The Data Act Ends Switching Charges. Exit Costs Remain.

Switching charges end on January 12, 2027, for the data processing services that Regulation (EU) 2023/2854 covers. Exit cost does not end with them. The Data Act changes what a provider may charge for one flow, the switch. It does not change what it costs to stay able to move.

Most exit models carry provider egress as a single line, the shape Exit Cost Architecture describes alongside reconstruction and time. After January 12 that line splits into flows with different legal treatment, and a model that zeroes the whole line understates exit cost wherever the movement is not a switch. This post sorts the flows and adjusts the model. It does not explain Data Act compliance, and it is not legal advice.

What the Data Act Actually Prohibits

For a Cloud Strategy team, the useful question is which transfers the rule reaches. Two definitions carry it. Data egress charges are transfer fees for extracting data from one provider’s infrastructure to a different provider or to on-premises infrastructure (Article 2(35)). Switching charges are charges, other than standard service fees or early termination penalties, that a provider imposes for the switching actions the Act prescribes, and they include data egress charges (Article 2(36)).

Article 29 sets the timetable. From January 11, 2024, providers could impose reduced switching charges, capped at the costs directly linked to the switching process concerned. From January 12, 2027, Article 29(1) allows no switching charges at all. The prohibition attaches to the purpose of the transfer, not to the word egress. This table sorts the flows an exit model has to price.

Flow	Treatment under the Act	Basis
Data extracted to a different provider or to on-premises to complete a switch	Switching charge: reduced and cost-capped until January 12, 2027, none from that date	Arts. 2(35), 2(36), 29
Provider support actions the Act requires for the switch	Same as above	Arts. 2(36), 29
Support beyond what the Act requires	Outside the definition; governed by what you agreed in advance	Art. 2(36); contract
Egress for in-parallel use of a second provider’s service	Charge permitted only to pass on egress costs incurred; no end date	Art. 34(2), Recital 99
Operational movement that is not a switch: ordinary egress, inter-region transfer, failover, restore	Outside the switching-charge prohibition on the Act’s definitions; treatment follows the purpose of the transfer and your contract	Arts. 2(35), 2(36)
Standard service fees	Excluded from the definition of switching charges	Art. 2(36)
Early termination penalties	Excluded from the definition; whether a given penalty is proportionate or enforceable is a separate question	Art. 2(36), Recital 89

Parallel use is the row architects misread. Recital 99 separates the one-off egress a switch requires from the ongoing egress of running two services together. Article 34(2) lets a provider pass on egress costs for the second, at cost and without margin, and that treatment has no end date. It is a separate charge treatment. A charge’s label does not decide its category; the purpose of the transfer does. A multi-cloud design does not become free on January 12.

The second misread is the middle row. Sitting outside the switching-charge prohibition does not make every such charge valid or enforceable. It means this prohibition does not settle the question, and your contract and other law still do.

Where the Prohibition Does Not Apply

Article 31(1) takes Article 23(d), Article 29 and Article 30(1) and (3) out for data processing services whose majority of main features have been custom-built for an individual customer, or whose components were all developed for that customer, and which are not offered at broad commercial scale via the provider’s service catalogue. The exclusion turns on those stated conditions, so check whether your service meets them rather than assuming that customization does. It is not a full exemption from the switching chapter, and a provider relying on it should tell you before you contract which obligations do not apply.

Separately, the Commission’s November 2025 Digital Omnibus proposal, which commentary says would add relief for some custom-made services and some smaller providers, remains under negotiation and is not current law. The dates and obligations described here are those in the Data Act as currently in force.

Supervisors are not uniform either. The Dutch ACM published draft cloud guidance on October 19 for providers within its remit, and France’s Arcep issued draft guidelines in February on which costs may count toward switching other than egress, and separately on egress fees for multi-cloud customers. Neither is a universal interpretation.

Switching Charges Cover One Flow

An exit model usually names three ways exit gets expensive: a DR plan that cannot execute, a ransomware restore, and a multi-cloud design whose data layer sits with one provider. Run each through the table.

DR failover. A failover between one provider’s regions is ordinary operation, not a switch. A standing second-provider standby may look more like in-parallel use. Neither is a switch unless you are leaving, and neither is reached by the prohibition for that reason alone. Multi-Cloud Failover Is Mostly Theater covers why the data layer decides whether that failover works at all.

Ransomware restore. Restoring from backup is movement for recovery. It is a switch only if the restore is the last step of leaving the provider. Its cost follows your contract and your backup tier, the territory of Ransomware Recovery Time Is an Architecture Problem.

Multi-cloud operation. A second provider used alongside the first is in-parallel use: egress at cost, with no end date.

The classification questions repeat. Is the data leaving to complete a switch? Is the destination a second provider used alongside the first? Is the movement one-off or ongoing? What does the contract say? The Act answers the first question for the flow it names. The rest stay with your contract and your architecture.

What Stays in the Exit Cost Model

Exit Cost Architecture models exit as data movement, reconstruction and time. The Act touches the first, and only the part a provider levies for a switch. Reconstruction, the work of rebuilding the environment on the other side, and time, meaning transfer duration, sequencing and downtime, were never on the provider’s invoice. No January date removes them. Cloud exit validation asks whether the organization survives the move, and a cheaper switch does not answer that.

After January 12, these lines stay in the model:

• Egress for in-parallel use, at cost.
• Ordinary egress and inter-region transfer, per contract.
• Standard service fees and any early termination penalty.
• Reconstruction of the environment on the destination.
• Transfer time, sequencing and downtime.

Verify the Contract Before You Re-Baseline

Re-baselining starts from the contract, not from this post. Check these five items against the agreement and the order form.

Two limits apply. The article references in this post come from hosted reproductions of Regulation (EU) 2023/2854, not from a reading of the Official Journal, so check the wording there before you rely on it. Commentary also cites a Commission FAQ for the parallel-use treatment; this post does not rely on it. None of this is legal advice.

Additional Resources

View 2 more resources ↓

>_ External Reference

ACM: guidance on switching between cloud services

the Dutch regulator’s October 19 draft guidance announcement, in Dutch.

>_ External Reference

Arcep: cloud computing, February 2026

the French regulator’s draft guidelines on switching-charge costs other than egress.