Ask a room of developers what GDPR says about where data has to live and many will answer: in the EU. That is not what the regulation says. GDPR does not contain a rule that personal data must stay inside the European Union. What it does contain is a set of conditions on transferring personal data to third countries, and those conditions are strict enough that keeping data in-region is often the simplest way to satisfy them.
This post covers what the primary texts require, what the courts said about EU-US transfers, what the EU Data Act added, and how to pin workloads to a region on common platforms. This is an engineering guide, not legal advice, so have counsel review any decision that depends on it.
What GDPR Actually Restricts
Chapter V of GDPR governs transfers. Article 44 sets the general principle: any transfer of personal data to a third country or an international organisation may take place only if the conditions in that chapter are met, including for onward transfers from that third country to another one [1].
The chapter then gives you three routes. Under Article 45, a transfer may take place where the European Commission has decided that the destination country ensures an adequate level of protection [1]. Under Article 46, absent an adequacy decision, a controller or processor may transfer data only if it has provided appropriate safeguards and data subjects have enforceable rights and effective legal remedies [1]. Article 46(2) lists those safeguards, including binding corporate rules and standard data protection clauses adopted by the Commission [1]. Article 49 lists narrow derogations for when neither applies [1].
So the question is not where the database lives but which data leaves the EEA, to whom, and under which route. A Frankfurt-hosted database queried by a support engineer in another country can still involve a transfer. The EDPB says so explicitly: remote access from a third country, for example in support situations, and storage in a cloud outside the EEA are both considered transfers [2].
Schrems II, the Data Privacy Framework, and Latombe
The EU-US story is why this topic keeps resurfacing. On 16 July 2020, in Case C-311/18 (Schrems II), the Court of Justice declared invalid Decision 2016/1250, the adequacy decision behind the EU-US Privacy Shield [3]. In the same judgment it found nothing affecting the validity of Decision 2010/87 on standard contractual clauses [3]. Contracts survived, but signing clauses was no longer the end of the analysis.
The EDPB's Recommendations 01/2020 on supplementary measures, adopted in final form as version 2.0 on 18 June 2021, set out a six-step roadmap: know your transfers, identify the transfer tools you rely on, assess whether the Article 46 tool is effective in light of the destination's laws and practices, adopt supplementary measures, take the procedural steps, and re-evaluate at appropriate intervals [2]. Encrypted backups stored with a provider in a third country can be an effective supplementary measure if the encryption is state of the art and the keys are retained solely under the control of the exporter or a trusted entity in the EEA or an equivalent jurisdiction [2]. But where a cloud provider needs the data in the clear to do its job and the destination's surveillance powers go beyond what is necessary and proportionate, the EDPB says it cannot envision an effective technical measure, and that transport plus at-rest encryption do not suffice if the importer holds the keys [2].
On 10 July 2023 the Commission adopted a new adequacy decision for the EU-US Data Privacy Framework [4]. It covers transfers to companies in the United States that participate in the framework, and includes a redress mechanism for Europeans who complain about US intelligence collection [5]. On 3 September 2025, in Latombe v Commission (T-553/23), the General Court dismissed the action in its entirety, finding that on the date of adoption the US ensured an adequate level of protection for data transferred to organisations there [6]. It rejected the claim that the Data Protection Review Court lacks independence and held that Schrems II does not require bulk collection to be authorised in advance by an independent authority, only that it be subject to ex post judicial review at a minimum [6].
Two engineering implications follow. First, the framework only covers US recipients that participate in it, so it does not automatically cover every vendor in your stack [5]. Second, the General Court itself noted that the Commission must monitor the US legal framework continuously and may suspend, amend or repeal the decision if it changes, and the judgment could be appealed to the Court of Justice on points of law [6]. Two earlier US adequacy decisions were invalidated [6], so a regional fallback is cheap insurance.
The EU Data Act: Non-Personal Data and Switching
The EU Data Act, Regulation (EU) 2023/2854, applies from 12 September 2025 [7]. Unlike GDPR, it is not limited to personal data.
Article 32 requires providers of data processing services to take all adequate technical, organisational and legal measures, including contracts, to prevent international governmental access to or transfer of non-personal data held in the Union where that would conflict with EU or national law [7]. A third-country court or administrative decision demanding such data is recognised or enforceable only if based on an international agreement, such as a mutual legal assistance treaty [7]. Article 28 requires providers to publish, on their websites, the jurisdiction governing the ICT infrastructure for each service and a general description of the measures they use against unlawful governmental access [7].
The switching provisions are the other half. Article 23 obliges providers to remove commercial, technical, contractual and organisational obstacles to switching [7]. Article 25 sets a maximum transition period of 30 calendar days, extendable to up to seven months where the provider shows 30 days is technically unfeasible [7]. Under Article 29, from 12 January 2027 providers may not impose any switching charges at all; until then, charges are capped at direct costs [7]. That lowers the cost of keeping a credible second provider.
Residency Is Not Sovereignty
Data residency means bytes sit on disks in a chosen geography. Sovereignty is about whose law can compel access to them, and residency alone does not answer that. A provider headquartered outside the EU running a Paris region still has to deal with its home jurisdiction, which is exactly the scenario the EDPB's analysis and Data Act Article 32 are built around [2] [7]. Key custody, who operates the system, and subprocessor contracts close the gap that region pinning leaves.
