Migrating EU workloads now only beats waiting if projected egress fees before January 2027 exceed your one-off migration cost.
This answer changes when your measured monthly egress, multiplied by your provider's current per-TB rate and months until your switching window, stays below that cost.
What the Data Act Actually Requires
The Data Act, Regulation (EU) 2023/2854, is described in the gerrishlegal.com deep read as set to apply from September 2025. For cloud and SaaS providers, it obliges removal of barriers to switching, including egress-related switching charges. Treat the exact application date and any provider-specific effective date as items to verify in your provider’s documentation and contract forms, rather than as assumptions.
Treat the switching-charge phase-out as a contract-specific timetable to verify in your provider’s switching policy and master agreement. The gerrishlegal.com deep read frames the Act as applying from September 2025, but your provider’s phase-down date and contract commencement date determine which phase applies; do not replace those checks with a generic “egress is becoming free” statement.
Charges are phased out on a schedule tied to contract commencement dates, with complete removal expected by the Act’s transition deadline. Your provider’s phase-down date can differ from any headline date, so verify the exact date in your master agreement, order form, or published switching policy before you model any savings.
| Check | Where to verify | Why it matters |
|---|---|---|
| Application date used by provider | Provider terms, switching policy, or account notice | Confirms whether 12 September 2025 or another date governs your account |
| Contract commencement date | Master agreement or order form | Determines which phase of the charge phase-out applies |
| Provider phase-down date | Published switching schedule or support ticket | Prevents assuming free egress before your contract's window |
Use the gerrishlegal.com deep read as the framing source for the Act's switching-rights logic: customers can initiate migration, providers must maintain service during transition, and the contract terminates once migration is complete. Ask your provider to confirm in writing how that process maps to your account, including notice periods and any service-continuity commitments.
Do not treat "free egress" as an automatic entitlement on a calendar date. The practical rule is to match your contract's commencement date to the phase-out schedule, then obtain the provider's written phase-down date before deciding whether waiting is cheaper than migrating.
The $20/TB Claim: Verify, Don't Trust
The widely circulated “$20/TB” figure is a planning shorthand, not a price you can take to a budget meeting. Neither the Data Act text nor the sources reviewed for this guide publishes a statutory or confirmed per-terabyte at-cost egress rate: no such number appears in the regulation or in verified provider disclosures. Figures like $20/TB or $90/TB are illustrative market references only. The Act requires providers to charge no more than the costs directly linked to the switching process; it does not set a tariff, so your real number must come from your own account data.
To find it, run an invoice audit rather than relying on list pricing. Pull egress line items from your provider invoices for the past 90 days, sum the total egress fees, and divide by the total terabytes transferred in the same period. Compare that blended invoice rate against the provider’s published list rate for the same regions and transfer types. The gap between the blended rate and the list rate is your negotiation surface: it reveals discounts, tiering effects, or bundled credits that a headline rate obscures.
| Audit step | What to collect | What it tells you |
|---|---|---|
| Extract line items | 90 days of egress charges by region, service, and transfer type | Where egress spend actually concentrates |
| Calculate blended rate | Total egress fees ÷ total TB transferred | Your effective per-TB cost, not the advertised one |
| Compare to list rate | Provider’s published egress pricing for matching traffic | The discount or markup you can document and challenge |
Use the audit as the basis for a written request to your provider. Ask for an itemized breakdown of the costs directly linked to data transfer and switching—such as network transit, retrieval, and operational handling—and request the methodology used to allocate those costs to your account. Under the Data Act’s cost-reflective charging principle, described in Gerrish Legal’s analysis of the regulation’s switching provisions, providers should be able to justify charges against underlying costs rather than defaulting to standard list pricing.
Treat any quote that cannot be reconciled to invoice data or a documented cost methodology as provisional. Before you commit to a migration plan or a renewed contract, verify that the per-TB figure you model matches the blended rate from your audit and that the provider’s justification is specific enough to compare across alternatives. A verified number, even if imperfect, is a safer input than a market rumor.
Migrate Now vs Wait: The Math
The break-even test is exposure = measured TB per month × verified per-TB egress rate × months to your switching window, compared with the one-off migration cost. Use the earlier of your contract's switching window or the January 2027 phase-out as the month count. If cumulative exposure is greater than the migration cost, migrating now wins; if it is less, waiting wins; if the two are essentially equal, timing alone does not decide the case.
