Every Plex admin eventually considers moving to the cloud. The appeal is obvious: unlimited upload bandwidth, no ISP throttling, no hardware to maintain, no UPS to buy when the power flickers. But cloud Plex has its own set of brutal trade-offs — namely, the cost of storing terabytes of media and the GPU compute needed for transcoding. This guide breaks down the real-world costs and trade-offs of running Plex entirely in the cloud, keeping it self-hosted, or taking a hybrid approach.
The Full Cloud Model
A “full cloud” Plex server means everything — media storage, Plex application, and transcoding — runs on rented infrastructure in a data center. You access it remotely to manage your library, and your users stream from the data center’s network rather than your home connection.
The three cost pillars are storage, compute (CPU/GPU), and bandwidth. Let’s look at each one across the most popular providers used by the Plex community.
Storage Costs: The Big Number
A modest Plex library is 5–10 TB. A serious collector has 20–50 TB. A library with 4K HDR remuxes can easily exceed 100 TB. Storage is usually the largest ongoing cost of a cloud Plex server.
| Provider | Storage Cost | 10 TB / mo | 50 TB / mo | Notes |
|---|---|---|---|---|
| Hetzner Storage Box | ~$3.50/TB/mo | $35 | $175 | EU-only, NFS/CIFS/SFTP |
| OVH (Advance) | ~$5/TB/mo | $50 | $250 | Bare metal w/ disk |
| Scaleway (Block) | ~$8/TB/mo | $80 | $400 | EU, block storage SSD |
| AWS S3 (Standard) | ~$23/TB/mo | $230 | $1,150 | + egress fees |
| Self-hosted (HDDs) | ~$1.50/TB/mo* | $15 | $75 | *Amortized over 3yr |
The math is stark. Self-hosted storage costs roughly $15–20 per TB as a one-time expense (a shucked 18 TB WD drive runs about $200–250), amortized to about $1.50/TB/month over a 3-year lifespan. Cloud storage starts at $3.50/TB/month on the cheapest European providers and climbs to $23/TB/month on AWS. For a 50 TB library, the annual difference between self-hosted and Hetzner is $1,200. Against AWS, it’s over $12,000.
Compute Costs: CPU vs. GPU
Plex transcoding requires compute power. On a cloud server, you have two options: CPU-based transcoding (no GPU) or renting a GPU instance.
CPU-only cloud servers are cheap. A Hetzner AX41-NVMe (Ryzen 5 3600, 64 GB RAM) costs about $45/month and can handle 2–3 simultaneous 1080p software transcodes. For many small servers, that’s sufficient. But add 4K content, HDR, or more than 3 concurrent users, and you need GPU compute.
GPU cloud instances are expensive. A single NVIDIA T4 on AWS (the closest equivalent to a T600 for Plex) costs $0.526/hour for a g4dn.xlarge instance — about $380/month running 24/7. Hetzner doesn’t offer GPU instances. Scaleway offers L4 instances starting around $200/month. Even the cheapest GPU instances cost more per month than buying a T400 outright.
Bandwidth: The Hidden Cost
Most cloud providers include generous or “unmetered” bandwidth on dedicated servers and VPS instances. Hetzner includes 20 TB/month of outbound traffic. OVH includes “unmetered” bandwidth on their dedicated range (though they’ll throttle sustained multi-gigabit usage). This is the primary advantage of cloud Plex — your users stream from a datacenter with a 1 Gbps+ connection, not your home upload.
AWS is the exception. EC2 egress costs $0.09/GB after the first 100 GB/month. Streaming a 4K movie (~30 GB) to one user costs $2.70 in bandwidth alone. Three users watching different movies in an evening = $8 in AWS egress. Over a month with moderate usage, bandwidth costs can exceed the compute cost of the instance itself.
Pros of Full Cloud Plex
- Unlimited upload bandwidth. No ISP bottleneck. Every user gets the full bitrate the server can encode.
- Always-on reliability. Datacenter uptime beats a home server behind a consumer router on a shared circuit.
- No home hardware. No noise, no heat, no UPS, no maintenance. If a drive fails, the provider swaps it (on dedicated servers).
- Geographic flexibility. Choose a datacenter close to your users for lower latency.
- No ISP scrutiny. No warnings about “running a server on a residential connection.”
Cons of Full Cloud Plex
- Storage cost is ongoing. You pay every month forever. A 30 TB library on Hetzner costs $105/month ($1,260/year) vs. a one-time $750 for self-hosted drives.
- Uploading media is slow. Getting 30 TB of existing content to a cloud server over a 10 Mbps upload takes roughly 280 days. Even with 100 Mbps upload, it takes 28 days. Most users ship physical drives to the datacenter.
- GPU instances are prohibitively expensive. $200–400/month for a single GPU instance makes hardware transcoding uneconomical at cloud scale for personal use.
