Crusoe vs Hyperstack

Crusoe and Hyperstack are both GPU clouds that also rent plain CPU VMs you set up yourself. The big differences are which GPUs they carry, where they run, and what happens when you stop a VM: on Crusoe it only keeps paying for its disk, on Hyperstack it keeps billing at the full rate unless you hibernate it.

The companies and products

Crusoe

Crusoe calls itself "the AI factory company": it builds energy-first data centers and rents GPU and CPU machines on Crusoe Cloud.

Crusoe Cloud is infrastructure for training and serving models: Nvidia and AMD GPU VMs, general-purpose and storage CPU VMs, disks, managed Kubernetes and Slurm, and managed inference. You create Ubuntu VMs through a console, CLI, Terraform or API, log in over SSH and install what you need. It is not a sandbox product, so this page compares the CPU VMs where it overlaps with bochka.

Hyperstack

Hyperstack is the GPU cloud of NexGen Cloud, a UK company founded in 2020: you rent GPU and CPU virtual machines, Kubernetes clusters and storage by the minute.

A Hyperstack VM is a full virtual machine with root and SSH, booted from an Ubuntu image (with or without Docker) and sized from a fixed list of flavors, from one RTX A4000 to 8-GPU B300 nodes, plus CPU-only flavors from 4 to 32 vCPUs. It also runs managed Kubernetes, volumes, object storage and an AI Studio inference API. It is built for teams that train, fine-tune and serve models, not as a sandbox API for agents, so this page compares the parts that overlap: VMs as machines for agent work.

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. “43rd cheapest of 364” means 42 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 workloadCrusoe: price rankHyperstack: price rankbochka: price rank
Software factory
100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month
Both give a coding agent a plain VM with root and Docker, but neither has a desktop, built-in agents or a fork of a running machine; bochka has all three on every sandbox.
232nd of 367, with limits
$2,996 a month
343rd of 367, with limits
$11,458 a month
3rd cheapest of 367
$634 a month
Always-on AI employees
50 machines 24/7 · 2 vCPU / 4 GB
An agent that runs all month fits on both as a VM you watch yourself; Crusoe comes out far lower than Hyperstack here, and bochka adds a desktop and a browser.
234th of 366, with limits
$3,000 a month
337th of 366, with limits
$12,775 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
128th of 345, with limits
$268 a month
172nd of 345, with limits
$1,167 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 has HTTPS URLs for each machine, and Hyperstack does not publish how many VMs can run at once, so 300 user sandboxes is not confirmed there.
104th of 361, with limits
$951 a month
173rd of 361, with limits
$4,207 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
71st of 360, with limits
$4,055 a month
173rd of 360, with limits
$17,500 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
Crusoe fits agent evals fully, while Hyperstack's unconfirmed limit on VMs at once makes it a compromise; bochka fits too and starts each sandbox from a snapshot.
43rd cheapest of 364
$811 a month
204th of 364, with limits
$3,250 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 fit for build and test jobs on full VMs; Crusoe bills per second and Hyperstack per minute.
48th cheapest of 377
$395 a month
122nd cheapest of 377
$1,560 a month
4th cheapest of 377
$86 a month
24/7 dev boxes
1 machine 24/7 · 4 vCPU / 8 GB
Both keep the whole disk; a stopped Crusoe VM pays only for its OS disk, while a Hyperstack VM must be hibernated to stop paying for compute.
87th cheapest of 368
$121 a month
115th cheapest of 368
$475 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 graphical desktop, so a computer-use agent means installing and streaming one yourself; bochka ships a 60fps desktop with Chromium.
221st of 369, with limits
$811 a month
332nd of 369, with limits
$3,250 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
246th of 367, with limits
$405 a month
335th of 367, with limits
$1,750 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
268th of 346, with limits
$403 a month
327th of 346, with limits
$1,750 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
201st of 338, with limits
$8,055 a month
310th of 338, with limits
$35,000 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
201st of 360, with limits
$207 a month
309th of 360, with limits
$876 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
Hyperstack ranks 2nd of 98 for GPU eval jobs, well ahead of Crusoe, and bochka has no GPUs.
86th of 98, with limits
$7,803 a month
2nd cheapest of 98
$625 a month
Can'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

