Cloud vendor lock-in: welcome to hotel  Cloudifornia

What is vendor lock-in in cloud computing?

“There is no cloud. There are only other people’s computers”, It’s one of those old tech-industry jokes that started as a punchline and ended up as strategy. Because once you move your infrastructure to the cloud, you’re not soaring into some modernist abstraction. You’re renting rack space in someone else’s warehouse. And that someone, it turns out, would really prefer that you never leave.

Cloud vendor lock-in is the point at which convenience turns into dependence. It rarely happens overnight. Maybe your team needed a quick deploy, so you used a managed service. Then another. Then you wrote logic tied to a proprietary API. Then someone negotiated a great long-term pricing deal. By the time Finance asks about migrating to a second provider or bringing workloads in-house, the spreadsheet already says no.

It’s not a new story. It’s the digital equivalent of a company town: the provider owns the hardware, the services, the billing interface, and increasingly, the roadmap. Sure, you can leave. You just might not be able to take anything with you.

Cloud vendor lock-in is the point at which convenience turns into dependence.

Lock-in tends to show up in three flavors. Platform lock-in is when your entire system is architected around one provider’s stack. Data lock-in happens when you’re technically allowed to leave but financially discouraged, thanks to retrieval charges or opaque formats. Tool lock-in is when your DevOps pipelines only work in one environment and nobody remembers how to undo that.

The result is less flexibility, slower response to business changes, and a tendency to leave cost savings on the tableת not because no one wants them, but because no one can reach them.

Why is cloud vendor lock-in an issue?

Cloud computing, of course, was sold as the opposite of that. On-demand infrastructure. Zero upfront investment. Horizontal scaling. The pitch was so good that for a decade, no one asked what happens when you try to scale back. And when companies do finally try, they find out the hard way that cloud providers are more than happy to help you grow, but less excited about helping you go.

The blockers aren’t technical. They’re economic. Egress fees. Contractual friction. Architecture glued together with managed services that don’t translate. And it’s not an accident. It’s how the model works. The business of cloud isn’t about storage or compute, it’s about gravity.

When "pay-as-you-go" becomes "pay-to-leave"

Cloud providers love to advertise dirt-cheap storage and free uploads. Getting your stuff in is easy. But getting it out? AWS charges per gigabyte for standard transfers, plus some “modest” exit fees. For large enterprises, that modesty can add up to six-figure bills, just for the privilege of accessing their own data.

And egress fees are just the appetizer. According to Bloomberg, 90% of tech executives oppose these practices. The Coalition for Fair Software Licensing argues some vendors go a step further, threatening customers who consider switching clouds with “intrusive software audits” or “higher licensing fees.” When companies have to demand “freedom from retaliation” as a formal policy point, it’s a fairly clear signal that the market isn’t functioning properly.

Cold storage only sharpens the knife. Archive tiers promise radical savings, until you need to touch the data. Glacier Deep Archive runs at $0.00099 per gigabyte per month. Sounds great. But retrieving the data can cost orders of magnitude more. Businesses tend to discover this at the worst possible moment: during a compliance audit or a migration.

Architectural dependencies: the deeper lock-in

Vendor lock-in goes deeper than pricing. It's architectural. Applications built on Lambda, DynamoDB, CloudFormation, they don’t just pick up and move. Each proprietary service requires a replacement or a full rework. The initial speed of development turns into long-term technical debt. Infrastructure decisions harden into product decisions, roadmaps adapt to what the provider can do. Engineers specialize in one cloud’s tools, and Finance teams align to one cloud’s billing cycles. Before long, entire organizations are running on someone else’s logic.

Success stories: companies breaking free

Some companies have managed to get out. 37signals, the team behind Basecamp and HEY, grabbed headlines by announcing their cloud exit after AWS bills reached $3.2 million annually. Their calculations showed $7 million in savings over five years by returning to owned hardware. Dropbox did it earlier, building custom infrastructure to cut costs and boost performance. These are not margin-chasing optimizations. They’re strategic shifts by companies that know exactly what their workloads look like and no longer buy the cloud hype.

This tweet from DHH highlights the financial impact of their decision.

‍

‍
The sentiment is spreading. Retailers, banks, and healthcare companies are quietly repatriating workloads after failed contract renewals. Others hedge with multi-cloud setups, accepting the added complexity in exchange for a little leverage.

Planning your cloud exit strategy

None of this is new. Monopoly behavior is well-documented and famously resistant to self-correction. The classic remedy was structural: you could own the rails, or run a freight company, but not both. Banks couldn’t lend to companies they owned. The idea was simple, don’t let the referee play on the field.

Today, cloud vendors get to be both platform and provider, and somehow we’re surprised when they design the game to benefit themselves. Until regulators decide to revisit 19th-century antitrust lessons, companies need to plan accordingly.

Here’s what that looks like:

• Design for portability. Containers, abstraction layers, minimal proprietary glue.
• Negotiate exit terms at the start. Egress caps, retrieval waivers, portability clauses.
• Keep a map. Know what breaks if you leave.
• Measure switching costs quarterly. Code, training, downtime, track it like a liability.
• Consider structural separation. Maybe you need cloud services or cloud hosting, but not both from the same vendor.

Abraham Lincoln once insisted on buying rifles only from manufacturers who used interchangeable parts. The logic was war-tested: no general wants to lose a battle because one supplier cornered the market on a single bolt.
The cloud promised flexibility. But flexibility only matters if you can actually leave. Vendor lock-in isn’t a glitch. It’s the business model. And like every other monopolist before them, these platforms aren’t going to change, unless customers show up with a plan to escape.