64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.
Recommended
Documented guidance
KVM VPS or Dedicated Bare Metal with at least 4 dedicated vCPUs and 8 GB RAM.Machine learning indexing and background CLIP analysis are CPU-intensive.
Fast NVMe storage for database and thumbnails; scalable block storage or object storage for raw photo/video assets.Thumbnails generate hundreds of thousands of small file IOPS.
Capacity depends on your workload
Immich runs multiple microservices: Server, Microservices worker, Machine Learning container, PostgreSQL with pgvector, and Redis. Idle memory footprint is approximately 1.5–2.0 GB, scaling up to 4–6 GB during heavy bulk ingestion.
We only compare what providers publish. A RAM figure alone can't confirm compatibility or production capacity.
Proxmox LXC container (with Docker inside or native mounts)
The host must provide
Runtime
Docker Engine
>= 24.0
PostgreSQL with pgvector
>= 14 (vecto.rs/pgvecto.rs)
Redis
>= 7.0
Storage
Persistent data
Plan for two separate storage zones: (1) High-speed NVMe storage for PostgreSQL database, vector embeddings, and generated WebP thumbnails; (2) Scalable bulk storage for original RAW/JPEG files and MP4/MOV videos. Back up the PostgreSQL database and UPLOAD_LOCATION synchronously.
Step by step
Production deployment with Docker Compose
Download official configuration files
Create a dedicated directory /opt/immich, then download docker-compose.yml and example.env from the official Immich release repository.
Copy example.env to .env. Set UPLOAD_LOCATION to your designated persistent data disk, configure DB_PASSWORD with a strong random string, and verify IMMICH_VERSION.
Run docker compose up -d. Verify all containers (server, worker, redis, database, machine learning) are healthy via docker compose ps. Place Nginx or Caddy in front with SSL/TLS and client_max_body_size set to 50000M.
Never expose the raw port 2283 directly to the public internet without an authenticated reverse proxy (Nginx, Caddy, Cloudflare Tunnel, or Tailscale).
Enable HTTPS using Let's Encrypt with TLS 1.3 to protect mobile token transmission.
Configure password policies and OAuth/OIDC authentication if integrating with Authentik or Keycloak.
Cost
Beyond the server price
Compute cost: 4 vCPU / 8 GB RAM VPS starts at approximately $15–$25/month.
Storage cost: High-capacity NVMe or block storage costs between $0.05 and $0.10 per GB/month; consider Hetzner Storage Boxes or S3-compatible cold tiers for long-term archives.
Backup cost: Secondary offsite backup storage for large multi-terabyte photo libraries.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
Meeting documented runtime prerequisites does not prove production capacity. Traffic, workloads, software configuration, and shared host limits are not measured here.
64-bit Linux host (Ubuntu 22.04/24.04 or Debian 12 recommended) with Docker Engine and Docker Compose v2.This prerequisite has no comparable published plan field.
x86_64 CPU with AVX/AVX2 instruction set support (or specialized CPU fallback build for older processors).This prerequisite has no comparable published plan field.
Minimum 4 GB RAM (8 GB+ strongly recommended if facial recognition and video transcoding run concurrently).This prerequisite has no comparable published plan field.
Persistent local SSD/NVMe volume or high-throughput NFS/ZFS array for photo storage and PostgreSQL database.This prerequisite has no comparable published plan field.
