Hetzner Cloud vs Vultr

Both rent plain KVM servers by the hour that you set up and run yourself. Hetzner Cloud is the leaner list price with six locations and outbound firewalls, while Vultr spreads across 33 locations and adds GPUs and Windows.

The companies and products

Hetzner Cloud

Hetzner is a German hosting company founded in 1997; Hetzner Cloud rents virtual servers, storage and networking by the hour.

A Hetzner Cloud server is a Linux VM picked from shared vCPU lines (cost-optimized CX and CAX, regular CPX) or dedicated vCPU (CCX), with local NVMe disk, a public IP and up to 20 TB of traffic in the EU. You add volumes, private networks, firewalls and load balancers, and run everything yourself through the Console, the Cloud API, the hcloud CLI or Terraform. It is general infrastructure, not an agent sandbox, so this page compares the two where they overlap: giving an agent its own Linux machine.

Certifications
ISO/IEC 27001:2022

Vultr

Vultr is a general-purpose cloud, and Cloud Compute is its line of KVM virtual servers.

A Vultr instance is a Linux or Windows VM you rent by size, with root, a public IPv4, local SSD or NVMe and a monthly transfer allowance. You choose shared vCPU plans (Regular, High Performance, High Frequency) or dedicated vCPU plans (Optimized Cloud Compute, VX1), and run your own software next to Vultr block storage, file systems, VPCs, Kubernetes, bare metal and GPUs. It is a server host, not an agent sandbox, so this page compares the two where they overlap: giving an agent its own Linux machine.

Which use cases each one fits

For each use case, Battleships prices every sandbox provider on the same workload and sorts them from the cheapest monthly bill up. “51st cheapest of 364” means 50 of the other 363 providers run that workload, with every feature it needs, for less. “With limits” means the provider runs it but misses something, so it is listed after every full match. Open a use case for its whole ranking and what each provider misses.

Use case and its workloadHetzner Cloud: price rankVultr: price rankbochka: price rank
Software factory
100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month
Both give a full VM with root and Docker that you set up yourself; Hetzner says nested virtualization is not possible, and Vultr documents it only on VX1 and bare metal. bochka has /dev/kvm on every machine.
46th of 367, with limits
$287 a month
114th of 367, with limits
$1,060 a month
3rd cheapest of 367
$634 a month
Always-on AI employees
50 machines 24/7 · 2 vCPU / 4 GB
94th of 366, with limits
$325 a month
144th of 366, with limits
$1,000 a month
5th cheapest of 366
$657 a month
Code interpreter bursts
100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month
156th of 345, with limits
$649 a month
168th of 345, with limits
$1,000 a month
186th of 345, with limits
$2,000 a month
Sandboxes for your users
300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month
Neither Hetzner nor Vultr publishes its default server limit, so hundreds of user sandboxes need a support request first on both; bochka allows 100 to 2,000 sandboxes by plan.
46th of 361, with limits
$314 a month
98th of 361, with limits
$899 a month
Cheapest of 361
$209 a month
RL rollouts
1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month
104th of 360, with limits
$6,490 a month
179th of 360, with limits
$20,000 a month
34th of 360, with limits
$2,000 a month
Agent evals
250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month
Both fit agent evals as full VMs, and they land close together in the table; Hetzner's CX line is out of stock today, so the servers you can order cost more.
51st cheapest of 364
$1,010 a month
47th cheapest of 364
$893 a month
10th cheapest of 364
$200 a month
CI runners
40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month
Both bill a ten-minute job as a full started hour and leave the runner software to you; bochka bills per second and stops for free.
83rd cheapest of 377
$788 a month
91st cheapest of 377
$857 a month
4th cheapest of 377
$86 a month
24/7 dev boxes
1 machine 24/7 · 4 vCPU / 8 GB
All three fit a 24/7 dev box: Hetzner and Vultr are classic always-on servers with a monthly cap, and bochka keeps the disk with snapshots every minute.
32nd cheapest of 368
$42 a month
28th cheapest of 368
$40 a month
18th cheapest of 368
$26 a month
Computer-use agents
50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month
Neither ships a desktop: on Hetzner and on Vultr you install the desktop, VNC and the control loop yourself, while bochka streams a 60fps desktop with computer use.
134th of 369, with limits
$320 a month
252nd of 369, with limits
$1,190 a month
6th cheapest of 369
$200 a month
Browser-use agents
50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month
Neither ships a browser stack, so you install and run Chromium yourself on both; bochka has Chromium and browser control in every sandbox.
227th of 367, with limits
$312 a month
300th of 367, with limits
$894 a month
7th cheapest of 367
$200 a month
Scraping & bots (startup scale)
50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month
157th of 346, with limits
$62 a month
175th of 346, with limits
$89 a month
6th cheapest of 346
$90 a month
Mass scraping & bots (undetectable)
1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month
45th of 338, with limits
$1,248 a month
72nd of 338, with limits
$1,818 a month
71st of 338, with limits
$1,800 a month
Parallel red teaming
200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month
114th of 360, with limits
$52 a month
168th of 360, with limits
$150 a month
2nd cheapest of 360
$200 a month
GPU eval jobs
10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month
Only Vultr rents GPUs (Nvidia and AMD); Hetzner Cloud and bochka have none.
Can't run itCan't run itCan't run it
macOS agents
5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month
Can't run itCan't run itCan't run it

