In 2015, General Electric’s CIO Jim Fowler hosted the keynote address at AWS re:Invent, where he told a packed auditorium that the company was going through one of the “largest, and frankly most important transformations in [its] 140-year history.”

That transformation he was describing was GE’s planned (near) 100 percent move to the cloud. The company shut down its data centers, keeping only those few that would house its “secret sauce” and put its trust in Amazon Web Services (AWS) to handle what it does best - the digital infrastructure side.

In the years that followed, many other companies and organizations have followed suit, adopting a “cloud first” approach to IT buying, while tech vendors have embraced a new way of selling, moving away from traditional licensing agreements to offer their products in the cloud ‘as a service.’ The great migration continued at a pace in 2023, with organizations including Choice Hotels, the US stock exchange Nasdaq, and the UK’s NHS, hitting milestones on their cloud journeys.

The philosophy has been to lease, not buy, and for many organizations, this was the logical choice. If you aren’t a digital infrastructure or IT specialist, then why cling to that responsibility?

During the initial stampede to the cloud, conventional wisdom said it would be cheaper because it eliminated the upfront CapEx costs of hardware. Jamal Mazhar, founder and CEO of cloud technology Kaavo, said once: “Cloud computing is empowering, as anyone in any part of the world with [an] Internet connection and a credit card can run and manage applications in the state of the art global data centers; companies leveraging cloud will be able to innovate cheaper and faster.”

But the very nature of the cloud business model means that users can incur substantial hidden costs, such as data egress fees. This is where cloud companies, particularly the hyperscale providers such as the market-leading trio of Amazon Web Services (AWS), Microsoft Azure, and Google Cloud, will let a business send data to the cloud free of charge, but will then charge a certain amount per gigabyte to access it. This can add up fast, leaving users with dramatic and unexpected bills.

David Friend, CEO of Wasabi Technologies, a cloud storage company, told DCD: “Egress fees are an integral part of Amazon, Microsoft, and Google's business models - we have customers who have told us that 50 percent of their cloud bill was egress fees.”

Egress fees vary depending on the type of cloud storage used, but, according to Dennis Hahn, principal analyst of data center storage at Omdia, we can think of them simply as a charge incurred any time the data needs to leave a cloud provider’s storage and go back onto the Internet to be sent anywhere else.

“Most of the traffic, I would say, goes out of object storage onto the Internet. That is typically somewhere around seven cents a gigabyte,” said Hahn. “That doesn't sound like much, but as a data center guy, I’ve got terabytes, not a gigabyte. So all of a sudden that starts to add up.”

Cloud market domination

Egress fees are one of the charges that have led to a slew of antitrust investigations into the cloud market after concerns that it essentially causes “vendor lock-in,” where the cost of switching providers becomes too great for many enterprises to contemplate.

Hahn offered a bite-sized example: “A 32 terabyte drive, which is fairly common and the highest capacity drive you can buy right now, costs around $700 - or for an enterprise this may be less, closer to $500. If you start working out how much that is at seven cents per gigabyte, you get closer to $2,240.”

When this starts getting multiplied by 10s, 100s or even 1000s, it is easy to see how the cost of migrating away from a provider becomes much less appealing.

The problem is exacerbated, critics say, by the dominance the ‘big three’ have in the sector. Between them, AWS, Azure and Google Cloud controlled 66 percent of the market in the fourth quarter of 2023, according to figures from Synergy Research Group. During this period, worldwide cloud market spending amounted to $74 billion.

The US Federal Trade Commission (FTC) kicked off its antitrust inquiry in March 2023, releasing a Request for Information on the business practices of cloud computing providers.

One of the key pillars of that investigation surrounded market power and business practices affecting competition. The FTC noted that, for instance, “if there are few cloud provider options, it is possible that customers may have to agree to less competitive contract terms.”

Pricing practices, the regulator continued, “can also be confusing due to the wide array of services offered by cloud computing providers, each with complex pricing structures – making it difficult for companies to compare contract terms and forecast or control their spending on cloud.”

