November 15, 2026
Vendor Lock-In: What It Is, Types, Costs and How to Avoid It
Vendor lock-in is a situation where your business depends so heavily on one supplier’s products, data formats or contracts that switching to another supplier becomes too costly, slow or risky to be practical. It is not always a mistake, and it is not only a cloud problem. It can build up quietly in ERP systems, custom software built by an outside agency, SaaS platforms and, increasingly, AI tools. This guide explains how lock-in happens, the four main types, what it really costs, and how to reduce it without giving up the benefits of good vendors.
What Is Vendor Lock-In?
Vendor lock-in (also written “vendor lock in” or “proprietary lock-in”) means a customer cannot easily leave a vendor without major cost, effort or disruption. The vendor may not have done anything sinister. Often the lock-in is a side effect of convenience: you adopt proprietary features because they are fast to use, and years later those features are woven into everything.
A useful test: if your vendor doubled the price tomorrow, could you realistically leave within a year? If the honest answer is no, you are locked in.
Why Does Vendor Lock-In Happen?
Lock-in rarely comes from one decision. It is the sum of many small ones:
- Proprietary services. Managed databases, serverless functions and AI platforms that have no equivalent elsewhere.
- Closed data formats. Data you can technically export but cannot use anywhere else.
- Undocumented systems. When a vendor built your software and the knowledge lives in their heads, not yours.
- Commercial terms. Multi-year commitments, volume discounts and prepaid credits that punish leaving.
- Switching costs. Retraining staff, rewriting integrations and running two systems in parallel during a migration.
What Are the Main Types of Vendor Lock-In?
Most lock-in falls into four categories. Knowing which one you face determines the fix.
| Type | What it looks like | Typical example | How hard to escape |
|---|---|---|---|
| Technical / platform | Code depends on services with no equivalent elsewhere | Applications built on one cloud’s proprietary database or serverless runtime | High: needs rewrites |
| Data | Data is hard to export or unusable outside the platform | Proprietary file formats, custom metadata, per-gigabyte exit fees | Medium to high |
| Contractual / commercial | Contract terms make leaving expensive | Multi-year commitments, auto-renewals, prepaid credits | Medium: time-bound |
| Knowledge / operational | Only the vendor understands how your system works | Custom software with no documentation, owned in practice by the supplier | High: invisible until you try to leave |
The last row is the one most guides skip. It is also the one Promact sees most often with mid-sized companies: the contract may allow you to leave, but nobody on your side can explain how the system works. We come back to this below.
What Does Vendor Lock-In Cost?
The cost shows up in three places.
- Price. Vendors with captive customers can raise renewal rates. The best-known recent example is VMware after Broadcom’s acquisition. A CloudBolt survey of 302 IT decision-makers at North American organisations with 1,000 or more employees found that 14% had seen VMware costs rise by more than 100%, 12% by 50-99%, 33% by 24-49% and 31% by less than 25%. The same report says 86% of organisations are actively reducing their VMware footprint. These are industry figures from a single vendor-published survey, so treat them as directional rather than definitive.
- Speed. When every change waits in a vendor’s queue, your roadmap slows. Competitors ship integrations and AI features while you wait.
- Risk. If the vendor is acquired, changes its terms or retires a product, you inherit the consequences. Our post on AWS retiring Bedrock Agents Classic shows how even a major platform can force a migration on its customers.
What Are Examples of Vendor Lock-In?
- Cloud egress fees. Moving large datasets out of a cloud provider has historically been priced to discourage it. Google Cloud, AWS and Microsoft all waived exit fees in 2024, but only for customers leaving entirely and with conditions attached.
- Virtualisation licensing. Customers running core infrastructure on one hypervisor discovered how hard it is to move hundreds of virtual machines when pricing changed.
- Custom-built software. A company commissions an application from an outside team, never receives usable documentation, and finds every change requires the original supplier.
- AI platforms. Prompts, agent workflows, fine-tuned models and vector stores built around one provider’s API are costly to port. Our piece on the Cursor cutoff and change-of-control clauses covers a contract-level version of this risk.
Is Vendor Lock-In Always Bad?
No. Committing to one vendor can cut integration work, simplify support and earn volume discounts. The question is whether you are choosing the dependency with your eyes open, or drifting into it. A deliberate lock-in with a documented exit plan is a business decision. An accidental one is a risk.
How Do You Avoid Vendor Lock-In?
You cannot eliminate dependency, but you can keep your exit cost low enough that you retain negotiating power.
1. Own your data and keep it portable
Insist on export in open, machine-readable formats and test the export before you need it. Store critical data in standard formats such as standard SQL or S3-compatible object storage where you can.
2. Put a thin abstraction layer in front of proprietary services
Wrap vendor-specific calls behind your own interfaces so swapping a provider means changing one layer, not the whole application. This matters most for AI, where models and APIs change quickly. Our post on why single-vendor AI bets are ending explains why multi-model design is becoming standard.