Key features

FeatureHetzner CloudVultrbochka
What you getA general Linux VM you set up and run yourselfA general KVM server you set up and run yourselfA sandbox API for agents, on full VMs
IsolationEvery server is a full VM with its own kernelEvery instance is a KVM VM with its own kernelEvery sandbox is a full VM with its own kernel
CPU speed (Node.js build-and-test, 4 vCPU / 8 GB)8.88 runs/s on a shared CX33; CPX and CCX were not measuredNo comparable measurement26.69 runs/s on dedicated AMD Ryzen 9 9950X (Zen 5) hosts
Docker and nested virtualizationRoot and Docker; nested virtualization is "not possible on cloud server"Root and Docker; virtualization documented for VX1 and bare metal, not other Cloud Compute plansDocker, root, systemd and /dev/kvm on every machine
Billing unitEvery started hour, with a monthly capEvery started hour, capped at 672 hours a month on most plansPer second from plan credit
StoppingA powered-off server bills at the full rate until you delete itA stopped instance is billed in full until you destroy itA stopped sandbox costs nothing and keeps its disk
SnapshotsSnapshots you take yourself, or daily backups in 7 slots for 20% of the server priceManual disk snapshots that can take up to 30 minutes; paid scheduled backupsAutomatic snapshots every minute and on stop
ForkSnapshot, then create a new server from it (disk, no memory)Snapshot, then deploy new instances from it (disk, no memory)Fork a running sandbox in one call (disk, no memory)
Reaching the machinePublic IPv4 and IPv6 on any port; HTTPS needs your own certificates or a paid load balancerPublic IPv4 on any port; HTTPS needs your own domain, proxy and certificatesHTTPS URLs for up to 50 ports, token-protected; raw TCP; own IP
Outbound rulesFree Cloud Firewalls with inbound and outbound rules by IP, port and protocolVultr Firewall filters incoming traffic only; outbound rules go in the guestOpen internet on every plan; no provider-side allowlist
Desktop and coding agentsNone built in; you install a desktop and agents yourselfHeadless Linux; you install a desktop, VNC and agents yourself60fps desktop with Chromium and computer use; Claude Code, Codex, OpenCode and more preinstalled
Keeping keys out of the machineNo credentials proxy; secrets go on the serverNo credentials proxy; secrets go on the serverAI credentials proxy for model providers and any HTTPS API you register
GPUs and WindowsNo GPUs; no WindowsNvidia and AMD GPUs; Windows Server for a license feeNo GPUs; Linux guests only, Wine for Windows apps
RegionsGermany and Finland, plus Ashburn, Hillsboro and Singapore33 locations on six continentsEU only: Germany, Finland, France
Every feature Battleships tracks (43)
FeatureHetzner CloudVultrbochka
Full VM (own kernel)YesNot publishedYes
Docker insideYesYesYes
Nested virtualizationNoNot publishedYes
RootYesYesYes
SystemdYesYesYes
Desktop GUINoNoYes
BrowserNoNoYes
Computer-use APINoNoYes
SSHYesYesYes
Persistent diskYesYesYes
Pause/resumeYesNoYes
Automatic snapshotsNoNoYes
Fork/cloneYesYesYes
Live fork (no pause)YesNot publishedYes
Memory forkNoNoNo
Custom image (Docker or snapshot)YesYesYes
≥24 h sessionsYesYesYes
Idle auto-stopNoNoNo
Wake on requestNoNoNo
HTTPS preview URLsNoNoYes
Raw TCP inboundYesYesYes
Public IPv4YesYesNo
Egress allowlistYesNoNo
Secret proxyNoNoYes
Secret proxy for any APINoNoYes
Preinstalled agentsNoNoYes
Agent harness APINoNoYes
MCP serverNoNot publishedNo
GPUNoYesNo
WindowsNoYesNo
Arm64YesNoNo
Self-hostingNoNoNo
BYOCNoNoNo
Open sourceNoNoNo
Scoped API keysNot publishedNot publishedYes
Spend limitsNoNot publishedYes
Usage per sandbox (API)NoNoYes
Audit logsNot publishedYesNo
SSONoYesNo
SOC 2NoNot publishedNo
HIPAANoNot publishedNo
EU data residencyYesYesYes
Zero data retentionNot publishedNot publishedYes

