boat vs OVHcloud GPU
OVHcloud GPU instances are cloud VMs with an NVIDIA card attached, built for training and inference. boat is a sandbox for AI agents, with coding agents, a browser desktop and a disk that is saved every minute, but no GPU.
Where boat stands out
Coding agents and keys built in
Claude Code, Codex and five other agents come preinstalled, and you prompt, steer and follow them from outside through one API. Real API keys stay outside the sandbox in a credentials proxy, and a credential can switch to another account when one runs out of credits.
A desktop and browser for computer use
Every boat sandbox can stream a Linux desktop at 60fps with Chromium, and agents can see and click it through the computer-use API. You can embed the desktop in your own app.
Stops that keep the disk, every minute
boat saves the disk every minute and when a sandbox stops, and a stopped sandbox costs nothing. You can fork a running sandbox with its disk in one call.
The companies and products
boat
boat.dev runs sandboxes for AI agents: each one is a full Linux VM with its own kernel.
You start a machine with one command or one API call, then SSH in, run Docker, open a desktop, or let a built-in agent (Claude Code, Codex, OpenCode and others) work in it. The disk stays between sessions, and a stopped sandbox costs nothing.
- Isolation
- Full VM per sandbox, own kernel
- Regions
- EU: Germany, Finland, France
OVHcloud GPU
OVHcloud is a French cloud provider, and its GPU instances are Public Cloud VMs with one or more NVIDIA GPUs attached.
Each GPU instance is an OpenStack VM on the KVM hypervisor, with the GPU given to the VM through PCI passthrough so the guest controls it like a physical card. You pick a flavor such as l4-90, l40s-90 or h100-380, which fixes the GPU, vCores, RAM and disk, and run Linux or Windows on it. It is a general GPU cloud for machine learning, rendering and HPC, not an agent sandbox, so this page compares the parts where the two overlap.
- Kind of product
- GPU cloud VMs (IaaS), not a managed agent sandbox
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. “3rd cheapest of 367” means 2 of the other 366 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 | boat: price rank | OVHcloud GPU: price rank |
|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month A coding agent needs CPU, a disk and Docker, not a GPU, so on OVHcloud you pay for a graphics card the agent never uses. | 3rd cheapest of 367 $634 a month | 352nd of 367, with limits $15,488 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB An always-on assistant on OVHcloud runs on a GPU flavor all month, while boat runs it on a small VM with agents and a desktop built in. | 5th cheapest of 366 $657 a month | 358th of 366, with limits $32,120 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month | 186th of 345, with limits $2,000 a month | 189th of 345, with limits $2,933 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month Giving each of your users a sandbox needs fast starts, free stops and hosting URLs, which boat has and plain GPU VMs leave to you. | Cheapest of 361 $209 a month | 185th of 361, with limits $10,230 a month |
| RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month | 34th of 360, with limits $2,000 a month | 205th of 360, with limits $44,000 a month |
| Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month | 10th cheapest of 364 $200 a month | 209th of 364, with limits $4,400 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month CI jobs start and stop many times a day; boat stops for free with the disk kept, while an OVHcloud VM is billed until you shelve or delete it. | 4th cheapest of 377 $86 a month | 104th cheapest of 377 $1,056 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB Both give you a full VM with SSH that runs as long as you like, but OVHcloud's cheapest fit here still carries a GPU. | 18th cheapest of 368 $26 a month | 133rd of 368, with limits $642 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month boat streams a Linux desktop at 60fps with Chromium for computer use; OVHcloud gives you a bare VM, or RDP on its Windows images, and you build the rest. | 6th cheapest of 369 $200 a month | 345th of 369, with limits $4,400 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month | 7th cheapest of 367 $200 a month | 353rd of 367, with limits $4,400 a month |
| Scraping & bots (startup scale) 50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month | 6th cheapest of 346 $90 a month | 341st of 346, with limits $4,400 a month |
| Mass scraping & bots (undetectable) 1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month | 71st of 338, with limits $1,800 a month | 329th of 338, with limits $88,000 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month | 2nd cheapest of 360 $200 a month | 337th of 360, with limits $2,200 a month |
| GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month This is the one use case where OVHcloud wins outright: it rents NVIDIA GPUs from L4 to H200, and boat has no GPUs. | Can't run it | 5th cheapest of 98 $748 a month |
| 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 |
Key features
Every feature Battleships tracks (43)
| Feature | boat | OVHcloud GPU |
|---|---|---|
| Full VM (own kernel) | Yes | Yes |
| Docker inside | Yes | Yes |
| Nested virtualization | Yes | Not published |
| Root | Yes | Not published |
| Systemd | Yes | Not published |
| Desktop GUI | Yes | No |
| Browser | Yes | No |
| Computer-use API | Yes | No |
| SSH | Yes | Yes |
| Persistent disk | Yes | Yes |
| Pause/resume | Yes | Yes |
| Automatic snapshots | Yes | Yes |
| Fork/clone | Yes | Yes |
| Live fork (no pause) | Yes | Yes |
| Memory fork | No | No |
| Custom image (Docker or snapshot) | Yes | Yes |
| ≥24 h sessions | Yes | Yes |
| Idle auto-stop | No | No |
| Wake on request | No | No |
| HTTPS preview URLs | Yes | Yes |
| Raw TCP inbound | Yes | Not published |
| Public IPv4 | No | Yes |
| Egress allowlist | No | Yes |
| Secret proxy | Yes | No |
| Secret proxy for any API | Yes | No |
| Preinstalled agents | Yes | No |
| Agent harness API | Yes | No |
| MCP server | No | No |
| GPU | No | Yes |
| Windows | No | Yes |
| Arm64 | No | No |
| Self-hosting | No | No |
| BYOC | No | No |
| Open source | No | No |
| Scoped API keys | Yes | Not published |
| Spend limits | Yes | Not published |
| Usage per sandbox (API) | Yes | Yes |
| Audit logs | No | Not published |
| SSO | No | Yes |
| SOC 2 | No | Yes |
| HIPAA | No | Yes |
| EU data residency | Yes | Yes |
| Zero data retention | Yes | Not published |
Performance
There is no comparable measurement for OVHcloud GPU instances. They were not part of the hpc-sandbox-benchmarks CPU run where boat scored 26.69 runs per second at 4 vCPU / 8 GB, and a GPU card does not speed up the CPU-bound builds and tests that agents spend most of their time on.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. OVHcloud GPU 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.
