Nebius vs Crusoe
Nebius and Crusoe both rent GPU machines and plain CPU VMs that you set up yourself. Nebius is Nvidia only, costs less for CPU VMs in every scenario we priced and runs in mainland Europe, the UK, Israel and the US, while Crusoe adds AMD GPUs and GB200 and runs in the US, Iceland and Norway.
The companies and products
Nebius
Nebius is an AI cloud company, listed on Nasdaq and based in Amsterdam, that rents GPUs and the infrastructure around them.
You create a Compute VM from a preset (GPU or CPU only), add disks and an SSH key, and get a full Linux VM with root and an optional public IP. Around it Nebius sells Managed Kubernetes, Slurm clusters (Soperator), Serverless AI jobs and endpoints, object storage and networking. It is built for teams that train, fine-tune and serve AI models, not as a sandbox for agents.
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.
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. “29th cheapest of 364” means 28 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 workload | Nebius: price rank | Crusoe: price rank | bochka: price rank |
|---|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month Both are a compromise for a software factory: no coding agents, no desktop, and forks by hand, though Nebius can snapshot a running disk and Crusoe must shut the VM off first. | 197th of 367, with limits $2,272 a month | 232nd of 367, with limits $2,996 a month | 3rd cheapest of 367 $634 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB | 210th of 366, with limits $2,261 a month | 234th of 366, with limits $3,000 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 Both GPU clouds beat bochka here, at $202 a month on Nebius and $268 on Crusoe, because bochka's starts per minute make a big burst a compromise. | 111th of 345, with limits $202 a month | 128th of 345, with limits $268 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 300 user sandboxes start above Nebius's default quota of 12 VMs per region and likely need a support ticket on Crusoe, which publishes no default quotas; bochka ranks first here. | 81st of 361, with limits $716 a month | 104th of 361, with limits $951 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 | 52nd of 360, with limits $3,049 a month | 71st of 360, with limits $4,055 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 | 29th cheapest of 364 $610 a month | 43rd cheapest of 364 $811 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 work as plain CI runners, with Nebius 27th and Crusoe 48th; bochka ranks 4th and adds snapshots every minute and free stops between jobs. | 27th cheapest of 377 $297 a month | 48th cheapest of 377 $395 a month | 4th cheapest of 377 $86 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB A 24/7 dev box is a 16 GiB machine at $91 a month on Nebius and $121 a month on Crusoe, since both sell 4 GB of memory per vCPU. | 68th cheapest of 368 $91 a month | 87th cheapest of 368 $121 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 Nebius nor Crusoe has a desktop or computer-use product, so you would install and stream a desktop yourself; bochka ships a 60fps desktop with Chromium. | 187th of 369, with limits $610 a month | 221st of 369, with limits $811 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 | 223rd of 367, with limits $305 a month | 246th of 367, with limits $405 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 | 253rd of 346, with limits $302 a month | 268th of 346, with limits $403 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 | 174th of 338, with limits $6,049 a month | 201st of 338, with limits $8,055 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 | 176th of 360, with limits $156 a month | 201st of 360, with limits $207 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 Nebius ranks 18th of 98 for GPU eval jobs at $1,127 a month, while Crusoe ranks 86th at $7,803 a month because its H100 and H200 machines come only as 8-GPU VMs; bochka has no GPUs. | 18th cheapest of 98 $1,127 a month | 86th of 98, with limits $7,803 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 it | Can't run it | Can't run it |
Key features
| Feature | Nebius | Crusoe | bochka |
|---|---|---|---|
| Kind of product | AI cloud: GPU and CPU VMs, Managed Kubernetes, Slurm clusters and serverless AI jobs | GPU and CPU cloud with managed Kubernetes, Slurm and managed inference | Sandboxes for AI agents: full VMs with agents, desktop and hosting ready |
| Isolation | Full VM behind a hypervisor, with its own kernel and root | Full VM on cloud-hypervisor | Full VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts |
| GPUs | Nvidia L40S, RTX PRO 6000, H100, H200, B200 and B300 | Nvidia L40S, A100, H100, H200, B200, B300 and GB200, plus AMD MI300X and MI355X | None |
| Spot GPUs | Self-serve preemptible VMs at a spot price that updates every 15 minutes | Spot quoted through sales; reservations bill for the whole term | None |
| Root, Docker and nested virtualization | Root on your VM and container VMs from Docker images; nested virtualization not documented | sudo user (root login deprecated); Docker preinstalled on GPU images; nested virtualization not documented | Root, systemd, Docker and /dev/kvm on every machine |