It also queried the “structure and design of offerings from cloud providers,” stating they “may make it more difficult for customers to switch to another provider for the majority of their cloud usage, or to mix and match offerings from multiple providers.”

The November 2023 findings of that inquiry cited egress fees as one of the most common concerns: “According to some commenters, these egress fees could have the effect of discouraging customers from using multiple cloud providers or switching from one provider to another.” it said. Beyond that, the competition was found to be impacted by cloud providers incentivizing customers to consolidate their use of services to just one provider, and also offering “minimum spend contracts.”

Shortly before the FTC investigation concluded, the UK’s Competition and Markets Authority (CMA) launched its own probe, which at the time of writing, remains ongoing. The CMA’s inquiry also holds egress fees as a key point of contention, describing them as a “significant barrier” to switching providers or using a multi-cloud set-up, where businesses pick and choose different services from different platforms.

The CMA investigation comes off the back of an Ofcom report which found that the egress fee charges by hyperscalers are “unlikely to be necessary for cost recovery and that egress list prices are likely to be higher than the incremental costs of providing the service.”

Similar investigations and conversations in France, Japan, and the Netherlands have noted the detrimental impact these anti-competitive practices have on other cloud providers, and by extension, the enterprises and organizations seeking to use their services.

Throughout these debates, the European Commission has been working on the EU's Data Act, which officially went into force on January 11, 2024.

The Data Act will require public and private cloud computing service providers to remove "obstacles to effective switching" between their own and competing cloud services, including commercial, contractual, technical, or organizational hurdles. This includes egress fees.

Currently, cloud providers have until 12 September, 2025, to comply with this, meaning that a dramatic shift is unlikely to occur in the immediate future.

Despite this, Google released an announcement also on January 11, 2024, stating that it would be removing exit fees from its offering. In a blog post, the cloud provider’s head of platform Amit Zavery wrote that 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 out of Google Cloud. This applies to all customers globally."

Zavery went on to say that "restrictive and unfair licensing practices" are a fundamental issue, suggesting that "certain legacy providers" leveraged their software monopolies to create cloud monopolies that "lock in customers and warp competition."

Despite condemning those “certain legacy providers,” it is worth noting that Google had fees that functioned in the same way, until January 10, 2024.

In light of this, DCD reached out to AWS and Microsoft. Microsoft did not respond, but AWS shared via email: “AWS designs cloud services to give customers the freedom to choose technology that best suits their needs. Since 2021, over 90 percent of our customers pay nothing for data transfers out of AWS because we provide them with 100 gigabytes per month for free to use for any purpose.

"In accordance with the European Data Act, for the small number of remaining customers, AWS will not charge eligible EU customers more than cost for data transfers out to support switching or in parallel use of other relevant providers. Restrictive licensing practices remain a far bigger issue to customers who want the choice of working with their preferred cloud provider.”

Based on Hahn’s estimated seven cents a gigabyte, that represents about $7 of free data transfer a month.

On March 3, 2024, AWS announced that it had followed Google's lead, and removed exit fees within some stipulations. Corey Quinn, chief cloud economist of the Duckbill Group, a company that helps companies optimize their AWS bill, took us through one of his test AWS accounts - of which he says he has 18 - to take a look at how much he uses of the free allotment.

The test account, amusingly dubbed his “Shitposting Account,” uses up around a fifth of this free data transfer allotment.

“What they're saying is that 90 percent of our customers don't pay anything,” Quinn says. "Yeah, because most of your customers are tiny test accounts. I don't know anyone who works on this stuff professionally, that doesn't have their own independent AWS account just for learning purposes.

“It's a question of how many of those 90 percent are actual businesses.”

When it comes to Google Cloud’s offer, Solange Viega Dos Reis, chief legal officer at cloud provider OVHcloud, said it amounts to “the first step in the right direction and that it is so small. If you read the announcement well, you realize that.”

Google’s move, while touted as a game changer by many, only applies to those customers who wish to fully exit Google Cloud, which must first be approved by the Google Cloud support team. Customers will then have 60 days to fully exit and must then terminate their agreement. In other words, you can’t take advantage of the “no exit fees” if you just want to move some workloads or applications to another provider. This is a full divorce, and nothing less.

