Alternative UE-uniquement à Fly.io.
Fly.io ("Fly") is a US-headquartered edge compute platform that runs Firecracker microVMs in 30+ regions including Amsterdam, Frankfurt, Paris, Madrid and Stockholm. Fly Inc. is a Delaware corporation, the EU regions are EU-located but US-controlled, and the CLOUD Act applies. Fly's technical approach (microVMs at the edge, near-instant cold start, simple `fly deploy`) is genuinely innovative; replacing it with a sovereign EU stack means trading that specific multi-region edge model for either a region-fixed deployment or a self-managed equivalent on EU infrastructure.
Une "région UE" n'est pas la souveraineté. Quatre questions tranchent.
La résidence des données indique où sont les bits. La souveraineté indique quel système juridique peut contraindre à l'accès. La réponse doit tenir sur les quatre points — sinon la stack n'est pas souveraine.
Où les données sont-elles physiquement stockées ?
Pas "dans le cloud" — quel datacenter, dans quel pays, sous quelle juridiction.
Qui d'autre est dans votre chemin de données ?
Chaque fournisseur qui touche les données : le CDN, le relais e-mail, le tracker d'erreurs, le pipeline analytics.
Quelles lois peuvent contraindre à la divulgation ?
Un fournisseur dont le siège est aux États-Unis relève du FISA 702 et du CLOUD Act — même lorsque les données se trouvent à Francfort.
Qui détient réellement les clés de chiffrement ?
Si le fournisseur cloud détient à la fois les données et les clés, il peut les lire — quel que soit le DPA.
Échoue sur la juridiction et la garde des clés.
Bits en UE, maison mère américaine, sous-traitants américains dans le chemin par défaut, clés gérées par le fournisseur.
Réussit sur les quatre.
Hébergé en UE sur une infrastructure au siège européen. Zéro sous-traitant américain dans le chemin par défaut. Clés détenues par le client ou par un KMS européen. Nommés dans votre DPA Article 28.
Pourquoi les équipes partent Fly.io
Fly.io exits we have scoped come from regulated workloads (healthcare SaaS, fintech) where the multi-region edge pattern was nice-to-have but the US-jurisdictional processor was a blocker. The honest answer for these workloads: most don't actually need 30 regions, they need 2-3 EU regions with low latency. That requirement is met by Hetzner Falkenstein + Hetzner Helsinki, or OVH Roubaix + Frankfurt, with a CDN like Bunny.net for static assets — which collectively serves EU users with sub-50ms latency and full EU jurisdiction.
Fly.io services et leurs équivalents UE-uniquement
Une migration n'est pas "échanger une boîte contre une autre". La cartographie ci-dessous est ce que nous exécutons pour les clients qui quittent Fly.io pour des raisons Schrems II — pleine juridiction UE, pas de maison mère US dans le chemin des données.
| Fly.io service | Alternative UE-uniquement | Note d'ingénierie |
|---|---|---|
| Fly Machines (microVMs) | Hetzner Cloud, OVH Public Cloud, Scaleway, self-hosted Firecracker on EU bare metal | For most workloads, regular VMs with multi-region deployment via DNS GeoIP cover the use case. For true microVM-per-request, self-hosted Firecracker on EU compute is the sovereign answer. |
| Fly Apps (PaaS layer) | Coolify on Hetzner with multi-server clustering, Scaleway Serverless Containers, Dokku | Coolify's multi-server feature handles multi-region deployment patterns. |
| Fly Postgres (clustered) | OVH Managed PostgreSQL with replicas, Aiven multi-region EU, self-managed Patroni cluster | Patroni on EU compute is the open-source pattern that powers Fly's own Postgres offering. |
| Fly Redis (Upstash) | OVH Managed Redis, Aiven Redis, Dragonfly self-hosted | Note: Upstash itself is US-headquartered, so Fly Redis is US-on-US. |
| Fly Volumes | Hetzner Volumes, OVH Block Storage, Scaleway Block Storage | Standard NVMe-backed volumes; size-equivalent. |
| Fly Proxy (Anycast) | Bunny.net + EU origin, Cloudflare → Bunny migration first, self-managed Anycast on EU IXP peering | Bunny.net offers Anycast routing across EU; for true global Anycast on EU jurisdiction, self-managed via BGP at an IXP is the path. |
