Editing Openai/695158d7-1b68-800b-a02f-4817bf1f6d25 (section)

=== Assistant: Good question — this gets to the heart of how Qubes actually stores things. ===
Good question — this gets to the heart of how Qubes actually stores things.

I’ll give you realistic numbers, not marketing ones, and explain why.

==== For one AppVM (“qube”), ignoring user data and installed programs: ====

: 

That’s the baseline storage footprint of an AppVM.

==== In Qubes, an AppVM’s disk is layered: ====

===== - Shared =====
* Stored once, no matter how many AppVMs use it
* Does not count per qube

➡️ 0 GiB per qube

===== This is the main per-qube cost. =====

Typical size:
* Default: 2 GiB
* Actually used on disk (empty): ~300–600 MiB
* Grows as you add files

Contains:
* /home/user
* App settings
* Browser profiles
* Downloads (unless redirected)

===== - Ephemeral =====
* Backed by sparse storage
* Typically <100 MiB when idle
* Discarded on shutdown (unless persistent)

===== - VM config =====
* Device mapping
* Xen bookkeeping

➡️ ~50–150 MiB

==== |     |     | ====
| --- | --- |
|     |     |
|     |     |
|     |     |
|     |     |

Rounded up safely:

: 

==== Because: ====
* Default private.img limit = 2 GiB
* People leave browser cache / downloads there
* Safety margin

But the actual disk usage is usually lower.

==== Templates are the real storage users. ====

Typical TemplateVM sizes:
* debian-12: ~4–6 GiB
* fedora-39: ~6–8 GiB

But:
* Each template is stored once
* Shared by unlimited AppVMs

==== Let’s say: ====
* 3 templates → ~20 GiB
* 30 AppVMs → ~30 GiB
* dom0 → ~5–8 GiB

➡️ ~60 GiB total for a very VM-heavy setup

Plenty of headroom.

==== ### ====
* Leave private.img at 2 GiB
* Increase only for specific qubes (browsers, dev)

===== - Media → media qube =====
* Downloads → dedicated storage qube
* Dev builds → separate work qube

==== - Baseline per qube: ~1 GiB ====
* Configured limit: usually 2 GiB
* Templates dominate storage
* Qubes scales very well in storage usage

If you want, next I can:
* Estimate how many qubes you can realistically run
* Suggest which qubes need larger private storage
* Help you pre-plan LVM logical volume sizes

Just tell me 👍