness of investments in cloud technologies and services. According to a March 2021 study by IDC, "Future-proofing Storage: Modernizing Infrastructure for Data Growth Across Hybrid, Edge, and Cloud Ecosystem", a staggering 99% of cloud storage users reported incurring planned or unplanned egress fees with these fees amounting to nearly 6% of an enterprise's total cloud storage bill.
What? 6% of all the money you're spending on cloud storage simply goes towards accessing your own data. Plain and simple – egress fees only benefit the cloud provider and they don't make sense. Go all the way back to mainframe days – you paid to submit and process a job, not to read the output and delete the job. Another example is that you might want to build your own car, motorcycle or bicycle. One common way of doing this is to rent time in someone else's shop to gain access to work space and specialized tools. It's accepted in this model that you pay for parts and time using tools – there's no separate charge when you drive your creation home.
Cloud providers are under pressure and they know it. 37 Signals is saving millions, X is saving tens of millions, Ahrefs is saving hundreds of millions. When pulling out of the cloud saves you 60% or more, it's time to get smart about what you're building and where you're running it.
What Google Really Said
Reading the Cloud Exit FAQ with a fine-toothed comb, it becomes evident that there are enough conditions, restrictions and judgment calls that you're really going to have to jump through hoops if you want free data egress from Google.
Customers are only eligible if the intention is to leave GCP forever – all accounts must be terminated when the migration is complete. Buried in the FAQ was a statement that customers that only want a partial exit should reach out and inquire also.
The first thing disgruntled customers must do is contact their assigned Google account team member and announce their intention to leave. Then they fill out a "free data transfer form" for Google Support to review. The form asks how many GB customers intend to migrate out of BigQuery, Cloud Bigtable, Cloud SQL, Cloud Storage, Filestore, Persistent Disk and Spanner.
Google Support then reviews the form and decides whether or not the customer qualifies for free data egress. If approved, customers have 60 days to terminate their account. They will be charged the data egress fees, but those fees will be credited back to their final invoice.
The thing is, this is what Google has been doing for years anyway. We're all familiar with the customer retention process to reduce churn. When you tell Google you want to leave, they would key all these values into a spreadsheet to decide how much your business is worth and then offer a discount to stay. All that's changed today is that you have to complete the spreadsheet that helps them decide what you're worth.
Why Google Said It
So why all this friendliness for churning customers all of a sudden?
It seems that a little thing like an antitrust investigation conducted by the EU comes into play here. A recent report by the UK's Competition Markets Authority voiced concerns about anti-competitive practices like egress fees, interoperability, software licensing and personalized discounts.
Rather than address their own business practices that engender lock-in, AWS and GCP have chosen to attack Microsoft over its licensing fees. They've made so much noise that both the EU and the US FTC have launched antitrust investigations into the major cloud providers over anti-competitive behavior. According to Google, they have to engage in anticompetitive practices to counter the "on-premises software monopoly [being used] to create cloud monopolies, using restrictive licensing practices that lock in customers and warp competition."
Let's be clear. This is simply positioning. All three (AWS, GCP and Azure) use egress and data fees to lock you in. Google would like you to think they have taken a different approach - but they really have not. They probably hope that this will spur a deeper look at egress fees by regulators - and as the third place player, that would likely benefit them incrementally by increasing data mobility. That is the silver lining.
Free Data Egress or Death
Allowing customers to work with their data when, where and how they see fit is the right solution. It allows them to optimize their workloads and makes the competition in the cloud around the services, not the infrastructure. These freedoms are core to the principles of the cloud operating model. As the cloud operating model grows in popularity, the Big Three find themselves in conflict with the very principles that allowed them to be successful in the first place – and it is driving a renaissance in on-prem deployments.
The value is in the data. You know it, AWS knows it, GCP knows it, even your CEO knows it. We've previously written that the cloud isn't a place, it is an operating model. There is a movement afoot to repatriate data from clouds to corporate data centers. The source of truth data lake resides on-premise in cloud-native object storage such as MinIO and datasets are pushed to the public cloud provider with the best combination of services and fees. This is the essence of cloud competition, it is 100% enabled by S3-API object storage, Kubernetes and open-source software.
Your brand new AI initiative is likely to suffer from egress fees. Do you want to push training data between on-premise and various clouds to take advantage of GPU processing or a machine learning service? You'd better think twice because egress fees can cause underlying costs to double. Can you really guarantee ROI when startup costs for ML training are twice what they should be? Don't let cloud providers hold your data hostage.