Measure egress before you trust any rate. Instrument 30 days of actual egress for each bucket, convert that total to terabytes per month, and multiply by the rate you can verify from invoices or contract terms rather than a list price. Count months from today to the earlier of your switching window or the January 2027 phase-out; if no switching window exists, use the phase-out date as the outer bound.
| Input | How to obtain it | Role in the test |
|---|---|---|
| Measured egress | 30-day instrument per bucket | TB/month |
| Verified rate | Invoice or contract audit | per TB |
| Months to window | Earlier of switching window or January 2027 phase-out | multiplier |
| Migration cost | Written quote | comparison threshold |
Compare cumulative exposure with a migration quote you actually hold, not a category average. Grounding sources report small migrations at $50,000–$200,000 and enterprise transformations at $500,000 or more (oski.site, Mar 2026), but those ranges are context only; use your vendor's or integrator's written quote, including data transfer, re-platforming, testing, and cutover labor.
If cumulative exposure is below the quoted migration cost, waiting wins on this arithmetic; if it is above, migrating now wins. When the two are close, the tie-breaker is contract terms—notice periods, minimum terms, and whether the provider's rate can change before the window—not the formula. Re-run the calculation whenever measured egress, rates, or quotes change, because a 30-day instrument goes stale if traffic patterns shift.
Options Compared
The four exit paths are not interchangeable; each wins in a different measured band. Choose by instrumented egress, not by a headline rate or a vendor quote.
Stay and wait costs nothing upfront and keeps full exposure until the switching-charge phase-out completes. It can be rational when your measured exposure is small relative to a written migration quote, but that comparison must come from your own 30-day egress measurement and quote rather than a fixed volume band. Negotiate at-cost terms now uses the Data Act’s cost-reflectiveness obligation as leverage; it may be worth pursuing before a renewal when your invoice audit shows a material gap between the blended rate and the provider’s documented cost basis.
Partial migration moves only the datasets with the highest egress multiplier. It wins when 20% of datasets drive 80% of egress, but only after per-bucket measurement; otherwise you risk moving cold data and paying for migration that the phase-out may make free. Full migration is the fastest exit and the right default only when measured exposure already exceeds the quoted one-off migration cost and the switching window is close; otherwise it is the most expensive answer to a problem time may solve.
| Exit path | Winning band | Decision rule |
|---|---|---|
| Stay and wait | Under roughly 10 TB/month | Zero migration cost; exposure is small enough to absorb |
| Negotiate at-cost terms now | Over 100 TB/month with renewal inside 12 months | Use cost-reflectiveness leverage before renewal |
| Partial migration (hot data only) | Any volume where 20% of datasets drive 80% of egress | Move only highest-egress buckets after per-bucket measurement |
| Full migration | Exposure already exceeds quoted one-off migration cost | Fastest exit, but highest upfront cost |
Use published 2026 migration-cost ranges only as a sanity check on the quote: one guide cites small migrations at $50,000-$200,000 and enterprise transformations at $500,000 or more. If your measured exposure does not clear that number before the switching window, waiting or negotiating beats moving everything.
Numbers That Decide It
The decision reduces to three numbers you can measure or obtain, refreshed on a 30-day cycle and compared against the migration-cost benchmark from the prior section.
| Input | How to produce it | Common error to avoid |
|---|---|---|
| Measured egress | 30-day per-bucket totals from provider billing exports | Using provisioned capacity or last year's averages |
| Verified blended rate | Invoice egress fees divided by invoice TB over 90 days, by egress class | Blending intra-region, cross-region, and internet egress |
| Quoted migration cost | Fixed-price quote covering transfer tooling, dual-running storage, and engineering time | Using a planning estimate or an unitemized vendor number |
| Checkpoint cadence | Refresh all three inputs every 30 days and after pricing or workload changes | Letting one input go stale while the others stay current |
First, measure egress rather than capacity. Export 30 days of provider billing data and total egress per bucket or workload. Use the provider's metered quantity and its TB definition; do not substitute provisioned throughput, storage size, or last year's averages, because either can overstate or understate the bytes that actually leave the platform.
Second, compute a verified blended rate from invoices, not list prices. For each egress class—intra-region, cross-region, and internet—divide 90 days of egress fees by 90 days of metered TB. Keep the classes separate; they are different line items with different rates, and blending them into one average hides the class that drives most of the exposure.
Third, get a fixed-price migration quote rather than a planning estimate. It should itemize data transfer tooling, dual-running storage, and engineering time. Published 2026 cost guides place small migrations at $50,000–$200,000 and enterprise transformations at $500,000 or more (oski.site), but use those only as a sanity check against your quote, not as your number.
