Serious compute.
Ridiculously low friction.
Linux VPS and Windows RDP instances that boot in under a minute, billed by the hour, on hardware you'd actually pick yourself — EPYC 9004, Xeon 6, and RTX 5090s across 32 datacenters worldwide.
➜ ~ nodara deploy --plan nv-4x --region fra1 --image ubuntu-24.04 ▸ matching offer .. 4 vCPU · 16 GB · 300 GB NVMe · 10 Gbps ▸ allocating .. fra1-c7-rack-14 / node 08 ▸ imaging .. ubuntu-24.04-cloudinit · verified ▸ booting .. cloud-init ok · ssh ready ✓ instance nv-8f2a10c3 is up in 47s root@nv-8f2a10c3:~#
Trusted by builders at
Fewer surprises,
more room to build.
We looked at every line item that felt like a gotcha on other hosts and either killed it or included it. Here's what that gets you.
Hourly, prorated to the second
Spin an instance up for a load test, tear it down when the run ends. You pay for exactly 6m 12s if that's what you used — no 24h minimums, no per-hour rounding.
Hardware you'd actually choose
AMD EPYC 9004 (Zen 4c), Xeon 6 Granite Rapids, RTX 5090 and H200 GPUs. Gen5 NVMe as standard, ECC RAM everywhere. See the exact CPU model before you rent.
10 Gbps uplinks, unmetered egress
Every plan ships with a symmetric 10 Gbps port. Egress is free within the plan's monthly bandwidth cap — after that it's $0.005/GB, capped at $30.
Snapshots that don't lie
Copy-on-write snapshots at the storage layer. Rollback in under 4 seconds, ship an image to another region without leaving the dashboard.
Windows RDP that isn't cursed
Real Windows Server 2022/2025, activated, with your choice of licensed apps preinstalled. Not a hacked-together QEMU image with a keygen.
SSH keys, firewall, backups — free
Weekly backups, private networking between your instances, IPv6 by default, DDoS scrubbing up to 200 Gbps. We don't nickel-and-dime the basics.
32 points of presence,
one flat egress bill.
Our backbone is peered at DE-CIX, LINX, AMS-IX, Equinix and SIX, with dedicated waves between our core sites. Choose the closest region and let anycast handle the rest.
- p95 intra-region
- 0.34 ms
- p95 inter-continent
- 82 ms
- peering sessions
- 147
- tier-1 transits
- 6
Three ways teams put a
Nodara box to work.
Ship a game server on Sunday
Minecraft, Valheim, Palworld — spin up on the closest edge, hand your friends an address, delete it Monday morning. You pay $0.94.
➜ ~ bash -c "$(curl nodara.host/g/valheim)" ▸ rented nv-4x @ waw1 · 4c · 16 GB ▸ port 2456/udp opened ✓ server up: waw1.nod.rocks:2456
Fleet of a hundred bots
Discord scrapers, market-making bots, always-on cron work. Use our CLI to fan out — same image, per-region overrides, no lock-in.
# fleet.yaml image: alpine-3.20 plan: nv-1x fanout: fra1: 40 iad1: 40 sin1: 20 # → nodara fleet up → 100 boxes in 62s
Windows dev box you can throw away
Real Server 2022 with Visual Studio and Office preinstalled. Connect from any RDP client, close the tab when you're done.
➜ ~ nodara rdp --sku win-2022-devbox ▸ region: fra1 · 4c · 8 GB · gen5 nvme ▸ password auto-rotated → dashboard ✓ open nodara.host/rdp/8f2a
Priced in seconds,
not spreadsheets.
Same box, whether you use it for 8 minutes or 30 days. Windows RDP and GPU tiers on the full pricing page.
Nano
nv-1xSidecars, cron jobs, small bots.
- 1 vCPU
- 2 GB
- 40 GB NVMe
- 2 TB / mo
- IPv4 + IPv6
Micro
nv-2xStaging APIs, hobby SaaS, discord bots.
- 2 vCPU
- 4 GB
- 80 GB NVMe
- 4 TB / mo
- IPv4 + IPv6
Real people.
Real workloads.
“We use Nodara for chaos-testing our staging cluster. Terraforming 40 nodes takes about a minute, tearing them down takes seven seconds, and the bill for a two-hour run is under nine dollars. The math wasn't close.”
“I rent Windows RDPs to test my game on 'a real user's machine' — not a VM that pretends. Nodara is the only place where the box actually feels responsive on RDP over a hotel wi-fi.”
“H200 by the hour is what convinced my PI. We now run 3-4 short fine-tunes a week instead of queuing on the university cluster. Sub-minute allocation matters more than raw price/hour.”
The honest FAQ.
If your question isn't here, ask us in the dashboard chat — median response time is 4 minutes during business hours.
Yes — any hourly instance can convert to a monthly commit from the dashboard. The change is instant and prorated: the hours you already spent that month count against the monthly rate, so you never pay twice.
You get an alert within 15 seconds. Instances with a snapshot enabled (all of them, by default) are re-provisioned on a healthy host in the same region automatically, with the same IP where possible. If we can't recover in under 4 minutes we credit the outage at 100x the pro-rated rate.
Yes — noVNC-based console in the dashboard for every instance, plus a hardware serial console for Linux. RDP boxes get an emergency VNC channel that bypasses the guest OS.
Every account gets a real human on the abuse team, not a form. First-time reports come with a 48-hour window to remediate before any action is taken. We never null-route without notice.
Yes. Every Windows RDP instance is activated under Microsoft's Services Provider License Agreement (SPLA). You'll see the activation status in your dashboard, and the license is tied to the instance for its lifetime.
On Scale tier and above. Bring a /24 or larger, we announce it from any of our core PoPs. BYOIP setup is a one-line ticket, usually done within 24 hours.