FeatureCrusoeHyperstackbochka
What you getA plain Ubuntu VM you set up yourself, plus GPUs, Kubernetes and SlurmA plain Ubuntu VM you set up yourself, plus GPUs, Kubernetes, volumes and object storageA sandbox for agents: a full VM with tools, a desktop and coding agents ready
IsolationFull VM on cloud-hypervisorFull VM, on flavors mapped 1:1 to physical resourcesFull VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts
Root, Docker and nested virtualizationsudo user (root login deprecated); Docker preinstalled on GPU images; nested virtualization not documentedRoot and SSH; Ubuntu images with or without DockerRoot, systemd, Docker and /dev/kvm on every machine
Stopped machinesNo compute charge when stopped; the 128 GB OS disk keeps billing until the VM is deletedA stopped VM bills at the full rate; hibernation bills only disk and a kept IP, and a restore needs the flavor in stockA stopped sandbox costs nothing and keeps its whole disk
SnapshotsNo snapshots of disks or running VMs; a custom image of the boot disk while the VM is shut offSnapshots on request, with the VM shut off, one at a time, up to 100 GBDisk snapshotted every minute and on stop, while the sandbox keeps running
ForkBoot new VMs from a custom image, in the same project and regionNo fork; snapshot, make a custom image and deploy a new VMFork a running sandbox into a new one with the same disk
Desktop and computer useNone; install and stream a desktop yourselfSSH and text console logs; no graphical desktop60fps desktop and Chromium, computer use and browser control
Coding agentsNone; an MCP server lets an assistant inspect your Crusoe infrastructureNone; a hosted MCP server lets your own agent manage HyperstackClaude Code, Codex, OpenCode and more preinstalled, driven from outside with a prompt, steer and events API
Keeping keys out of the machineNo credentials proxy; secrets live on the VMNo credentials proxy; keys go in with cloud-init, files or environment variablesAI credentials proxy for model providers and any HTTPS API; switch a credential to another account when credits run out
Reaching the machinePublic IPv4, static or dynamic; TCP and UDP load balancers; no managed HTTPS URLsA public IPv4 you can keep static; open ports with firewall rules and run your own TLSToken-protected HTTPS URLs for up to 50 ports; raw TCP; dedicated IPv6 or IPv4
Network rulesVPC firewall with allow rules by IP, port and protocol, in and outPer-VM firewall: inbound closed by default, outbound rules by IP, protocol and portOpen internet out; inbound through HTTPS URLs and TCP ports
GPUsNvidia L40S to GB200, plus AMD MI300X and MI355X; on-demand, spot or reservedNvidia RTX A4000 to H100, H200, B200 and B300; on-demand, spot or reservedNone
BillingPer second while running, no plan fee, no data transfer charges todayPer minute from a prepaid balance, no bandwidth chargesMonthly plan from $20, all of it usage credit spent per second on running sandboxes
RegionsUS (Texas, Virginia, Ohio, Nevada) and EU (Iceland, Norway)Norway, Canada (Quebec) and the US (Texas)EU only: Germany, Finland, France
Every feature Battleships tracks (43)
FeatureCrusoeHyperstackbochka
Full VM (own kernel)YesYesYes
Docker insideYesYesYes
Nested virtualizationNoNot publishedYes
RootYesYesYes
SystemdNot publishedNot publishedYes
Desktop GUINoNoYes
BrowserNoNoYes
Computer-use APINoNoYes
SSHYesYesYes
Persistent diskYesYesYes
Pause/resumeYesYesYes
Automatic snapshotsNoNoYes
Fork/cloneYesNoYes
Live fork (no pause)NoNoYes
Memory forkNoNoNo
Custom image (Docker or snapshot)YesYesYes
≥24 h sessionsYesYesYes
Idle auto-stopNoNoNo
Wake on requestNoNoNo
HTTPS preview URLsNoNoYes
Raw TCP inboundYesNoYes
Public IPv4YesYesNo
Egress allowlistYesYesNo
Secret proxyNoNoYes
Secret proxy for any APINoNoYes
Preinstalled agentsNoNoYes
Agent harness APINoNoYes
MCP serverYesYesNo
GPUYesYesNo
WindowsNoNoNo
Arm64YesNoNo
Self-hostingNoNoNo
BYOCNoNoNo
Open sourceNoNoNo
Scoped API keysNot publishedNot publishedYes
Spend limitsNoNot publishedYes
Usage per sandbox (API)YesNoYes
Audit logsYesNot publishedNo
SSOYesNoNo
SOC 2YesYesNo
HIPAAYesNoNo
EU data residencyYesYesYes
Zero data retentionNot publishedNot publishedYes