Refresh all three inputs every 30 days, and immediately after a pricing change, contract amendment, workload migration, or new bucket. The egress measurement, rate calculation, and quote should describe the same workload and the same billing period; otherwise the comparison mixes numbers that never describe the same period.
What the Evidence Does Not Establish
Even when the break-even arithmetic favors waiting, several conditions fall outside what the available evidence can confirm. The switching-charge phase-out reduces one category of cost, but it does not resolve every contractual, technical, or jurisdictional barrier that can keep a workload expensive or immobile. The cases below identify when the wait-for-2027 rule can break.
Measure each bucket separately for 30 days; do not use list prices or planning shorthand. Multiply measured TB by your verified rate, then by months until your contract's switching window. If the total exceeds the migration quote, migration is justified on egress alone; if not, waiting is cheaper unless non-egress factors change the decision. A 30-day sample catches weekly traffic patterns and can be rerun before a quote expires.
Checklist: instrument 30 days per bucket; verify the blended rate from invoices; include every billable egress class; use your actual switching window; rerun after any volume change; keep the worksheet with the quote so finance can see which assumption moved.
What to do next
| Step | Action | Why it matters |
|---|---|---|
| 1 | Instrument 30 days of actual egress per bucket. | Creates the measured TB/month input for the migration decision. |
| 2 | Multiply measured TB/month by your provider's current egress rate and by months until your contract's switching window. | Estimates egress exposure before the January 2027 free-egress phase. |
| 3 | Compare that projection with your quoted one-off migration cost. | The canonical rule migrates only when projected fees exceed migration cost. |
| 4 | If the projection exceeds the quote, start migration now; otherwise wait for January 2027. | EU Data Act free egress from January 2027 under Regulation (EU) 2023/2854 changes the economics. |
| 5 | Re-run the comparison at $20/TB at-cost and $90/TB rates, and at small ($50,000–$200,000) versus enterprise (over $500,000) migration bands. | Rate and migration-cost bands scale total exposure and the threshold. |
| 6 | Confirm the workload is in EU scope and verify the switching window before committing funds. | Keeps the calculation tied to EU workloads and contract timing. |
Frequently Asked Questions
When does migrating EU workloads now make more financial sense than waiting?
Migrating EU workloads now only beats waiting if projected egress fees before January 2027 exceed your one-off migration cost.
How can I calculate whether projected egress fees will stay below my migration cost?
You can determine this by multiplying your measured monthly egress by your provider's current per-TB rate and the months until your switching window to see if it stays below that cost.
What does the EU Data Act require from cloud and SaaS providers regarding switching?
For cloud and SaaS providers, the Data Act obliges the removal of barriers to switching, including egress-related switching charges.
What determines which phase of the switching-charge phase-out applies to my contract?
Your provider’s phase-down date and contract commencement date determine which phase of the switching-charge phase-out applies.
Where should I verify the exact application or provider-specific effective dates?
The exact application date and any provider-specific effective date should be verified in your provider’s documentation and contract forms, rather than treated as assumptions.
Can I assume that egress charges will be free based on a general headline date?
You should not replace checks of your provider’s phase-down date and contract commencement date with a generic “egress is becoming free” statement.
Quick answers
| When does migrating EU workloads now beat waiting? | Migrating EU workloads now only beats waiting if projected egress fees before January 2027 exceed your one-off migration cost. |
| When might waiting be a better option than migrating now? | Waiting may be better when your measured monthly egress, multiplied by your provider's current per-TB rate and months until your switching window, stays below that cost. |
| What does the EU Data Act require from cloud and SaaS providers regarding switching? | The Data Act, Regulation (EU) 2023/2854, obliges cloud and SaaS providers to remove barriers to switching, including egress-related switching charges. |
| How should providers' application and effective dates be treated? | The exact application date and any provider-specific effective date should be treated as items to verify in your provider’s documentation and contract forms, rather than as assumptions. |
| What determines which phase of the switching-charge phase-out applies? | Your provider’s phase-down date and contract commencement date determine which phase applies, so the exact date should be verified in your master agreement. |
Also worth reading: Enforcing data-residency policies at the object-storage layer: measured egress cost ($/TB) and P99 latency overhead of S3 Object Lock + bucket policy vs. gateway-side filtering across AWS, Azure Blob, and GCS: Enforcing data-residency policies at the · Data center cooling: move object-storage racks to direct-to-chip cooling above 30 kW: Data center cooling: move object-storage · 2026 SSE-KMS vs Azure SSE vs GCS CSEK: SOC2 & Latency: 2026 SSE-KMS vs Azure SSE