boat vs IBM Cloud VPC

IBM Cloud VPC rents Linux and Windows virtual servers in 13 regions, and you build the tooling around them. boat gives each agent a full VM that already has Docker, /dev/kvm, forks of running machines and coding agents you drive by API.

Where boat stands out

Docker and /dev/kvm on every machine

Every boat sandbox is a full VM with root, systemd, Docker and nested virtualization, so /dev/kvm is there for Android emulators, Firecracker or QEMU. IBM's docs say nested virtualization is not supported on its virtual servers. IBM Bare Metal Servers for VPC can run a hypervisor instead.

Coding agents you drive from outside

boat installs Claude Code, Codex, OpenCode and four other harnesses in each sandbox. You drive them from outside with a prompt, steer and events API. The optional paid credentials proxy keeps real AI keys out of the machine. An IBM server gives you SSH and the API to manage servers; you write the rest.

Fork a running machine in one call

boat forks a running sandbox from its last 1-minute snapshot, so writes from the last minute can be missing. boat saves the disk every minute and at each stop. A stopped sandbox has no compute charge, and boat keeps its disk for at least 14 days. On IBM a copy starts from a volume snapshot or image and boots a new server.

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.

IBM Cloud VPC

IBM Cloud VPC is IBM's general-purpose cloud for virtual servers: you rent Linux or Windows machines inside your own virtual private network.

It is a hyperscaler IaaS product, not a sandbox API. Each virtual server instance is a full VM with root and SSH, picked from balanced, compute, memory, Flex and GPU profiles. You set up networking, disks, images, firewalls and any agent tooling yourself. It suits teams that already run on IBM Cloud and want raw VMs they control end to end.

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. “5th cheapest of 365” means 4 of the other 364 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 workloadboat: price rankIBM Cloud VPC: price rank
Software factory
100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month
Both can host a software factory. IBM virtual servers have no nested virtualization, so emulators and microVMs need Bare Metal Servers for VPC. You also set up the images, HTTPS and agent driver yourself.
5th cheapest of 365
$634 a month
141st of 365, with limits
$966 a month
Always-on AI employees
50 machines 24/7 · 2 vCPU / 4 GB
An always-on assistant needs steady CPU. The IBM price uses a beta Flex server that guarantees 25% of each vCPU; a full-rate IBM profile costs more.
6th cheapest of 361
$657 a month
190th of 361, with limits
$874 a month
Code interpreter bursts
100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month
IBM fits short code runs well: a small burstable Flex server costs little when it is mostly idle, and starts are limited only by your vCPU quota. On boat, starts above the plan's included count cost extra.
189th of 336, with limits
$2,000 a month
62nd of 336, with limits
$32 a month
Sandboxes for your users
300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month
Giving each user a VM on IBM means you build the sandbox layer yourself (URLs, auth, cleanup). 300 servers of 2 vCPU need a quota raise past 200 vCPU per region. On boat, the $100 plan allows 300 sandboxes at once.
3rd cheapest of 349
$209 a month
62nd of 349, with limits
$428 a month
RL rollouts
1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month
IBM's burstable Flex servers and spot instances (up to 75% off, in early access) make RL fleets cheap if each rollout needs little CPU and can be preempted. A raise of the vCPU quota needs a support case.
39th of 349, with limits
$2,000 a month
25th of 349, with limits
$1,130 a month
Agent evals
250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month
Both fit. IBM suits eval batches you already script with its CLI or Terraform. boat lets you fork one prepared sandbox into many, within the plan's start limits.
10th cheapest of 359
$200 a month
13th cheapest of 359
$230 a month
CI runners
40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month
2nd cheapest of 374
$86 a month
57th cheapest of 374
$438 a month
24/7 dev boxes
1 machine 24/7 · 4 vCPU / 8 GB
19th cheapest of 371
$26 a month
47th cheapest of 371
$57 a month
Computer-use agents
50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month
On IBM you set it up: install a desktop, a VNC or RDP server and a browser on the server, then wire your own computer-use loop. On boat a 60 fps desktop and Chrome come with every machine.
7th cheapest of 367
$200 a month
153rd of 367, with limits
$253 a month
Browser-use agents
50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month
7th cheapest of 356
$200 a month
242nd of 356, with limits
$140 a month
Scraping & bots (startup scale)
50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month
6th cheapest of 343
$90 a month
245th of 343, with limits
$90 a month
Mass scraping & bots (undetectable)
1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month
73rd of 327, with limits
$1,800 a month
56th of 327, with limits
$1,349 a month
Parallel red teaming
200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month
3rd cheapest of 351
$200 a month
175th of 351, with limits
$69 a month
GPU eval jobs
10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month
Only IBM fits: it rents Nvidia L4 and L40S servers, plus A100, H100, H200 and Intel Gaudi 3 in select availability. boat has no GPUs.
Not ranked: misses a requirement30th cheapest of 102
$3,080 a month
macOS agents
5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month
Not ranked: misses a requirementNot ranked: misses a requirement