Wasabi’s David Friend told DCD: “It's really window dressing, because the issue is not people wanting to leave a cloud provider. It's people who are using the cloud and have to pay egress fees every day.

“I think what they did was they looked at the proposed legislation in the EU and said, ‘Okay, we'll do something which has absolutely no meaning to 99.9 percent of our customers, but it satisfies the government.’”

Viega Dos Reis points out that it shows that the egress fees were not really necessary - “It confirms what the industry already knew,” she says. She also considers it likely that other cloud providers will soon follow suit, and has already been proven correct by AWS.

Similarly, Hahn is somewhat optimistic about the move. “It is a step in the right direction,” he says. “It’ll probably bring others to do that, but they may still have some work to do.”

Hahn notes that most people are unlikely to fully exit a cloud provider, so the concession is more or less irrelevant. Instead, he says the complexity of pricing models is the real issue.

“You end up with really unpredictable storage costs,” he explains. “As a user there’s a whole bunch of really complex fee structures, there are so many exceptions within them all. For example, there are at least six different fees for object storage, and it can add up quickly.”

Quinn agrees. “It's the ebb and flow of conducting business, while happily remaining AWS customers, that gets expensive,” he says. “There's a reason that Netflix still has its entire own home-built content delivery network. They don't stream any of their movies out of AWS because even at their scale, the cost would be ruinous.

“The underlying economics of data transfer does not reflect how the cloud providers price for it. We're still paying 1990s prices for bandwidth when we’re in the cloud.”

As for the trail-blazing Google announcement, Quinn describes it as “performative nonsense.”

“If they have three customers a year that take them up on that offer, I’d be surprised,” he says. “It’s genius on their part. They look like they’re doing something customer-friendly, and it costs them effectively $0.”

Competing as an underdog

While the presence of such fees can leave large organizations feeling like they are being stung by the lack of competition amongst large cloud providers, they can also provide an opportunity for smaller players in the market to offer something different.

Some cloud providers do not charge egress fees at all, or use a more simple payment structure, says Kevin Cochrane, CMO at one smaller player, Vultr.

“Bandwidth is the pricing vector that's the most unpredictable and the most egregious out of all of the overcharges,” argues Cochrane, noting the similarity across the board for virtual machines on all cloud providers for compute price, storage price, and plan.

Pulling up a PowerPoint slide, Cochrane offers a comparison: 3TB of egress is charged at $307.41 per month by AWS, $286.63 by Google, and $299.14 by Microsoft, while Vultr charges nothing at all.

“We cut prices on those core services by 40 percent, and we are still wildly profitable. That’s how badly they are overcharging,” he says.

Cochrane describes vendor lock-in as “Stockholm syndrome by proxy.”

“Industry influences have told me, ‘lock in isn’t the problem. People want to be locked in.’ They knowingly enter contracts to be locked in, because of all the benefits,” he says.

“If you talk to anyone, they’ve all done past optimization initiatives. They’ve negotiated discounts from Amazon, Azure, and Google, but the way they do that is by doubling down on investment and saying ‘we’re going to build everything new on you,’ and that’s where vendor lock-in comes in.”

One method companies use to avoid this is to store their data on-premise, or with a cloud storage provider that doesn’t charge egress fees. They can then make copies to send to another cloud provider to process their data, only to abandon it there, thus avoiding extra charges.

“I have seen some use cases of that,” says Omdia’s Hahn. “Not necessarily targeted at egress fees, but as a way to kind of sidestep some of the costs of storage as well as to keep it under your control.”

This can also help businesses to avoid lock-in. If they own their data, they can choose where it goes, according to the best offerings. They can then abandon that version of the data reassured in the knowledge that the original information is safe on their own server.

The Ofcom report found that some of the hyperscalers are charging between five to 10 times as much egress fees as other cloud providers like OVHcloud and Oracle, while other cloud providers make no charge at all.

Viega Dos Reis told DCD that OVHcloud only charges egress fees in two of its services and that both have very few data transfers - being used for storage rather than ‘live’ data. “It’s a very low price and for the rest of the services - say 99 percent - egress fees don’t apply,” she says.