Performance

Hetzner Cloud8.88 runs/s
VultrNot measured
bochka26.69 runs/s

The only Hetzner number is 8.88 Node.js build-and-test runs per second on a shared CX33, measured through bochka's temporary fallback capacity, against 26.69 for bochka on its own hosts, about a third of bochka's speed. Vultr has no comparable measurement, so there is no fair speed ratio between Hetzner and Vultr. Hetzner's CPX and CCX servers were not measured either.

Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Hetzner Cloud measured on: Shared vCPU, Cost-Optimized (CX, Intel/AMD), EU. Vultr has no published run in this suite yet, so there is no like-for-like number to compare.

Pricing in five scenarios

Each bill is priced from the providers' public price lists by the open Battleships engine, with the cheapest plan that runs the scenario.

ScenarioHetzner Cloud, a monthVultr, a monthbochka, a month
One coding agent, workdays
4 vCPU / 8 GB, 50 GB disk, 8 hours a day, 22 days a month, files kept between days
$3 · Pay-as-you-go (postpaid) · with limits$10 · Usage account$20 · $20/mo
One agent that never stops
4 vCPU / 8 GB, 50 GB disk, running 24/7 all month
$10 · Pay-as-you-go (postpaid) · with limits$40 · Usage account$26 · $20/mo
50 agents in parallel
50 machines of 4 vCPU / 8 GB at once, 8-hour sessions, 1,100 sessions a month, Docker inside
$141 · Pay-as-you-go (postpaid) · with limits$524 · Usage account$317 · $20/mo × 3 seats
Sandboxes for your users
2 vCPU / 4 GB, 30,000 sessions of 10 minutes a month, up to 300 at once
$312 · Pay-as-you-go (postpaid) · with limits$894 · Usage account$200 · $100/mo × 2 seats
1,000 short jobs at once
2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month
$2,080 · Pay-as-you-go (postpaid) · with limits$5,960 · Usage account$2,000 · $2000/mo · with limits

Hetzner Cloud: Hetzner bills each server by the started hour with a monthly cap, and keeps billing until you delete it, even when it is off. The prices here use the shared CX line plus the public IPv4, and CX is currently out of stock, so the CPX and CCX servers you can order cost more.

Vultr: Vultr bills each plan by the started hour, capped at 672 hours a month on most plans (VX1 has no cap), from deploy to destroy, including while stopped. The prices here use the shared Regular plans; snapshots, backups and extra IPs cost extra.

Vultr: You pick a monthly plan from $20, and all of it becomes usage credit spent per second on running sandboxes. A stopped sandbox costs nothing and keeps its disk.

Limits in those scenarios

Every Hetzner scenario is priced on the CX line, which Hetzner's site marks as currently unavailable, so none fit as priced today. Vultr's Regular plans fit every scenario on price. Neither publishes how many servers a new account can run at once, so 50 agents, 300 user sandboxes or 1,000 jobs at once need a support request on both. On bochka the limit is a sandbox count (100 to 2,000 by plan) plus starts per minute, so the 1,000-job burst gets queued and needs more starts per day than the $2,000 plan allows.

ScenarioHetzner CloudVultrbochka
One coding agent, workdays
  • stock-limited: all CX types "currently unavailable" on hetzner.com (2,026-09-28)
None that biteNone that bite
One agent that never stops
  • stock-limited: all CX types "currently unavailable" on hetzner.com (2,026-09-28)
None that biteNone that bite
50 agents in parallel
  • stock-limited: all CX types "currently unavailable" on hetzner.com (2,026-09-28)
None that biteNone that bite
Sandboxes for your users
  • stock-limited: all CX types "currently unavailable" on hetzner.com (2,026-09-28)
None that biteNone that bite
1,000 short jobs at once
  • stock-limited: all CX types "currently unavailable" on hetzner.com (2,026-09-28)
None that bite
  • max 2,400 starts/day (needs ~6,667)
  • max 600 starts/hour (needs ~1,000)
  • bursts above 90 starts/min are queued (a burst of 333 takes ~4 min to start)

Which one should you use?

