Docker containers themselves consume negligible power - the cost question is really about what hardware they run on and for how long. A container is just a process. Running five Docker containers on a Synology DS925+ costs the same as running the NAS idle, plus a small CPU and RAM overhead per active container. The meaningful power draw comes from the underlying hardware: NAS models range from 8W idle (ARM entry-level) to 30W+ under load (x86 with spinning drives). This guide gives you the numbers to calculate your actual electricity cost and identify the containers where 24/7 operation is worth it versus where scheduled or on-demand operation makes more sense.
In short: An x86 NAS running Docker 24/7 draws roughly 15-25W under typical multi-container load. At an illustrative electricity usage rate of $0.30/kWh, that works out to $40-65 AUD per year. Entry-level ARM NAS hardware draws 8-12W with containers running, costing $20-30 AUD per year. Container power overhead depends on the processes and workload being run and cannot be estimated reliably from container count alone. Most containers are worth running 24/7; the exceptions are resource-hungry workloads like video transcoding or AI model inference that can be triggered on-demand instead.
Note on electricity rates: Australian electricity rates vary significantly by state and retailer. The figures in this article use $0.30/kWh as a representative rate. Check your latest electricity bill for your actual rate. Western Australian and Tasmanian rates differ notably from NSW and VIC. Calculated annual costs scale linearly: if your rate is $0.35/kWh, multiply the figures below by 1.17.
NAS Hardware Power Draw: The Baseline
Before calculating container costs, understand your hardware baseline. NAS power draw has three components:
- Idle system power: CPU at minimal utilisation, RAM powered, no active disk I/O. This is the floor everything else builds on.
- Drive spin-up and active I/O: Drive power varies by model, capacity and operating state; use the installed drive's datasheet rather than a single HDD or SSD allowance. HDD hibernation time varies widely with services, logging, mapped volumes and background tasks.
- Container CPU/RAM overhead: Most background containers (Nextcloud, Vaultwarden, Home Assistant) draw minimal CPU when idle, adding 1-3W each. CPU-intensive containers (Plex transcoding, Jellyfin, AI inference) can add 10-20W or more during active use.
NAS Hardware Idle Power Draw by Category
| Entry ARM (e.g. DS223J) | Mid x86 (e.g. DS425+) | Performance x86 (e.g. DS925+) | High-end x86 (e.g. DS1525+) | |
|---|---|---|---|---|
| Typical idle (no HDD spin) | Idle draw varies with the exact NAS, installed drives, services and test method; use the manufacturer's stated test figure or measure the configured system at the wall. | Idle draw varies with the exact NAS, installed drives, services and test method; use the manufacturer's stated test figure or measure the configured system at the wall. | Idle draw varies with the exact NAS, installed drives, services and test method; use the manufacturer's stated test figure or measure the configured system at the wall. | Idle draw varies with the exact NAS, installed drives, services and test method; use the manufacturer's stated test figure or measure the configured system at the wall. |
| Idle with 2x HDD spinning | 17-21W | 23-28W | 25-30W | 28-34W |
| Under CPU load (Docker active) | 12-18W | 20-30W | 25-35W | 30-45W |
| Annual cost at idle (no HDD). AUD $0.30/kWh | ~$18-24 | ~$34-42 | ~$39-47 | ~$47-58 |
| Annual cost under load. AUD $0.30/kWh | ~$31-47 | ~$53-79 | ~$66-92 | ~$79-118 |
| Docker support | Yes (ARM; image and package-version limits apply) | Yes (model/version dependent) | Yes (model/version dependent) | Yes (model/version dependent) |
| RAM capacity (max) | 1GB (fixed) | 6GB | 32GB | 32GB |
Power Cost Per Container Type
Not all containers are equal in power terms. The difference is between containers that sit idle most of the time (web servers, password managers, sync servers) versus containers that actively process data (media servers, AI tools, monitoring with frequent polling). The following figures are illustrative only: incremental draw varies with the NAS, storage, application configuration, traffic and measurement method, so verify it at the wall on your own system.
