Skip to main content
NexGPU

Virtual computing

A powerful machine, on demand

Not every GPU need is called training or rendering. Sometimes you just need a machine stronger than the one in front of you, for a few hours, and then you do not need it any more.

Sign up freeView pricing
Cloud host per hour, from
$0.193
Full system access
root
Can stay resident
24/7

01 — When it applies

When you need a temporary strong machine

A laptop's performance ceiling is thermal, not silicon. A thin machine writes code, runs meetings and does design work perfectly well — but the moment a task saturates it for hours it either throttles or becomes uncomfortable to touch. What you need then is not a new laptop but a machine to connect to.

A second case is isolation. You need a clean Linux environment to verify a deployment script, test an open-source project of uncertain provenance, or run an experiment that will make a mess of the system. Do all of that in a rented instance and, if it breaks, destroy it and start again — your own machine is untouched.

A third is collaboration. When several people need the same environment, one shared instance beats each person configuring their own: consistent setup, reproducible results. Hourly billing means the machine only costs money while it is being used.

02 — Workloads

How cloud hosts get used

What they share: you need a machine, not ownership of one.

  • Remote development

    Connect with VS Code Remote or JetBrains Gateway and keep only the editor local — compilation, execution and debugging all happen in the cloud. Your battery and your fans will thank you.

  • Clean-environment validation

    Test deployment scripts, verify a Dockerfile, run an open-source project you are not sure about. If it breaks, destroy and reboot; your own machine carries zero risk.

  • One-off heavy jobs

    Bulk data processing, a large project build, batch transcoding. These arrive occasionally, and upgrading your own machine for them is poor value while renting a few hours is exactly right.

  • GPU-backed visual work

    Run a remote desktop inside the instance when you need a GUI. Pair it with the visual control panel for files, processes and scheduled tasks and it stays usable without the command line.

03 — Access

Three ways to connect

Command line through graphical, whichever suits you.

  • Direct SSH

    Every instance has a public IP and direct ports. Download the key and connect with any SSH client — no bastion hop, so latency is simply your distance to the host.

    SSH · Direct ports · Key auth

  • Browser terminal

    No client to install: open the console and you have a root shell. The quickest way to check a log or edit a config.

    Web terminal · No install · root

  • Visual control panel

    Install aaPanel with one click to manage files, processes, scheduled tasks and services from the browser. After installation, the panel URL and credentials appear on the instance page.

    aaPanel · File manager · Visual

04 — Choosing a GPU

Sizing a cloud host

These uses are rarely VRAM-limited, so pick by what you actually intend to run. When unsure, start cheap and move up if it is not enough.

Use caseVRAM neededCheapest availableNotes
Remote dev, builds, scripting6GBGTX 1660 S$0.083/hrWhen the GPU is not the bottleneck, take the cheapest card — you are really using CPU, RAM and bandwidth.
Light GPU work and experiments12GBRTX 3060$0.100/hrRunning inference, trying things out, learning a framework. Sufficient and inexpensive.
Graphics workstation24GBTesla V100$0.188/hrFor a GUI plus real GPU performance, the 24GB tier is comfortable.
Dedicated heavy-duty host48GBQ RTX 8000$0.508/hrFor a resident machine running demanding work, more VRAM means fewer compromises in tuning.

"Cheapest available" is derived from live inventory — the lowest per-GPU rate among models that clear the VRAM bar — and moves with the market. Multi-GPU nodes rent whole.

05 — Getting started

Booting a cloud host

  • 01

    Size it for the purpose

    Take the cheapest card when the GPU is not the bottleneck. Remember port mappings are fixed at creation — allocate enough if you will expose services.

  • 02

    Pick an image and connect

    Clean Ubuntu or a framework image, either way. Once provisioned, the instance page gives you the SSH command, the web terminal and the port mapping table.

  • 03

    Stay resident or destroy

    Both patterns are supported. One caveat: stopped is not free — disk continues to bill per GB-month, so if you will not need it soon, destroying is the cheaper choice.

06 — FAQ

GPU cloud hosts

Can I use it as a long-term development machine?

Yes — there is no time limit and it runs as long as the balance holds. Be aware the billing model differs from a monthly VPS: charges accrue hourly, so a full month is the hourly rate times 720. If you genuinely need something resident long-term, get in touch about a longer-term arrangement.

Is there a GUI? Can I use remote desktop?

Images are command-line by default, but you have full root and can install a desktop environment and remote desktop server yourself. Note that remote desktop needs an exposed port, so allocate enough ports at order time.

Is my data safe? Can the platform see my files?

The instance is a container you hold exclusively; we neither read nor back up its disk. The corollary is that there is no automatic backup — sync anything important elsewhere, because disk contents are unrecoverable once the instance is destroyed.

What is the network like? Can it reach the internet?

Instances have full public network access plus a public IP and a range of direct ports for serving. Egress is metered per GB at a median around $0.0081/GB. Bandwidth varies considerably between nodes, and the marketplace lists each machine's up and down rates.

Can I use it without knowing Linux?

Yes. The instance page has a one-click button to install a visual control panel; once installed you can manage files, processes, scheduled tasks and services from the browser, with the credentials shown on the instance page. Day-to-day use rarely needs the command line.

Related solutions

Other ways to use it

Same compute network — swap the image and it becomes a different production line.

Start building on NexGPU

Enterprise R&D team or solo developer — either way, your first job can be running in minutes.

Sign up freeTalk to sales

Sign up to browse live network pricing. No payment method required.