Key features

FeatureboatIBM Cloud VPC
What you getAn agent sandbox: a full VM per sandbox, created and driven by APIA general cloud virtual server; you set up images, networking and tooling
IsolationFull VM with its own kernelFull virtual server with its own kernel, on shared or dedicated hosts
Docker and nested virtualizationDocker, root, systemd and /dev/kvm on every machineRoot and Docker work; nested virtualization is not supported on virtual servers (Bare Metal Servers for VPC can run a hypervisor)
CPU you getFull VM on a host dedicated to boat (AMD Ryzen 7 9700X or Ryzen 9 9950X); vCPUs are shared, not dedicated coresFull-rate profiles, or beta Flex servers with a 10%, 25% or 50% CPU baseline that can burst to twice that
Desktop and computer use1920x1080 desktop and Chrome streams at 60 fps, with computer tools for agentsA VNC or serial console for troubleshooting. For a desktop, you set it up: install a desktop and a VNC or RDP server on the server
Files after a stopNo compute charge while stopped; boat keeps the disk for at least 14 days, and saved snapshots keep it longerVolumes are kept until you delete them; compute stops billing, but storage and floating IPs keep billing, and stopped servers count toward the device quota
SnapshotsAutomatic disk snapshots every minute and at each stop; memory and running programs are not savedVolume snapshots you take by hand, by API or on a backup policy schedule; disk only, no memory
Fork and cloneFork a running sandbox in one call, from its last 1-minute snapshotYou set it up: snapshot the boot volume or make an image, then create new servers that boot from it
Reaching the machineHTTPS URLs for up to 50 ports, token-gated by default; public IPv6; SSH over IPv4; no static IPv4Floating public IPv4 and any port through security groups; HTTPS through a load balancer you add, or certificates you run yourself
Outbound internetOpen internet on every plan; no provider-side allowlistSecurity groups and network ACLs allow or deny traffic by IP range, protocol and port
Keeping keys out of the machineOptional AI credentials proxy ($1 per day on plus $1 per 1,000 requests); with it on, real keys of supported providers stay out of the sandboxSecrets Manager stores secrets that the server reads, so code inside can see them; a key-injecting proxy is yours to set up
Coding agentsSeven harnesses installed (Claude Code, Codex, OpenCode and more), with a prompt, steer and events APINone built in. You set it up: install the agent CLIs with cloud-init or a custom image, and write your own driver
GPUs and WindowsNo GPUs; Linux only, with Wine for Windows programsNvidia L4 and L40S servers, plus A100, H100, H200 and Intel Gaudi 3 in select availability; Windows Server 2016 to 2025 images
RegionsGermany, Finland and France; no US or Asia region, and you cannot pick one13 regions across the Americas, Europe, Japan, Australia and India
Compliance and transparencySOC 2 is in progress. We're open about where we stand: live status and data retention policy.SOC 2
Every feature Battleships tracks (43)
FeatureboatIBM Cloud VPC
Full VM (own kernel)YesYes
Docker insideYesYes
Nested virtualizationYesNo documented way
RootYesYes
SystemdYesYes
Desktop GUIYesNo documented way
BrowserYesNo documented way
Computer-use APIYesNo documented way
SSHYesYes
Persistent diskYesYes
Pause/resumeYesYes
Automatic snapshotsYesNo documented way
Fork/cloneYesNo documented way
Live fork (no pause)YesNo documented way
Memory forkNo documented wayNo documented way
Custom image (Docker or snapshot)YesYes
≥24 h sessionsYesYes
Idle auto-stopNo documented wayNo documented way
Wake on requestNo documented wayNo documented way
HTTPS preview URLsYesNo documented way
Raw TCP inboundYesYes
Public IPv4No documented wayYes
Egress allowlistNo documented wayYes
Secret proxyYesNo documented way
Secret proxy for any APIYesNo documented way
Preinstalled agentsYesNo documented way
Agent harness APIYesNo documented way
MCP serverNo documented wayNo documented way
GPUNo documented wayYes
WindowsNo documented wayYes
Arm64No documented wayNo documented way
Self-hostingNo documented wayNo documented way
BYOCNo documented wayNo documented way
Open sourceNo documented wayNo documented way
Scoped API keysYesNot published
Spend limitsYesNot published
Usage per sandbox (API)YesNo documented way
Audit logsNo documented wayNot published
SSONo documented wayYes
SOC 2No documented wayYes
HIPAANo documented wayYes
EU data residencyNo documented wayYes
Zero data retentionYesNot published