| Snapshots and fork | Manual disk snapshots of a running VM; fork by making a disk from the snapshot, then a VM | Custom image of the boot disk, only while the VM is shut off; no disk snapshots | Snapshots every minute and on stop; fork a running sandbox in one call |
| Stopped machines | Compute stops billing; disks keep billing on their full provisioned size | Compute stops billing; the 128 GB OS disk bills until the VM is deleted | A stopped sandbox costs nothing and keeps its disk |
| Desktop and computer use | None; a text serial console | None; install and stream a desktop yourself | 60fps desktop and Chromium, computer use and browser control |
| Coding agents | None built in; you install agents over SSH | None; an MCP server lets an assistant inspect your Crusoe infrastructure | Claude Code, Codex, OpenCode and more preinstalled, with a prompt, steer and events API |
| Keeping keys out of the machine | SecretStash stores secrets that the VM fetches with its service account | No credentials proxy; secrets live on the VM as files or variables | AI credentials proxy for model providers and any HTTPS API you register |
| Reaching the machine | Optional public IPv4, SSH and Nebius Tunnels; you set up TLS yourself | Static or dynamic public IPv4, SSH, TCP and UDP load balancers; no managed HTTPS URLs | HTTPS URLs for up to 50 ports, token-protected; raw TCP; dedicated IPv6 or IPv4 |
| Network rules | VPC security groups allow or deny traffic in and out by IP range, port and protocol | VPC firewall allow rules by IP, port and protocol, in and out | Open internet on every plan; no provider-side allowlist |
| Regions | Finland, France, Spain, United Kingdom, Israel and the US (Missouri, Minnesota) | US (Texas, Virginia, Ohio, Nevada) and EU (Iceland, Norway) | EU only: Germany, Finland, France |
| Quotas | Published default: 12 VMs and, in most regions, 200 vCPU per region; raise on request | Per-project quotas, none published; raise them through support | Self-serve plans with published sandbox counts and start rates |
Every feature Battleships tracks (43)
| Feature | Nebius | Crusoe | bochka |
|---|---|---|---|
| Full VM (own kernel) | Yes | Yes | Yes |
| Docker inside | Yes | Yes | Yes |
| Nested virtualization | Not published | No | Yes |
| Root | Yes | Yes | Yes |
| Systemd | Not published | Not published | Yes |
| Desktop GUI | No | No | Yes |
| Browser | No | No | Yes |
| Computer-use API | No | No | Yes |
| SSH | Yes | Yes | Yes |
| Persistent disk | Yes | Yes | Yes |
| Pause/resume | Yes | Yes | Yes |
| Automatic snapshots | No | No | Yes |
| Fork/clone | Yes | Yes | Yes |
| Live fork (no pause) | Yes | No | Yes |
| Memory fork | No | No | No |
| Custom image (Docker or snapshot) | Yes | Yes | Yes |
| ≥24 h sessions | Yes | Yes | Yes |
| Idle auto-stop | No | No | No |
| Wake on request | No | No | No |
| HTTPS preview URLs | Yes | No | Yes |
| Raw TCP inbound | Not published | Yes | Yes |
| Public IPv4 | Yes | Yes | No |
| Egress allowlist | Yes | Yes | No |
| Secret proxy | No | No | Yes |
| Secret proxy for any API | No | No | Yes |
| Preinstalled agents | No | No | Yes |
| Agent harness API | No | No | Yes |
| MCP server | No | Yes | No |
| GPU | Yes | Yes | No |
| Windows | No | No | No |
| Arm64 | No | Yes | No |
| Self-hosting | No | No | No |
| BYOC | No | No | No |
| Open source | No | No | No |
| Scoped API keys | Not published | Not published | Yes |
| Spend limits | Not published | No | Yes |
| Usage per sandbox (API) | Yes | Yes | Yes |
| Audit logs | Not published | Yes | No |
| SSO | Yes | Yes | No |
| SOC 2 | Yes | Yes | No |
| HIPAA | Yes | Yes | No |
| EU data residency | Yes | Yes | Yes |
| Zero data retention | Not published | Not published | Yes |
Performance
There is no comparable measurement for Nebius or Crusoe: neither was part of the StarSling hpc-sandbox-benchmarks run we use, so we cannot say which CPU VM is faster. 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. Nebius CPU VMs run on AMD EPYC Genoa or Intel Ice Lake, and Crusoe's on AMD Genoa or Turin.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Nebius and Crusoe 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.
| Scenario | Nebius, a month | Crusoe, a month | bochka, 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 | $25 · Pay as you go | $32 · On-demand | $20 · $20/mo |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $91 · Pay as you go | $121 · 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 | $1,127 · Pay as you go | $1,488 · 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 | $305 · Pay as you go | $405 · On-demand | $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 | $610 · Pay as you go | $811 · On-demand | $2,000 · $2000/mo · with limits |
Nebius: Nebius bills running VMs per second for vCPU and RAM, with no plan fee: from 1 October 2026, AMD Genoa costs $0.015 per vCPU-hour plus $0.0045 per GiB-hour. Disks bill by provisioned size even while the VM is stopped.
Crusoe: Crusoe bills each VM per second while it runs, at a flat rate per vCPU that includes 4 GB of memory, and every VM pays for a 128 GB OS disk each month, running or stopped. There is no plan fee and no data transfer charge today; reservations bill for the whole term.
