The results
Time from order to "running", Ubuntu 22.04 LTS, standard cloud plans, in seconds, rounded:
| Provider | Typical (median) | 9 out of 10 ready within |
|---|---|---|
| Hetzner | about 20 | about 25 |
| DigitalOcean | about 35 | about 55 |
| Vultr | about 45 | about 55 |
| Linode | about 85 | about 95 |
Hetzner is the most consistent: nine out of ten servers were running within roughly 25 seconds. DigitalOcean is quick on a typical order but varies more from one deploy to the next. Linode takes longer on average and is predictable, because even its slowest deploys were ready in under two minutes.
"Running" is not "you can log in"
A server that the provider calls running still has to finish booting and start the SSH daemon. For a smaller group of orders we also recorded when port 22 started accepting connections, which is the moment you can actually work.
The order of the providers stays the same, and the extra wait after "running" was roughly 10 seconds on Hetzner, 15 to 20 on Vultr, 25 on DigitalOcean and 30 on Linode. In round numbers, you can log in after about 30 seconds on Hetzner, a minute on Vultr and DigitalOcean, and two minutes on Linode.
Across all images, not only Ubuntu 22.04, the typical times were close to the same: about 20 seconds on Hetzner, 35 on DigitalOcean, 45 on Vultr and 75 on Linode. We use Ubuntu 22.04 above because a single image keeps the comparison fair.
Does the city change the time?
The city changes VPS deploy time a little within one provider. On Hetzner the European sites came out the same. On DigitalOcean the typical time ranged between about 35 and 45 seconds across European cities, and on Linode between about 65 and 85. The provider matters much more than the location: the slowest city of the fastest provider here still beats the best city of the slowest one.
Windows and bare metal take longer
Windows Server is installed and configured on first boot, and a physical machine has to be prepared, so both can take several minutes on a slow order. The typical Windows deploy was still about a minute and a half, but plan for a few minutes when a script waits on one. If you create machines from a pipeline, our guide to deploying a VPS from the API covers how to wait for readiness properly instead of guessing a sleep.
What this means for you
If you create a server once and keep it for months, 20 seconds against 85 does not matter. It matters in three cases:
- CI runners, scrapers and test environments that are created and destroyed many times a day, where a minute of waiting per machine adds up.
- Autoscaling and failover, where a replacement has to be serving traffic quickly.
- Scripts that create a server and immediately connect: they need a retry loop on SSH, because the server is "running" a few seconds to half a minute before it accepts a login.
Speed is one factor out of several. The right provider also depends on your region, your bandwidth and your payment method, and our comparison of the four clouds goes through the rest. When you order on NoctHost, the provider is a choice you make per server, and the price list shows the same plan on each backend.
How we measured
The timer starts when the order is accepted on our side and stops when the provider's API reports the instance as running. The time to SSH is measured by a probe that connects to port 22 and counts the seconds from the same start.
Orders restored from a snapshot are excluded, because they copy a disk and take longer by design. The sample covers late October to early November 2026 and the images and plans customers actually chose, so it reflects real use, not a controlled benchmark. Times shift as providers change their infrastructure, so we re-measure and update the figures here.