Performance

CrusoeNot measured
HyperstackNot measured
bochka26.69 runs/s

There is no comparable CPU measurement for Crusoe or Hyperstack: neither was in the StarSling hpc-sandbox-benchmarks run we use. bochka did 26.69 Node.js build-and-test runs per second at 4 vCPU / 8 GB in that run, on dedicated AMD Ryzen 9 9950X (Zen 5) hosts. Crusoe's general-purpose CPU VMs run on AMD Genoa or Turin processors.

Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Crusoe and Hyperstack have 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.

ScenarioCrusoe, a monthHyperstack, 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
$32 · On-demand$114 · Prepaid on-demand$20 · $20/mo
One agent that never stops
4 vCPU / 8 GB, 50 GB disk, running 24/7 all month
$121 · On-demand$475 · Prepaid on-demand$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
$1,488 · On-demand$5,720 · Prepaid on-demand$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
$405 · On-demand$1,750 · Prepaid on-demand · with limits$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
$811 · On-demand$3,500 · Prepaid on-demand · with limits$2,000 · $2000/mo · with limits

Crusoe: Crusoe bills each VM per second while it runs, at a flat rate per vCPU that includes 4 GB of memory per vCPU, and every VM also pays for a 128 GB OS disk each month, running or stopped. It has no plan fee and does not charge for data transfer today.

Hyperstack: Hyperstack bills a running VM per minute from a prepaid balance, at a fixed hourly price per flavor, with no bandwidth charges. A stopped VM keeps billing in full, so the totals assume you hibernate it between runs.

Hyperstack: 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 whole disk.

Limits in those scenarios

Neither sells a 4 vCPU / 8 GB CPU VM: the scenarios use Crusoe's 4 vCPU / 16 GB machine and Hyperstack's 8 vCPU / 8 GB flavor, so part of the price gap is size. Crusoe sets CPU and GPU quotas per project and raises them through support, so 50 agents or 300 user sandboxes may need a ticket first. Hyperstack publishes no limit on VMs at once, so 300 user sandboxes and 1,000 short jobs at once are not confirmed there. On bochka the limit is a sandbox count plus starts per minute, so the 1,000-job burst gets queued and needs more starts per day than the $2,000 plan allows.

ScenarioCrusoeHyperstackbochka
One coding agent, workdaysNone that biteNone that biteNone that bite
One agent that never stopsNone that biteNone that biteNone that bite
50 agents in parallelNone that bite
  • how many machines can run at once: limit not published, probably fine at 50 (the typical published limit is 50)
None that bite
Sandboxes for your usersNone that bite
  • not confirmed: how many machines can run at once (you need 300; the typical published limit is 50)
None that bite
1,000 short jobs at onceNone that bite
  • not confirmed: how many machines can run at once (you need 1,000; the typical published limit is 50)
  • 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 Crusoe when

you need AMD GPUs or Nvidia GB200, Slurm and shared disks for cluster jobs, or CPU VMs that cost little while stopped.