The company also encourages customers to take a multi-cloud approach, understanding that other cloud providers may be able to better provide certain services. The company offers customers discounts based on usage - in the same way you might get a “three for the price of two” deal at the supermarket - but exclusivity is never included in any contract.

OVHcloud is also open-source, something it says makes it different from the hyperscalers. Dos Reis explains that this means clients “can easily move data from their data center and to our cloud, or then move to another cloud provider, another server. The data is easily interoperable with other environments.”

She adds: “It's not what we see with the hyperscalers, they have their own models. And when you remove your data from Microsoft to put it on a different cloud, there is a big amount of time to translate the format of the data from Microsoft to an open source format.”

Cloud repatriation

Despite the financial and technical barriers, some companies are undertaking “cloud repatriation,” where workloads are moved off the cloud and back on-premises. A UK-based study conducted by Citrix found that, of the 350 companies surveyed, a quarter have moved half or more of their cloud-based workloads back to their own infrastructure, or are considering doing so.

Among the list of motivations, Citrix noted unexpected costs (33 percent of respondents), performance issues, security concerns, compatibility issues, and service downtime. Twenty-two percent listed financial concerns as the main motivation for repatriation.

Fifty percent of respondents identified data transfer fees as a significant contributing factor to unexpected cloud costs.

Omdia’s Hahn has seen more repatriation of late. “In the last year or two years, companies have been a lot more selective,” he says. “There used to be a trend of cloud first, cloud all in, but now it seems more like companies think, ‘okay, we’ve got these workloads, some of them make sense to go to the cloud, some make sense to go on-premise.’”

But Quinn disputes that this is actually cloud-repatriation at all.

He agreed that, on occasion, the Duckbill Group might recommend moving workloads away from the cloud, but that it is “not something we ever recommend lightly.” He argues that far more than “repatriation,” he is seeing “expansion.”

“People are always wondering about [cloud repatriation],” Quinn says. “What I have seen is people are declining to move individual workloads into the cloud, because they'll do basic math with a calculator and realize how pants-shittingly expensive it is and say, ‘well, glad we dodged that bullet.’

“But that's not really a repatriation. We'll see workloads occasionally migrating from one place to another when there's a business case behind it.”

Quinn adds: “But spinning that up in Google Cloud does not mean that Google won a workload from AWS or they’ve migrated from one to the other, or built an on-prem data center offering, that’s not repatriation. That’s an expansion.”

He goes on to explain that many companies won't even migrate to a different cloud region at the same service provider purely because of how complicated that is.

An example offered is that of review site Yelp, which he says moved to the AWS US West region in California. Only after migrating, did the company realize that it would have been far cheaper to be in AWS US West (Oregon) region. DCD has contacted Yelp for comment.

True cloud repatriation - a true divorce - is not a common story. In October 2022, web software company 37Signals announced it was moving its Basecamp and Hey services off the cloud, and claimed to have saved $7 million with an investment of $600,000 in servers.

In December 2023, LinkedIn revealed it was canning a long-planned migration and favour of keeping much of its data on its own servers, supplemented by cloud services. This was somewhat surprising given that it was due to migrate to Azure, the cloud platform owned by its parent company Microsoft. “We are using both Azure to complement our infrastructure needs and further investing in our data centers,” a LinkedIn spokesperson told CNBC at the time.

Back in 2016, online storage firm Dropbox said it was migrating off of the cloud as part of its “Magic Pocket” plan. Within a few years, Dropbox had moved around 34 petabytes of analytics data back to AWS and remains a customer.

It seems organizations considering the cloud will have to keep evaluating their costs, and use a complex variety of storage solutions alongside cloud computing tools if they want to derive the best value.

Certainly it seems some egress fees and other hidden charges are here to stay. Vultr’s Cochrane describes some of the behavior of the hyperscalers as “terribly abusive,” but says moves to cancel exit fees are “a step in the right direction.”

He adds: “They are admitting that there’s a problem, and the first step to recovery is admitting you have a problem."