- Vendor lock-in. Your media sits on someone else’s servers. Moving providers means re-uploading everything or paying egress fees.
- Complexity. Managing remote Linux servers, VPN tunnels, rclone mounts, Docker containers, and Plex configuration remotely requires significant technical knowledge.
Self-Hosted: The Baseline
For comparison, a well-equipped self-hosted Plex server looks like this:
- Used Dell Optiplex SFF: $100–150
- 4× 18 TB shucked drives (72 TB raw): $800–1,000
- NVIDIA T400 for transcoding: $80–100
- Electricity (~100W 24/7): $8–12/month
Total upfront: $1,000–1,250. Monthly ongoing: $8–12 for power. Amortized over 3 years, that’s roughly $35–45/month total cost of ownership — less than a single Hetzner dedicated server with equivalent storage, and a fraction of any GPU cloud instance.
The trade-off: your upload bandwidth caps concurrent remote streams, your ISP might object, and hardware failures are your problem.
The Hybrid Approach: Keep Storage Local, Offload Compute
The hybrid model combines the strengths of self-hosted and cloud. You keep your media on local drives (cheap, fast, no monthly storage fees) and offload just the expensive part — transcoding and delivery — to remote compute.
This is exactly what PlexBeam does. Your Plex server runs at home with your local storage. When a remote user starts watching and needs a transcode, PlexBeam routes the file data to a GPU worker in a datacenter. The worker transcodes the stream and delivers it directly to the user from the datacenter’s connection. Your home upload only carries the raw file data to the worker — not every user’s individual stream.
| Aspect | Self-Hosted | Full Cloud | Hybrid (PlexBeam) |
|---|---|---|---|
| Storage | Local (cheap) | Cloud ($3–23/TB/mo) | Local (cheap) |
| Transcoding | Local GPU/CPU | Cloud CPU (or expensive GPU) | Remote GPU worker |
| Upload bandwidth | Home ISP (limited) | Datacenter (fast) | Datacenter delivery |
| Monthly cost (50 TB) | $10–15 | $175–500+ | $10–15 + PlexBeam sub |
| Hardware maintenance | You | Provider | Local drives only |
| Setup complexity | Low | High | Low (plugin/cartridge) |
When Full Cloud Makes Sense
Full cloud Plex is the right choice in a narrow set of circumstances:
- Your home connection has extremely poor upload (under 5 Mbps) with no fiber option coming
- Your ISP actively blocks or throttles server traffic
- You live in a location with unreliable power or internet
- Your library is small (under 5 TB) and you can afford cheap European hosting
- You want zero physical hardware to manage
For everyone else — and that’s most Plex admins — the economics of local storage plus cloud transcoding (the hybrid model) are significantly better. You keep the cost advantage of buying hard drives once, avoid the complexity of managing a remote Linux server, and still get the bandwidth and transcoding benefits of datacenter GPU hardware.
The rclone Mount Problem
Many cloud Plex setups use rclone to mount cloud object storage (Google Drive, Dropbox, or Backblaze B2) as a local filesystem on a cheap VPS. This technically works — Plex sees the rclone mount as a local drive — but the performance is terrible for media serving. Random seeks for subtitle extraction, library scanning, and metadata generation hammer the cloud storage API with thousands of small requests. Playback often stalls for 5–10 seconds during seek operations as rclone buffers the new file position.
Google’s crackdown on API abuse (rate limiting heavy rclone users and revoking API keys) has made this approach increasingly unreliable. The Plex community still uses it, but the golden age of unlimited Google Drive storage mounted via rclone ended in 2023 when Google Workspace started enforcing storage caps.
Data Transfer: The Initial Upload Problem
Even if you decide cloud is right for you, getting your media there is a logistical challenge. A 30 TB library at 20 Mbps upload takes 139 days of continuous transfer. At 100 Mbps, it still takes 28 days. Some providers (Hetzner, OVH) accept shipped hard drives for a fee, but the process is manual and takes 1–2 weeks.
With the hybrid PlexBeam model, this problem disappears entirely. Your media stays on your local drives. There’s no bulk upload, no rsync jobs, no waiting. New content is available to PlexBeam workers the moment Plex scans it into your library. The only data that travels to the cloud is the active stream being transcoded, and that happens on-demand in real time.
The Bottom Line
Full cloud Plex costs $100–500+/month depending on library size and provider, requires significant technical effort to set up and maintain, and still can’t match the transcoding performance of local GPU hardware without spending $200–400/month on a GPU instance. Self-hosted Plex costs $35–45/month amortized but is limited by your home upload bandwidth. The hybrid approach — local storage plus remote GPU transcoding — gives you the best of both worlds at a fraction of the full-cloud cost.