Crusoe: 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
Both sell CPU VMs with 4 GB of memory per vCPU, so the 4 vCPU rows use 16 GiB machines on both against bochka's 8 GB sandbox. Nebius starts each region at 12 VMs and, in most regions, 200 vCPU across CPU-only VMs, while Crusoe publishes no default quotas and raises them through support, so 50 agents, 300 user sandboxes and 1,000 short jobs at once likely need a bigger quota on both first. On bochka the limit is a sandbox count by plan plus starts per minute, so the 1,000-job burst is queued and needs more starts per day than the $2,000 plan allows.
| Scenario | Nebius | Crusoe | bochka |
|---|---|---|---|
| One coding agent, workdays | None that bite | None that bite | None that bite |
| One agent that never stops | None that bite | None that bite | None that bite |
| 50 agents in parallel | None that bite | None that bite | None that bite |
| Sandboxes for your users | None that bite | None that bite | None that bite |
| 1,000 short jobs at once | None that bite | None that bite |
|
Which one should you use?
Use Nebius when
you want lower CPU VM rates, self-serve spot GPUs, snapshots of a running disk, or a region in Spain, France, the UK or Israel.
- Lower rates on CPU machines. For CPU-only work, Nebius came in below Crusoe in every scenario we priced, from one agent to 1,000 short jobs.
- Snapshots without stopping, and self-serve spot GPUs. Useful if you checkpoint often: Nebius snapshots a running disk, and its preemptible GPUs need no sales call.
- More regions in Europe and beyond. Useful if you need Spain, France, the UK or Israel: Nebius runs there as well as in Finland and the US.
Use Crusoe when
you want AMD GPUs or GB200, several US sites or the Nordics, or TCP and UDP load balancers with no data transfer charge.
- AMD GPUs and GB200. Useful if you want AMD MI300X or MI355X, Nvidia GB200 or A100: Crusoe rents them and Nebius does not.
- Several US sites and the Nordics. Useful if you want Texas, Virginia, Ohio or Nevada, or Iceland and Norway for your data.
- Load balancers and free data transfer. Useful for serving traffic: Crusoe has TCP and UDP load balancers and does not charge for data transfer today.
Where bochka fits
Nebius and Crusoe are GPU clouds, and bochka has no GPUs, so model training and GPU evals belong on one of them. If what you are really running is agents on CPU machines, both hand you a plain Ubuntu VM and leave the agent, desktop, snapshots and hosting to you. bochka ships those ready on every sandbox.
Work that saves itself and forks live
On Nebius the last good state is your last manual snapshot, and a fork means a new disk, then a new VM. On Crusoe you must shut the VM off to image it, which halts the agent.
A desktop and coding agents inside
Neither ships a desktop, a browser or coding agents: Nebius gives SSH and a serial console, Crusoe a bare Ubuntu server. You build the desktop stream, install each agent and write the runner that prompts and watches it.
Keys and previews handled for you
Nebius SecretStash hands the secret to the VM and Crusoe keeps it on the VM, so agent code can read it. Neither gives HTTPS URLs per machine, so previews need your own proxy, certificates and auth.
Use bochka when what you run is agents: full VMs that save themselves every minute, fork while running, cost nothing while stopped, and come with a desktop, coding agents, HTTPS hosting and a keys proxy. Compare it one to one: bochka vs Nebius and bochka vs Crusoe.
Questions people ask
Is Nebius cheaper than Crusoe?
For CPU VMs, yes in every scenario we priced. One agent running all month is $91 a month on Nebius and $121 on Crusoe, and 50 agents in parallel are $1,127 against $1,488. bochka comes to $26 and $317 a month for the same two.
Which is faster, Nebius or Crusoe?
We have no comparable measurement: neither was in the StarSling benchmark run we use. Both sell CPU VMs on AMD Genoa, and Crusoe also offers Turin. bochka did 26.69 Node.js build-and-test runs per second at 4 vCPU / 8 GB in that run.
Can Nebius and Crusoe run Docker?
Yes. Nebius gives you root on your VM and can start container VMs from Docker images, and Crusoe gives you a sudo user with Docker preinstalled on its GPU images. Neither documents nested virtualization; bochka has Docker, systemd and /dev/kvm on every machine.
Can I snapshot or fork a VM on Nebius or Crusoe?
On Nebius you take disk snapshots of a running VM by hand, then make a new disk and VM to fork. On Crusoe you shut the VM off and make a custom image of its boot disk, with no snapshots of data disks. bochka snapshots every minute and forks a running sandbox in one call.
Which has more GPU types, Nebius or Crusoe?
Crusoe has more: Nvidia L40S, A100, H100, H200, B200, B300 and GB200, plus AMD MI300X and MI355X. Nebius rents Nvidia L40S, RTX PRO 6000, H100, H200, B200 and B300. bochka has no GPUs.
Where do Nebius and Crusoe run?
Nebius runs in Finland, France, Spain, the United Kingdom, Israel and two US regions. Crusoe runs in Texas, Virginia, Ohio and Nevada, plus Iceland and Norway. bochka runs only in the EU (Germany, Finland and France), with EU data residency.
Try bochka
Install the CLI, then run bochka new.
Product preview