| Scenario | boat, a month | OVHcloud GPU, 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 | $20 · $20/mo | $155 · Pay as you go · with limits |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $642 · Pay as you go · with limits |
| 50 agents in parallel 50 machines of 4 vCPU / 8 GB at once, 8-hour sessions, 1,100 sessions a month, Docker inside | $317 · $20/mo × 3 seats | $7,744 · Pay as you go · with limits |
| Sandboxes for your users 2 vCPU / 4 GB, 30,000 sessions of 10 minutes a month, up to 300 at once | $200 · $100/mo × 2 seats | $4,400 · Pay as you go · with limits |
| 1,000 short jobs at once 2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month | $2,000 · $2000/mo · with limits | $8,800 · Pay as you go · with limits |
boat: Plans start at $20 a month, and the whole fee is usage credit. You pay per second while a sandbox runs, and a stopped sandbox costs nothing and keeps its disk.
OVHcloud GPU: Each GPU flavor has one hourly price that covers the GPU, vCores, RAM and local disk, or a monthly price you choose at launch and cannot switch back from. The VM bills until you shelve or delete it (on monthly billing, unshelving even briefly bills the whole month), and Savings Plans take up to 30% off for a 12 or 36 month commitment.
Limits in those scenarios
The cheapest OVHcloud flavor the estimator could use for these scenarios is t2-45, a V100S GPU VM with 15 vCores and 45 GB at $0.88 an hour, so every scenario pays for a GPU the agent does not use. The table marks them as not fitting because that smallest size is far bigger than each scenario needs. In practice the bigger limits are GPU stock in GRA11 and project quotas, which a new account may need to raise by request. boat has no session cap, but at 1,000 short jobs at once its $2,000 plan caps starts per hour and per day and queues big bursts.
| Scenario | boat | OVHcloud GPU |
|---|---|---|
| One coding agent, workdays | None that bite |
|
| One agent that never stops | None that bite |
|
| 50 agents in parallel | None that bite |
|
| Sandboxes for your users | None that bite |
|
| 1,000 short jobs at once |
|
|
Which one should you use?
Use OVHcloud GPU when
the job needs an NVIDIA GPU, such as training, fine-tuning, inference or rendering, or when you need Windows with a GPU.
- Real NVIDIA GPUs. For training, fine-tuning, inference or rendering, it rents NVIDIA GPUs from L4 up to eight H200s in one VM.
- Windows GPU machines. Teams that need Windows with a GPU can launch special Windows images and connect over RDP.
- Network security groups and Canada. It can filter traffic in and out by port and IP range, and it also runs GPUs in Canada.
Use boat when
your agents write code, run builds and tests, use Docker or drive a browser, and you want them on a fast CPU machine with agents, a desktop and free stops built in.
- Coding agents and keys built in. An OVHcloud VM starts as a plain OS image. You install the agent, write the harness that drives it, and leave real API keys on a machine the agent controls.
- A desktop and browser for computer use. OVHcloud offers no managed desktop or browser on its GPU instances. You would install a desktop, a VNC server and a browser on each VM, then keep them working.
- Stops that keep the disk, every minute. On OVHcloud a paused or suspended VM bills at the full GPU rate. Shelving bills only the snapshot, but it wipes the NVMe passthrough drives on some flavors, and forking means a backup and a new instance.
Questions people ask
Is boat cheaper than OVHcloud GPU instances?
For agent work without a GPU, yes: boat costs less in all five scenarios on this page, though neither fully fits 1,000 short jobs at once. One coding agent on workdays costs $20 a month on boat against $155 on OVHcloud, and an agent that never stops costs $26 against $642, because every OVHcloud flavor includes a GPU.
Does boat have GPUs?
No. boat runs CPU sandboxes on dedicated AMD Ryzen 9 9950X hosts. For GPU work such as training or inference, a GPU cloud like OVHcloud is the right tool.
Can I run AI coding agents on an OVHcloud GPU instance?
Yes, it is a full VM with SSH, so you can install any agent. You build the harness, the browser, the desktop and the key handling yourself, which boat ships ready to use.
What happens to an OVHcloud GPU instance when I stop it?
Shelving saves the disk as a snapshot and bills only that snapshot, but it wipes the NVMe passthrough drives on T1/T2-180 flavors. Pausing or suspending keeps billing at the full rate. A boat sandbox that stops costs nothing and keeps its disk.
Do both keep data in Europe?
Yes. boat runs only in the EU (Germany, Finland, France). OVHcloud GPU instances run in Gravelines, France, and in Beauharnois, Canada, so you pick the region at launch.
Is boat SOC 2 compliant?
SOC 2 is in progress. We're open about where we stand: live status at status.ascii.dev and our data retention policy at docs.boat.dev/data-retention.
Start a sandbox
Install the CLI, then run boat new.
7-day free trial