  • AMD GPUs and the newest Nvidia racks. If you want AMD MI300X or MI355X, or Nvidia GB200, Crusoe rents them; Hyperstack lists only Nvidia.
  • Cheaper to leave stopped. For VMs that sit idle between jobs, a stopped Crusoe VM pays only for its OS disk, with no hibernate step.
  • Slurm and shared disks. For cluster jobs, Crusoe adds Slurm images and shared network disks next to managed Kubernetes.

Use Hyperstack when

you want a wide range of Nvidia cards by the minute, a region in Canada, or snapshots you take by hand.

  • A wide range of Nvidia cards. For GPU evals and fine-tuning, Hyperstack rents single cards from an RTX A4000 up to 8-GPU B300 nodes, by the minute.
  • Canada as well as the US and Norway. If your data or users are in Canada, Hyperstack has a region in Quebec; Crusoe lists no Canadian region.
  • Snapshots you can take by hand. If you want to save a VM's disk now and then, Hyperstack takes snapshots on request, with the VM shut off.

Where bochka fits

If what you are really running is agents, not model training, both Crusoe and Hyperstack hand you a bare Ubuntu VM and leave the rest to you. bochka gives each agent a full VM that saves itself every minute, forks while running, and comes with a desktop, coding agents, HTTPS hosting and a keys proxy. Keep GPU work on Crusoe or Hyperstack and run the agent's workspace on bochka.

Work that saves itself and forks live

An agent that breaks its machine needs a recent copy to roll back to. Crusoe can only image a shut-off boot disk, and Hyperstack snapshots by hand with the VM off, so both halt the agent to save or branch it.

Free stops with the whole disk kept

Agents sit idle between tasks. A stopped Hyperstack VM bills in full unless you hibernate it and hope its flavor is in stock to restore; a stopped Crusoe VM still pays for its OS disk.

A desktop, agents and keys handled for you

Neither ships a desktop, coding agents or a credentials proxy, so you build a desktop stream, install each agent, write the code that drives it, and leave keys on a machine the agent's code can read.

Use bochka when your agents write and run code for hours on a full VM that saves itself every minute, forks while running, costs nothing while stopped, and comes with a desktop, HTTPS URLs and coding agents built in. Compare it one to one: bochka vs Crusoe and bochka vs Hyperstack.

Questions people ask

Is Crusoe cheaper than Hyperstack?

For CPU agent work in our scenarios, yes: one coding agent on workdays comes to $32 a month on Crusoe and $114 on Hyperstack, one agent running all month to $121 versus $475, and 50 agents in parallel to $1,488 versus $5,720. The VM sizes differ, so part of the gap is size. On bochka the same three come to $20, $26 and $317 a month.

Which is faster, Crusoe or Hyperstack?

We have no comparable CPU measurement for either one: neither was in the StarSling benchmark run we use. bochka did 26.69 Node.js build-and-test runs per second at 4 vCPU / 8 GB in that run.

Do stopped VMs cost money on Crusoe and Hyperstack?

On Crusoe a stopped VM has no compute charge, but its 128 GB OS disk bills until you delete the VM. On Hyperstack a stopped VM bills at the full rate; a hibernated one bills only its disk and any kept IP, and restoring it needs the flavor in stock. A stopped bochka sandbox costs nothing.

Can Crusoe and Hyperstack run Docker?

Yes. Both are full VMs: Crusoe has Docker preinstalled on its GPU images, and Hyperstack's Ubuntu images come with or without Docker. bochka sandboxes come with Docker, Docker Compose, systemd and /dev/kvm ready.

Which has better GPUs for evals?

In our GPU eval scenario Hyperstack ranks 2nd of 98 at $625 a month, against $7,803 for Crusoe. Crusoe carries AMD MI300X and MI355X and Nvidia GB200, which Hyperstack does not list. bochka has no GPUs.

Where do Crusoe and Hyperstack run?

Crusoe runs in the US (Texas, Virginia, Ohio, Nevada) and in Iceland and Norway. Hyperstack runs in Norway, Quebec and Texas. bochka runs only in the EU, in Germany, Finland and France.

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.