Performance

boat28.06 runs/s
IBM Cloud VPCNot measured

IBM Cloud VPC is not in the public benchmarks we use, so we cannot say which is faster. In StarSling's run of 3 October 2026 (run 37073680030), boat did 28.06 V8 Web Tooling runs/s at 4 vCPU / 8 GB, second of 15. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure. IBM's speed depends on the profile you pick, and a Flex server's guaranteed share depends on its CPU baseline.

V8 Web Tooling Benchmark at a 4 vCPU / 8 GB target, runs per second, higher is better. Source: StarSling's run of 3 October 2026 (run 37073680030), median of 5 sandboxes. Providers ran on different hardware in different regions. StarSling gives a shared rank when the run cannot separate two providers. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure. IBM Cloud VPC is not in the public benchmarks we use, so we cannot say which is faster.

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. Start limits, per-start fees and plan minimums are in the price, boat's included.

Scenarioboat, a monthIBM Cloud VPC, 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$18 · Usage account · with limits
One agent that never stops
4 vCPU / 8 GB, 50 GB disk, running 24/7 all month
$26 · $20/mo$57 · Usage account · 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$673 · Usage account · 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$197 · Usage account · with limits
1,000 short jobs at once
2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month
$8,800 · $2000/mo · with limits$344 · Usage account · with limits

boat: Plans cost $20, $100, $500 or $2,000 a month. The whole plan price comes back as VM time for that month, billed per second while a sandbox runs; unused time does not carry over. A stopped sandbox has no compute charge. Saved snapshots above the plan's count and the AI credentials proxy still charge while stopped. Starts above the plan's included count cost $0.03 to $0.07 each.

IBM Cloud VPC: IBM bills each server per second while it runs, with no plan fee. Boot volumes, floating IPs and egress are billed on top, and storage keeps billing while a server is stopped. The IBM figures in the table use burstable Flex servers with a 25% CPU baseline, the lowest-priced fit, at US list rates. Burstable Flex baselines are in beta at IBM; full-rate profiles cost more.