Data also moves where nobody drew it: logs shipped to observability vendors, backups replicated cross-region, CDN edges caching responses globally, support engineers connecting from anywhere. Subprocessors call their own subprocessors, which the EDPB tells exporters to include when mapping transfers [2].
Pinning Workloads and Data on Common Platforms
Vercel runs functions in Washington, D.C. (iad1) by default for all new projects; you change that with the regions key in vercel.json, the dashboard, or the CLI's --regions flag, with per-function overrides [8]. Its own example pins an api/eu-data.js function to Paris (cdg1) with failover to London (lhr1) [8]. Note that London is outside the EU, so check failover regions as carefully as primary ones. Vercel also caches static content in its CDN globally by default and deploys Routing Middleware to all regions regardless of your region settings [8].
Cloudflare's Data Localization Suite is a paid Enterprise add-on with three parts: Regional Services, which controls which data centers may decrypt and process HTTPS traffic; Customer Metadata Boundary, for logs and analytics; and Geo Key Manager, for where private keys are stored [9]. Regional Services still accepts traffic at any Cloudflare data center worldwide, forwarding it encrypted to an in-region data center where TLS termination, caching, WAF and Workers run [10]. Cloudflare notes that for single-country regions other than the US there is no automatic failover if in-country capacity is unavailable [10].
On AWS, region is explicit per resource, and Regions introduced after March 20, 2019 must be enabled before use [11]. Of the European regions, eu-west-2 is London and eu-central-2 is Zurich, neither in the EU [11]. On Google Cloud, the gcp.resourceLocations organization policy with the in:eu-locations value group restricts where new resources can be created [12]. Google is clear about its limits: it is not retroactive, does not remediate existing violations, does not cover global resources, and is not a data storage commitment [12].
Supabase's region list includes Frankfurt, Ireland, Paris and Stockholm, and warns that the general Europe option can include London and Zurich, which are not EU member states [13]. Projects are bound to a region at the infrastructure level; changing it means creating a new project and migrating [14].
A Practical Checklist
1. Map every personal data flow, including backups, logs, error tracking, support tooling and subprocessors' onward transfers [2]. 2. For each flow that leaves the EEA, name the Chapter V route: adequacy, Article 46 safeguards, or an Article 49 derogation [1]. 3. For US recipients relying on the Data Privacy Framework, confirm each one actually participates [5]. 4. Pin compute and storage explicitly, including failover regions and backup replication targets [8] [12]. 5. Decide what the CDN may cache and where TLS terminates [10]. 6. Hold encryption keys in-region and under your control wherever the provider does not need plaintext [2]. 7. Restrict and log support access from outside the region [2]. 8. Read each provider's Data Act Article 28 jurisdiction disclosure [7]. 9. Document an exit path to a second region or provider, which the Data Act's switching rules make more realistic [7]. 10. Re-evaluate on a schedule and when the legal landscape shifts [2].
How AW3 Host Fits
AW3 Host is a deployment control plane across 17 providers, including Vercel, AWS, GCP, Azure, Cloudflare, Supabase and Hetzner. It helps with the mechanical parts of residency: deploying the same service to a chosen region on more than one provider, geo routing so EU users hit EU deployments, failover with 60-second health checks so a regional outage has a planned answer, and encrypted secrets. It does not decide which flows are transfers, sign your contractual clauses, control where a provider's support staff sit, or stop a vendor's logging pipeline from replicating elsewhere.
Takeaways
GDPR restricts transfers; it does not mandate EU storage [1]. The EU-US Data Privacy Framework survived its first General Court challenge in September 2025, but it covers only participating companies and remains under review [5] [6]. The Data Act adds obligations around governmental access to non-personal data and makes switching cheaper from January 2027 [7]. Pin the region, then go find the logs, backups, caches and support sessions that quietly leave it.
Sources
- Regulation (EU) 2016/679 (General Data Protection Regulation) — EUR-Lex, 2016
- Recommendations 01/2020 on measures that supplement transfer tools, version 2.0 — European Data Protection Board, 2021
- Press Release No 91/20: Judgment in Case C-311/18 (Schrems II) — Court of Justice of the European Union, 2020
- Adequacy decision for the EU-US Data Privacy Framework — European Commission, 2023
- EU-US data transfers — European Commission, 2023
- Press Release No 106/25: Judgment in Case T-553/23, Latombe v Commission — General Court of the European Union, 2025
- Regulation (EU) 2023/2854 (Data Act) — EUR-Lex, 2023
- Configuring regions for Vercel Functions — Vercel, 2026
- Data Localization Suite — Cloudflare, 2026
- Regional Services — Cloudflare, 2026
- AWS Regions — Amazon Web Services, 2026
- Restricting resource locations — Google Cloud, 2026
- Available regions — Supabase, 2026
- Change Project Region — Supabase, 2026

Writing about cloud infrastructure, DevOps best practices, and the future of application deployment. Follow for weekly insights on multi-cloud strategies and platform engineering.