3. Negotiate exit terms up front
Before signing, ask: What is the notice period? Are there exit or egress fees? Who owns derived data, configurations and documentation? What happens on acquisition or product retirement?
4. Demand documentation and code access
For custom-built software, make source code, architecture diagrams and runbooks a contractual deliverable, not a favour. Undocumented systems are the hardest form of lock-in to reverse.
5. Prefer open standards and portable tooling
Containers, open APIs and common protocols lower switching costs. For sovereignty-sensitive workloads, see our guide to sovereign AI for business.
6. Review your dependencies on a schedule
Once a year, list every critical vendor and score it for exit difficulty, as in the table below.
| Question | Low risk | High risk |
|---|---|---|
| Can we export all our data in an open format today? | Yes, tested | Never tried |
| Does our team understand how the system works? | Documented and owned in-house | Only the vendor knows |
| Could we replace this vendor within 12 months? | Yes, plan exists | No plan |
| Are renewal terms predictable? | Capped increases | Uncapped |
| Is a proprietary service used with no equivalent? | No | Yes, widely |
Does the EU Data Act End Cloud Lock-In?
It helps, but it does not remove it. Under the European Commission’s Data Act explainer, the cloud switching rules have applied since 12 September 2025, and switching charges, including data egress charges, are removed entirely from 12 January 2027. Until then providers may still charge costs incurred for switching. Platform and software providers must also offer open interfaces and export data in a commonly used, machine-readable format.
Two caveats. First, reporting by The Register shows the big cloud providers’ early compliance offers are narrow: Google’s free transfers cover moves between a customer’s own services, and Microsoft and AWS ask EU customers to file requests. Second, removing fees does not remove the harder costs: rewriting code built on proprietary services, retraining staff and rebuilding integrations.
What If You Are Already Locked In?
If you are already deep in a vendor relationship, the first step is not a rewrite. Rewrites are expensive and risky, and most companies cannot afford the downtime. The practical path is:
- Assess the dependency. Find out what you actually own versus what lives with the vendor, where knowledge is trapped, and whether a transition is feasible.
- Reconstruct what is missing. Map architecture and dependencies and rebuild documentation so your own team can understand the system.
- Plan a phased transition. Move in stages that protect daily operations rather than attempting a big-bang switch.
This is the approach behind Promact’s Reclaim service, which starts with a vendor dependency assessment, builds system documentation and delivers a continuity and modernization roadmap. It is aimed at companies where software delivery has slowed, costs have risen, or switching feels impossible.
Conclusion: Practical Takeaways
- Define lock-in by exit cost, not by contract length: could you leave within a year if you had to?
- Identify which of the four types you face. Knowledge lock-in is the easiest to miss.
- Keep data portable, put abstraction layers in front of proprietary services, and negotiate exit terms before you sign.
- Treat documentation and code access as deliverables, not courtesies.
- Review critical vendors yearly, and give AI tools the same scrutiny as cloud and ERP.
- If you are already locked in, start with an assessment, not a rewrite.
Frequently Asked Questions
What is vendor lock-in in simple terms?
It is when you depend on one supplier so much that switching is too expensive, slow or risky. The dependency can be technical, contractual, data-related or knowledge-related.
What is an example of vendor lock-in?
A company builds its applications on one cloud provider’s proprietary database and serverless services. Years later, moving to another provider means rewriting large parts of the code, even if the data itself can be exported.
Is vendor lock-in always bad?
No. Committing to one vendor can reduce complexity and cost. It becomes a problem when it is accidental, undocumented, or leaves you with no realistic exit and no negotiating power.
How can companies avoid cloud vendor lock-in?
Keep data in open formats, wrap proprietary services behind your own interfaces, negotiate exit and egress terms before signing, and review dependencies yearly. Multi-cloud can help, but it adds its own cost and complexity, so use it where the risk justifies it.
Does the EU Data Act ban cloud egress fees?
It removes switching charges, including data egress charges for switching, from 12 January 2027. Providers can still charge cost-based fees in some situations, such as running services in parallel, so check the rules against your own contracts.
How do I get out of vendor lock-in if the vendor built my software?
Start by assessing what you own, what the vendor controls and what knowledge sits only with them. Then rebuild documentation and plan a phased transition rather than a full rewrite.
Sources
- European Commission: Data Act explained - official summary of the Data Act’s cloud switching rules and dates
- The Register: Big clouds scramble as EU Data Act brings new data transfer rules - how Google, Microsoft and AWS responded to the Data Act
- TechCrunch: Amazon follows Google in announcing free data transfers out of AWS - AWS’s March 2024 egress waiver and its conditions
- Petri: Enterprises Rethink Virtualization Roadmaps as VMware Pricing Fears Persist - coverage of CloudBolt’s 2026 survey of VMware customers
- Promact Reclaim - Promact’s vendor dependency assessment and modernization service
Have a project like this in mind?
Tell us what you're building — we'll help you scope it and ship it.
Talk to us

