For years, the strongest argument against changing cloud or SaaS providers was not technical. It was commercial and contractual: egress fees, long notice periods, vague exit clauses and data exports in formats nobody could easily load elsewhere. The EU Data Act changes that balance. Its switching rules have applied since September 2025, and from 12 January 2027 providers of data processing services in the EU may no longer charge switching fees at all.
That is a significant shift for anyone responsible for data platforms, but legal rights do not move a single table. A provider may be obliged to let you leave within defined timelines, yet if your data is locked in proprietary formats, your pipelines are wired to provider-specific services and nobody has tested an export, you will not be able to use those rights in practice. This article explains what the Data Act switching rules require, what they do not solve, and provides a practical readiness checklist for data and platform teams.
This article is general information, not legal advice. Work with your legal and procurement teams on how the rules apply to your contracts.
What the Data Act Switching Rules Require
Chapter VI of the Data Act (Articles 23 to 31) covers switching between data processing services, which includes infrastructure, platform and software services. Legal commentary from law firm Maples and draft guidance from the Dutch competition and consumer regulator ACM summarize the core obligations as follows.
- Notice period: customers can start a switch with a maximum notice period of two months.
- Transitional period: the switch itself should complete within 30 calendar days. If that is technically unfeasible, the provider must notify the customer within 14 working days with a justification and propose an alternative period of no more than seven months.
- Data retrieval period: customers get at least 30 calendar days after the transitional period to retrieve their data.
- Exportable data: contracts must specify exhaustively which data and digital assets can be ported, with possible exclusions where trade secrets are at risk.
- Erasure: after the retrieval period, the provider must erase the exportable data and digital assets.
- Interoperability transparency: providers must maintain an up-to-date online register of data structures, formats and relevant interoperability specifications.
- Switching charges: between January 2024 and 12 January 2027, charges are limited to costs actually incurred; from 12 January 2027, switching charges are prohibited.
Important nuances
The ACM’s provisional guidance, published for consultation in October 2026, clarifies several points that matter for planning. After 12 January 2027, providers may still charge data transfer costs for parallel use, where a customer runs two providers simultaneously with data flowing between them, but not above costs actually incurred, and technical support for parallel use may not be charged. Early termination penalties remain lawful only if proportionate and disclosed before signing, and charges not described in the contract cannot be billed. The ACM guidance includes a self-assessment checklist and may change after consultation, so check the final version.
Some services are treated differently. Maples notes that services whose main features are custom-built for an individual customer and not offered at broad commercial scale are partly exempt, and that the European Commission’s Digital Omnibus proposal would add further exemptions for certain custom-made services and smaller providers. These proposals were still in the legislative process at the time of writing.
Functional equivalence and interoperability
The switching chapter also reaches into technical obligations. Article 30 ties providers’ technical duties to the type of service, aiming to let the destination provider achieve functional equivalence, which an Irish legal analysis published via Legal 500 describes as a comparable customer experience after switching, while the source provider supplies reasonable help such as export tools and migration support. Article 26 requires the online register of data structures, formats, standards and open interoperability specifications mentioned above, and Chapter VIII adds broader interoperability requirements, including machine-readable dataset descriptions. For data teams, these provisions are useful levers: the register tells you what the provider commits to export, and you can map it against what you actually need.
What the Rules Do Not Solve
The Data Act makes switching legally and commercially possible. It does not make it technically easy. Several gaps remain the customer’s responsibility.
- Proprietary formats and features. A provider can comply by exporting data in documented formats, but if your logic depends on proprietary functions, stored procedures or platform-specific services, the export is only part of what you need.
- Pipelines and integrations. Ingestion jobs, orchestration, identity integration and downstream consumers are usually configured around one platform’s services and must be rebuilt or reconfigured.
- Metadata and governance. Lineage, access policies, data quality rules and business definitions are often stored in platform-specific catalogs. Check whether they are exportable and usable elsewhere.
- Volume and time. Moving many terabytes or petabytes still takes time and bandwidth. A 30-day transitional window is tight for large estates without preparation.
- Skills and testing. Nobody can execute a switch they have never rehearsed.
The Data Act gives you the right to leave. Open formats, documented pipelines and a rehearsed exit plan give you the ability to leave. Only the second turns regulation into real negotiating power.
The Data Migration Readiness Checklist
Use this checklist to assess each major platform or SaaS service holding important data. Score each item as ready, partially ready or not ready.
1. Contracts and commercial terms
- Contracts have been updated to reflect Data Act switching obligations, including notice, transitional and retrieval periods.
- The contract includes an exhaustive list of exportable data and digital assets, and you have reviewed any exclusions.
- All switching charges, parallel-use data transfer charges and early termination penalties are described, and you know which disappear after 12 January 2027.
- You know whether the provider classes any of your services as custom-built or otherwise exempt.
2. Data inventory and formats
- You maintain an inventory of datasets per platform, with size, owner, sensitivity and business criticality.
- Core analytical data is stored in open formats such as Parquet with Apache Iceberg or Delta Lake, or can be exported to them.
- You have reviewed the provider’s published register of data structures and formats for the services you use.
- Unstructured content, logs and AI assets such as embeddings, prompts and evaluation data are included in the inventory, not just tables.
3. Logic, pipelines and metadata
- Transformation logic lives in version control, preferably in portable SQL or code, rather than only inside a proprietary UI.
- Orchestration definitions, connection configurations and schedules are documented and exportable.
- Catalog metadata, lineage, access policies and data quality rules can be exported in usable formats.