| Nextcloud (AIO or standard) | Idle: 1-2W above baseline. Light usage spikes to 3-5W. Annual cost addition: ~$3-8 AUD. Worth running 24/7: yes. |
|---|---|
| Vaultwarden (Bitwarden-compatible) | Idle: under 0.5W. Essentially invisible in power measurements. Annual cost addition: under $2 AUD. Worth running 24/7: yes. |
| Home Assistant | Idle: 1-3W above baseline (depends on integrations and automations). With many active integrations: 3-6W. Annual cost addition: $3-12 AUD. Worth running 24/7: yes. |
| Tailscale (container mode) | Under 0.5W. Background process only. Annual cost addition: under $2 AUD. Worth running 24/7: yes. |
| Jellyfin / Plex (media server) | Idle (no transcoding): 1-3W. Active software transcoding: 10-25W additional. Hardware transcoding (Intel QSV): 3-8W additional. Annual idle cost addition: $3-8 AUD. Transcoding sessions cost extra. 2-hour movie at 15W software transcode = ~$0.009. Worth running 24/7: yes for idle; transcoding sessions are on-demand anyway. |
| Immich (photo management) | Idle: 1-2W. Machine learning tasks (face recognition, CLIP): 15-30W burst during initial processing, then idle. Annual cost addition (post-setup): $3-8 AUD. Worth running 24/7: yes, ML processing runs on first use then scales back. |
| Uptime Kuma (monitoring) | Under 1W. Minimal CPU, mostly sleeping between checks. Annual cost addition: under $3 AUD. Worth running 24/7: yes. |
| Portainer | Under 0.5W when idle. Annual cost addition: under $2 AUD. Worth running 24/7: yes (management interface only). |
| Ollama / local LLM inference | Idle (model loaded, no queries): 2-5W. During inference: 30-100W+ depending on GPU passthrough or CPU inference. Annual cost for on-demand use: varies significantly. Worth running 24/7 with model loaded: borderline. Consider loading on-demand if used infrequently. |
Calculating Your Own Annual Container Cost
The formula is straightforward:
Annual cost (AUD) = Watts × 8,760 hours ÷ 1,000 × your electricity rate per kWh
Example: A DS425+ running three containers (Nextcloud, Vaultwarden, Home Assistant) with two HDDs in HDD sleep mode most of the time.
- Baseline idle (system + drives parked): 14W
- Container overhead (3 containers): +4W
- Total: 18W
- Annual energy: 18W × 8,760h ÷ 1,000 = 157.7 kWh
- Annual cost at $0.30/kWh: 157.7 × 0.30 = $47.30 AUD per year
Add drive spin-up time (if HDDs spin for database access a few hours per day): +6W for 4 hours = +0.024 kWh/day = +$2.63/year. Effectively rounding error for most setups.
Australian Electricity Rates by State
Approximate Residential Electricity Rates. Australia 2025-26
| NSW/ACT | VIC | QLD | SA | WA | |
|---|---|---|---|---|---|
| Typical usage rate (per kWh) | $0.28-0.35 | $0.28-0.34 | $0.28-0.33 | $0.38-0.50 | $0.28-0.32 |
| Annual cost: 18W 24/7 | $44-55 | $44-54 | $44-52 | $60-79 | $44-50 |
| Annual cost: 25W 24/7 | $61-77 | $61-74 | $61-72 | $83-110 | $61-70 |
| Annual cost: 35W 24/7 | $86-107 | $86-104 | $86-101 | $117-153 | $86-98 |
At the article's stated rates, a 25W continuous setup costs about $83-110/year in SA versus $61-77 in NSW/ACT. For SA users, hardware efficiency is proportionally more valuable. Choosing a lower-draw NAS or enabling HDD sleep aggressively has a larger dollar impact.
When to Run 24/7 vs On-Demand
Most home server containers should run 24/7. The overhead is small, and the operational benefit (availability when you need it, background sync, automations) is real. The containers worth reviewing for on-demand operation are those with high active-state power draw that you use infrequently:
- Local LLM inference (Ollama, etc.): If you query a local model occasionally rather than continuously, loading the model and stopping the service between sessions saves meaningful power. The $11 figure covers inference energy, while the $13 figure covers idle energy only; an always-on total must include both components.
- Video transcoding jobs: Handbrake, FFmpeg, or conversion queues should be run as jobs rather than persistent services. Trigger them via cron or manually, not as always-on services.
- Development environments: Database servers and app stacks spun up for testing should be stopped when not in use. A Postgres container at idle is cheap; a development stack with hot reload and frequent queries is not.