Use Hetzner Cloud when

you want the lowest list price for plain long-lived Linux servers, outbound firewall rules, or big dedicated machines, and the line you need is in stock.

  • Lower list prices. Useful if cost per server matters most: Hetzner's CX prices come out well below Vultr's Regular plans in every scenario here, while in stock.
  • Outbound firewall rules. Useful if an agent's machine must only reach a few addresses: free Cloud Firewalls can drop all other outbound traffic.
  • Big dedicated machines. Useful for heavy single machines: dedicated CCX servers go up to 48 vCPU and 192 GB of memory.

Use Vultr when

you need GPUs, Windows, many regions around the world, or shared servers you can order right now.

  • Regions around the world. Useful if your users are spread out: Vultr runs in 33 locations, from Sao Paulo and Johannesburg to Tokyo and Sydney.
  • GPUs and Windows. Useful if agents serve models or need Windows: Vultr rents Nvidia and AMD GPUs and Windows Server instances.
  • Servers you can order today. Useful if you need machines now: Vultr's shared Regular plans fit every scenario here, while Hetzner's CX line is out of stock.

Where bochka fits

Hetzner Cloud and Vultr are general clouds, so they win for GPUs, Windows, regions outside Europe or big long-lived servers. If what you are really running is agents, each on its own Linux machine, bochka gives every agent a full VM that is ready to work, with free stops, minute snapshots and agents built in.

Nested virtualization on every machine

Hetzner says VMs inside a cloud server are not possible, and Vultr documents it only on VX1 and bare metal. An agent that runs emulators or its own VMs needs /dev/kvm, which every bochka machine has.

Idle agents without paying for idle servers

Both Hetzner and Vultr bill a stopped server in full until you delete it, so you script snapshot, delete and recreate yourself. bochka stops for free, saves the disk every minute and forks a running sandbox in one call.

Agents, desktop and keys already handled

Both start as an empty OS image: you install each agent, a desktop and HTTPS, and put API keys where the agent can read them. bochka ships coding agents, a 60fps desktop, HTTPS on 50 ports and a credentials proxy.

Use bochka when what you are really running is agents: each on its own full VM with /dev/kvm, coding agents, a desktop, HTTPS ports and minute snapshots built in, fork while running, and stops that cost nothing. Compare it one to one: bochka vs Hetzner Cloud and bochka vs Vultr.

Questions people ask

Is Hetzner Cloud cheaper than Vultr?

On list price, yes. One agent that never stops comes to $10 a month on Hetzner's CX line against $40 on Vultr, and 50 agents in parallel come to $141 against $524. But the CX line is currently out of stock, so the Hetzner servers you can order today cost more. bochka comes to $26 and $317 for the same two scenarios.

Which is faster, Hetzner Cloud or Vultr?

There is no fair comparison. A shared Hetzner CX33 did 8.88 Node.js build-and-test runs per second, and Vultr was not measured in the same benchmark. bochka did 26.69 runs per second at 4 vCPU / 8 GB.

Can Hetzner Cloud or Vultr run Docker and nested VMs?

Docker works on both, since each server is a full VM with root. Hetzner says nested virtualization is not possible on cloud servers, and Vultr documents it only for VX1 and bare metal. On bochka, Docker, systemd and /dev/kvm are documented on every machine.

Do I pay for a stopped server on Hetzner or Vultr?

Yes, on both. Each keeps billing a stopped server until you delete or destroy it, so to stop paying you snapshot, delete and recreate it later. On bochka, a stopped sandbox costs nothing and keeps its disk.

Which has more regions?

Vultr, with 33 locations on six continents. Hetzner Cloud runs in Germany, Finland, Ashburn, Hillsboro and Singapore. bochka runs only in the EU, in Germany, Finland and France.

Can I block outbound traffic from an agent's machine?

On Hetzner, free Cloud Firewalls take outbound rules by IP, port and protocol. Vultr Firewall filters incoming traffic only, so outbound rules go inside the guest. bochka has open internet on every plan and no provider-side allowlist.

Try bochka

Install the CLI, then run bochka new.

Product preview

More comparisons

Bochka product preview · Pricing, benchmarks and commands are illustrative. Sandbox provisioning is not connected.