- Dependencies on provider-specific functions and services are listed with candidate replacements.
4. Execution capability
- You have estimated transfer time for your largest datasets given available bandwidth and tools.
- An exit runbook exists with roles, sequencing, validation steps and rollback criteria.
- You have tested an export and reload for at least one representative dataset end to end.
- Reconciliation tests (row counts, aggregates, key metrics) are automated and reusable.
5. Governance and compliance
- Data residency, retention and legal hold requirements are mapped to the target environment.
- You have a process to confirm erasure by the outgoing provider after the retrieval period.
- Exit plans align with other regulatory expectations, such as DORA requirements for exit strategies at financial entities.
A Step-by-Step Path to Exit Readiness
- Prioritize platforms. Rank services by data criticality, spend and strategic risk. Start with the top three.
- Review contracts with legal. Confirm Data Act alignment, list exportable assets and note charges that change in January 2027.
- Adopt open formats where cheap. For lakehouse and warehouse data, prefer open table formats and catalogs with open APIs, so data can be read by other engines without conversion.
- Codify logic. Move transformation and orchestration definitions into version control and reduce dependence on proprietary-only features where there is a portable alternative.
- Rehearse. Run an export-and-reload drill for one dataset and one pipeline into an alternative environment. Measure time, effort and gaps.
- Write the runbook. Turn the drill into a documented exit plan with owners, timelines that fit the Data Act windows and validation steps.
- Review annually. Platforms, contracts and regulations change; refresh the assessment at least once a year and when major contracts renew.
Using Switching Rights Strategically
The end of switching charges does not mean most organizations should switch providers. Migration still costs engineering time and carries risk. The real value is strategic:
- Negotiation leverage. A credible, tested exit plan changes the dynamics of renewal negotiations.
- Multi-cloud and best-of-breed options. Lower exit barriers make it more realistic to run specific workloads, such as AI training or sovereign workloads, on different providers while keeping data in open formats.
- Sovereignty planning. Organizations evaluating EU sovereign cloud offerings can plan moves for sensitive workloads with clearer rules on timelines and charges.
- Parallel running for migrations. Because parallel-use transfer charges must be cost-based after January 2027, phased migrations with dual running become easier to budget.
Architecture Principles That Keep Data Portable
Readiness is cheaper to build into architecture than to retrofit during a switch. Five principles keep data platforms portable without giving up the benefits of managed services.
- Separate storage from compute. Keep analytical data in object storage using open table formats, so engines can change without rewriting data.
- Choose catalogs with open APIs. Catalogs that implement open specifications, such as the Apache Iceberg REST catalog, let other engines and future platforms read governed tables directly.
- Keep logic as code. Express transformations in SQL or general-purpose code managed in Git, with tests, rather than only in proprietary visual tools.
- Wrap provider-specific services. Where you depend on a proprietary service, isolate it behind a clear interface so replacement affects one component rather than every pipeline.
- Make metadata exportable. Prefer governance tools that export lineage, policies and definitions in documented formats, and include metadata exports in backups.
None of these principles prevent you from using managed services aggressively. They simply ensure the most valuable assets, your data, logic and metadata, remain under your control.
Worked Example (Hypothetical)
A hypothetical EU-based manufacturer runs its analytics on one hyperscaler’s data warehouse, uses a SaaS CRM, and stores IoT sensor data in a proprietary time-series service. Ahead of a renewal in early 2027, the CIO asks whether the company could realistically move any of these.
The assessment finds the warehouse data could be exported to Parquet, but about 40% of transformation logic sits in proprietary stored procedures and a visual ETL tool. The CRM contract has been updated for the Data Act and lists exportable data clearly. The time-series service exports only in a vendor-specific format, and nobody has tested a reload. The team prioritizes three actions: migrating new transformations to version-controlled SQL running on an open table format, rehearsing an export-and-reload of one year of sensor data to Iceberg, and writing an exit runbook that fits the 30-day transitional window. Six months later, the company enters renewal talks with a tested plan rather than a theoretical right.
Outlook: 2027–2030
After 12 January 2027, expect providers to compete more on interoperability, open formats and migration tooling, since lock-in through fees is no longer available in the EU. Expect regulators to refine guidance, as the ACM consultation shows, and the Digital Omnibus to adjust scope. Data teams that standardize on open table formats, portable logic and exportable metadata will find switching, multi-cloud and sovereign deployments steadily easier. Those that rely on proprietary features will find the legal right to leave much harder to use.
Common Pitfalls
- Confusing compliance with readiness. A compliant contract does not mean your team can move the data, logic and metadata within the windows the Act defines.
- Ignoring SaaS. Exit planning often focuses on infrastructure while critical data sits in SaaS applications with limited export options. Include them in the inventory.
- Forgetting AI assets. Embeddings, fine-tuned models, prompt libraries and evaluation datasets are increasingly valuable and frequently left out of export lists.
- Never rehearsing. Untested exit plans almost always underestimate transfer time and dependency work.
Conclusion
The EU Data Act removes many of the contractual and financial barriers to switching cloud and SaaS providers, culminating in the end of switching charges on 12 January 2027. Turning that right into real flexibility is a data engineering task: inventory your data, store it in open formats, codify logic and metadata, and rehearse an exit that fits the Act’s timelines. Start with your most critical platforms now, and you will enter 2027 with options rather than obligations.
How Hanoi Technology Can Help
Hanoi Technology assesses exit readiness, moves data to open formats and rehearses cloud migrations through projects, dedicated teams or managed services.