Short answer: On four official pages read on 2026-11-18, "upgrade" and "downgrade" are four different operations: OVHcloud upgrades in one click with a quick reboot but sends downgrades to a new plan and a data transfer; DigitalOcean lets CPU and RAM move both ways but powers the Droplet off and never shrinks the disk; Vultr performs an in-place plan change only when the host has room and keeps the IP in that case, otherwise it is a snapshot-and-new-server migration with a new IP; Hetzner rescales up or down inside the same architecture and refuses any disk downgrade. A plan table that lists only sizes shows none of these paths.
Every cell comes from an official provider page opened on 2026-11-18. Where the page printed no figure or statement, the cell says so and names the pages read.
| Provider | Upgrade path as printed | Downgrade path as printed | Reboot / power state | Data and IP as printed | Disk limit as printed | Official source (opened 2026-11-18) |
|---|---|---|---|---|---|---|
| OVHcloud (US) | "Upgrade your VPS resources with a single click -- just a quick reboot required, with no data migration or configuration changes needed"; "without changing servers" [S1] | "You will need to subscribe to a new plan, transfer your data, then cancel your old service" [S1] | "just a quick reboot required" for upgrade; no downgrade reboot statement on the page read [S1] | No manual data migration on upgrade; downgrade requires transferring data; IP behaviour on either path: Read not available on the page read [S1] | Read not available — the VPS page read prints no disk shrink or disk floor rule [S1] | us.ovhcloud.com/vps/ [S1] |
| DigitalOcean | Resize changes CPU and RAM in either direction; a second option also "permanently increases the size of a Droplet's disk" [S2] | "CPU and RAM only" can decrease; resize is allowed only to a plan with "equal or greater amount of disk space as the original Droplet" [S2] | Power off before resizing in the control panel or CLI; API resize powers down automatically; "Droplets may change hypervisors during a resize" [S2] | Backup recommended before resize because filesystem changes can lose data; IP behaviour on resize: Read not available on the page read [S2] | "You cannot decrease the size of a Droplet's disk"; volumes are suggested as removable storage instead [S2] | docs.digitalocean.com/products/droplets/how-to/resize/ [S2] |
| Vultr | Console Settings, Change Plan, Upgrade; an in-place upgrade is used when the physical host has room [S3] | Read not available — the docs page read covers in-place upgrade and migration to a new host only; the docs sitemap was searched and no downgrade-in-place article was found [S3][S4] | "Your server will restart with the new plan"; migration path requires snapshot, new deployment and removal of a rescue ISO [S3][S4] | "The server's IP address will remain the same after an in-place upgrade"; a snapshot-based migration gives the new server "a new IP address" [S3] | Upgrading to a larger plan increases the virtual disk, after which the partition and file system must be expanded manually; no disk shrink path printed on the pages read [S4] | docs.vultr.com/how-to-upgrade-your-plan-without-reinstalling [S3]; docs.vultr.com/resize-disk [S4] |
| Hetzner | "You can also rescale to and from a plan with shared resources or up- and downgrade to another plan"; the plan change is done in Hetzner Console [S5] | Rescale down is possible only to a plan whose disk is at least as large as the current disk; "we do not offer disk downgrades" [S5] | Reboot or power state on rescale: Read not available — the FAQ read prints no rescale reboot statement [S5] | CPU type may change during rescale; same architecture is required; IP retention on rescale: Read not available, though a Primary IP can be reassigned to a new server if auto-delete is off [S5] | "You can only rescale to server plans with a disk size equal to or larger than the current disk"; when scaling up you may keep the current disk size to preserve a later downgrade; partition must be enlarged manually via Rescue System [S5] | docs.hetzner.com/cloud/servers/faq/ [S5] |
A downgrade that ends in a new plan still leaves the disk floor to deal with; the VPS downgrade rollback page follows what happens after the plan change, including the copy to a smaller disk and the address rules on the replacement machine.
A range change is a different question from a plan change; the VPS 2027 config shift page records why the same configuration may require a different model when a provider replaces a range.
None of the four "upgrade" paths is a single atomic change: one is a control-panel click with a reboot, one is a CPU/RAM dial that may move the machine to another host, one is conditional on spare host capacity, and one is a plan rescale that cannot cross architectures.
vpshostingservice.co knowledgebase): a single vendor's control-panel steps; no cross-provider paths, no reboot or IP conditions side by side. This page supplies the four official paths.flipperhost.com billing knowledgebase): one billing workflow, not a comparison of what an upgrade means on four platforms. This page separates plan change from disk change.flipperhost.com, Dutch knowledgebase): the same vendor workflow in another language; no official provider documentation from the providers compared here. This page quotes only provider documentation.qhoster.com clients how-to): one hosting panel route; no statement about disk floors or address behaviour. This page prints both.vpscost.com blog): the title promises the same fields but the result is a third-party roundup, not a provider manual. This page uses only official pages opened on 2026-11-18.greencloudvps.com billing knowledgebase): another single-vendor billing article; no cross-provider path comparison. This page compares four vendors' own words.qhoster.com): a duplicate result for the same host's article; no additional fields. This page does not repeat a vendor's workflow as evidence.An existing page on this site, [VPS downgrade rollback](/cheap-vps-downgrade-rollback), records that DigitalOcean forbids shrinking a Droplet disk and that Hetzner refuses disk downgrades while allowing rescale to a plan with an equal or larger disk. The OVHcloud official page read today adds a third shape: its downgrade is not a dial at all but a new subscription, a data transfer and cancellation of the old service. The site's existing facts and today's OVHcloud fact are both official; together they show that one word, downgrade, covers a CPU/RAM dial, a disk floor and a service replacement.
DigitalOcean's resize guide recommends a snapshot before resizing and gives separate procedures for the control panel, CLI and API; it also says a resize can take longer when more disk data has to move over the network. [S2]
Vultr's upgrade article says an in-place upgrade takes several minutes and needs spare capacity on the physical host; when that capacity is absent, or when the server type changes, the documented route is a snapshot, a new deployment and a new IP. [S3] Its file-system guide explains that a larger plan may leave the old partition unchanged and that the file system must be expanded by hand from a rescue environment. [S4]
Hetzner's FAQ says the CPU type may change on rescale because hardware is allocated by availability, and that rescale is restricted to the same architecture type. [S5]
OVHcloud's VPS page says an upgrade needs no data migration or configuration changes, while a downgrade needs a new plan and a data transfer. [S1]