For everything else. Nextcloud, Vaultwarden, Home Assistant, media servers, monitoring, reverse proxies. Running 24/7 is the right default. The power cost is genuinely small and the convenience is significant.
Reducing Power Consumption Without Reducing Functionality
Practical steps to reduce 24/7 Docker power draw without compromising functionality:
- Enable HDD hibernation: Synology Control Panel → Hardware and Power → HDD Hibernation. Setting drives to sleep after 20 minutes of inactivity saves 5-8W per HDD most of the time. Whether drives hibernate depends on container I/O, logging, mapped volumes and other NAS services; verify actual hibernation rather than assuming containers will allow it.
- Consolidate containers onto fewer NAS devices: Running two NAS devices at 18W each costs $95/year. Running one NAS at 25W costs $66/year. Consolidation before adding more hardware pays off quickly.
- Use SSDs for the system volume: Placing Docker data on a supported SSD volume can reduce HDD access, but HDD hibernation still depends on services, configuration and bulk-data access. HDDs can then be used for bulk data and sleep more aggressively.
- Avoid idle containers you don't actually use: A downloaded image or stopped container does not itself create ongoing CPU load; only running workloads can add runtime power draw. Over a typical self-hosters stack of 15 containers with 5 genuinely unused, that's $13/year for nothing.
Related reading: our NAS buyer's guide, our NAS power consumption guide, and our NAS explainer.
Use our free NAS Power Calculator to calculate your exact running costs.
See also: our complete Synology NAS Australia guide.
See also: our complete QNAP NAS Australia guide.
How much does it cost to run a Synology NAS with Docker 24/7 in Australia?
For a mid-range x86 Synology (DS425+, DS925+) running several Docker containers with drives in sleep mode most of the time, expect 15-22W average draw. At $0.30/kWh that works out to $40-58 AUD per year. The added cost depends on drive count, each drive's measured active-versus-sleep draw, and the number of hours spent spinning. Annual cost for a DS925+ stack depends on the installed drives, hibernation behaviour, workload and tariff; measure average wall power and apply the formula above.
Does running more Docker containers significantly increase power use?
Not significantly for typical home server containers. Adding idle background services may have little effect, but incremental power depends on what each service actually does; compare whole-system wall power under the same conditions. The meaningful increases come from containers that actively process data. Plex transcoding, Immich machine learning, or LLM inference. Those can add 10-30W during active use, but that use is typically brief and on-demand rather than continuous.
Is it cheaper to run Docker on a NAS or on a dedicated mini PC?
The Beelink EQR6 is AMD Ryzen-based, not Intel N100-based; mini-PC idle draw varies by model, configuration and measurement conditions. At 10W draw, annual cost is $26/year at $0.30/kWh. Cheaper than a NAS running the same containers. However, you lose the NAS's storage management, RAID, and backup integration. The practical answer: if you already have a NAS, run Docker on it. If you are building from scratch and only need containers without a large NAS storage requirement, a mini PC is the more power-efficient choice for the compute portion.
Can I reduce NAS power use by scheduling containers to stop overnight?
Yes, for containers where overnight availability is not needed. Synology's Task Scheduler and QNAP's equivalent can run docker stop container-name and docker start container-name on a schedule. Stopping a container for 8 hours can save up to one-third of that container's incremental runtime energy, not one-third of the NAS baseline, and only if the workload is not deferred. For a constant 3W incremental load, that's about $2.63/year. The savings are modest for individual containers but meaningful if you stop 5-10 non-critical containers overnight.
Does running Plex or Jellyfin 24/7 cost significantly more than other containers?
Only when actively transcoding. The media server container itself idles at 1-3W. Transcoding a single stream in software adds 10-25W for the duration of playback. At $0.30/kWh, a 2-hour software-transcoded movie at the stated 10-25W additional draw costs approximately $0.006-0.015. Effectively nothing per view. Hardware transcoding (Intel Quick Sync on x86 NAS) adds only 3-8W per stream and is significantly cheaper. If your NAS supports hardware transcoding, enabling it in Plex or Jellyfin settings substantially reduces both power use and thermal load during playback.
Running Docker containers requires a NAS model whose operating system supports containers, sufficient RAM, and container images compatible with its CPU architecture; x86 is not universally required. For Australian pricing and model comparisons on Docker-capable Synology and QNAP hardware, see the current buying guide.