| Fly Postgres failover (multi-region) | Patroni multi-region on EU compute, OVH Multi-AZ Postgres | For EU-only multi-region (e.g. NL + DE active-active with regional failover), Patroni handles it. |
| Fly Secrets | Hashicorp Vault on EU infra, Coolify environment variables, Doppler self-hosted | Vault is the production-grade answer for any non-trivial secrets workload. |
| flyctl / fly deploy DX | Coolify CLI, GitLab CI deploy steps, custom Docker push to Scaleway | The DX gap is real but closeable. Coolify's `coolify deploy` is the closest equivalent. |
| Fly LiteFS (replicated SQLite) | Self-hosted LiteFS on EU compute, rqlite, dqlite | LiteFS is open-source; self-host on Hetzner with multi-region replication. |
Comment nous migrons depuis Fly.io
Une migration typique de mid-market se déroule en trois phases. Les chiffres ci-dessous supposent une équipe d'ingénierie de 6 à 10 personnes et une stack applicative modérément complexe.
Region strategy decision
Audit which Fly regions you actually use and which ones serve real traffic. For most EU-customer-facing apps, 2-3 EU regions cover it; for global apps, decide on the multi-region pattern (DNS GeoIP, Anycast, regional failover).
Database + storage migration
Fly Postgres replicated to EU managed PostgreSQL or Patroni cluster. Volumes mirrored. Secrets moved to Vault.
App cutover
Apps redeployed on Coolify or Scaleway. DNS migrated to GeoIP routing if multi-region needed. Fly account decommissioned after verification window.
5-year TCO on Fly exits varies more than other US-cloud exits because Fly's pricing model is unusual (per-second microVM billing). For steady-state workloads, Hetzner is dramatically cheaper. For very spiky workloads with long idle periods, Fly's scale-to-zero is hard to match cost-effectively in the EU sovereign space without serverless options like Scaleway Serverless Containers.
Questions fréquemment posées
Fly's "no surveillance" marketing — does it actually mean sovereignty?
Fly.io has been transparent about not being on the US hyperscaler critical-infrastructure surveillance lists. That's an operational claim, not a jurisdictional one. As a US Delaware corporation, Fly is subject to the CLOUD Act regardless of how they market themselves. The legal exposure is the same as any other US-jurisdictional provider.
How do we replace the "microVM cold start" feature?
For most workloads, microVM cold start is a nice-to-have rather than a hard requirement. If you genuinely need it, self-hosted Firecracker on EU compute (the same technology Fly uses) is the path. We deploy this for clients with specific microVM needs.
What about LiteFS for SQLite at the edge?
LiteFS is open-source. Self-host on EU compute with multi-region replication. The migration is mostly mechanical because LiteFS uses standard SQLite under the hood.
How long does a Fly.io exit take?
For a typical workload (a few apps, a Postgres cluster, some volumes): 2–4 weeks elapsed. For multi-region setups with Anycast and LiteFS: 4–8 weeks. The biggest schedule risk is replicating the multi-region pattern, not the technical work itself.
Are there any genuinely Fly-like EU options?
Not yet a 1:1 match in the EU sovereign space. The closest combination is Coolify multi-server + DNS GeoIP for routing + EU managed Postgres. It's not as polished as Fly's integrated experience, but it's sovereign and operationally simpler at the cost of some DX.
What about Fly's GPU offering?
Fly's GPU instances (A10, L40S) compete in inference workloads. Scaleway H100 instances are the EU sovereign alternative for current-generation GPU compute. For older generations, Hetzner has competitive pricing.
Planifiez votre sortie de Fly.io.
Appel de cadrage de 30 minutes. Nous cartographions votre stack par rapport aux alternatives UE-uniquement, estimons l'effort de migration et vous disons si c'est le bon choix.