Limits in those scenarios

Every IBM figure in the table uses a beta burstable Flex server that guarantees 25% of each vCPU and can burst to 50% only when the host has spare time. A busy agent runs well below the full machine, so none of the IBM rows fit as asked. IBM also caps each region at 200 vCPU by default, which 50 agents of 4 vCPU use up exactly, and stopped servers still count toward the account's device quota. On boat, the 1,000-job burst does not fit either: the $2,000 plan includes 90 starts a minute, 600 an hour and 2,400 a day. Starts above that are paid, and starts above 270 a minute fail with 429; boat never queues a start. boat's disk is fixed per size, and a stopped sandbox's disk is kept for at least 14 days unless you save a snapshot.

ScenarioboatIBM Cloud VPC
One coding agent, workdaysNone for this scenario
  • burstable CPU: 25% of each vCPU guaranteed, throttled above that
One agent that never stopsNone for this scenario
  • burstable CPU: 25% of each vCPU guaranteed, throttled above that
50 agents in parallelNone for this scenario
  • burstable CPU: 25% of each vCPU guaranteed, throttled above that
Sandboxes for your usersNone for this scenario
  • burstable CPU: 25% of each vCPU guaranteed, throttled above that
1,000 short jobs at once
  • 2,400 starts/day included, paid above that (needs ~6,667)
  • 600 starts/hour included, paid above that (needs ~1,000)
  • starts above 90/min are paid, and starts above 270/min fail with 429
  • burstable CPU: 25% of each vCPU guaranteed, throttled above that
boat's own limits
  • Plans start at $20 a month. The plan price comes back as VM time for that month. Unused time does not carry over. Source
  • Starts (create, fork, resume): the $20 plan includes 12 a minute, 60 an hour and 200 a day. The $2,000 plan includes 90 a minute, 600 an hour and 2,400 a day. Extra starts cost $0.03 to $0.07. A start fails with 429 above 3 times the included starts per minute, or 6 times per hour or per day. Source
  • A stopped sandbox has no compute charge. boat keeps its disk for at least 14 days. A saved snapshot keeps it longer; extra saved snapshots and the AI credentials proxy are paid. Source
  • Auto-stop defaults to 1 hour. You can turn it off on a paid plan. The trial allows 2 hours at most. Source
  • boat saves the disk every minute and at each stop. Memory, running programs and /tmp are not saved: a resume is like a reboot. A machine failure loses writes after the last good snapshot. Source
  • Machines run in Germany, Finland and France only. Snapshot copies can be stored outside the EU. Source
  • Public IPv6 and outbound IPv4. No dedicated IPv4. The IP changes at each resume. Source
  • Hosts are AMD Ryzen 7 9700X and Ryzen 9 9950X. vCPUs are shared. No GPU. You cannot boot your own Docker image. Source

Which one should you use?

Use IBM Cloud VPC when

you need GPUs, Windows, bare metal or a region outside the EU, or when your company already runs on IBM Cloud and wants raw VMs it controls end to end.

  • GPUs and Windows. Useful if agents train or test models, or need Windows: IBM rents GPU servers and Windows Server images.
  • Regions worldwide and network rules. Useful if data must stay in the Americas, Japan, Australia or India, or if traffic must follow IP rules you set.
  • Low-cost burstable and spot servers. Useful for agents that mostly wait: Flex servers with a small CPU baseline cost far less than full-rate ones.

Use boat when

you want agents to work in full VMs at once: Docker and /dev/kvm on every machine, forks of running sandboxes, a desktop, HTTPS URLs and coding agents you drive by API, without building that layer on top of raw servers.

  • Docker and /dev/kvm on every machine. An agent that tests Android apps or runs microVMs needs /dev/kvm. On IBM that work moves from virtual servers to bare metal servers, a different product with its own setup and billing.
  • Coding agents you drive from outside. On IBM you build and maintain the layer that installs agents, sends prompts, streams events and stores keys. With Secrets Manager the key still lands inside the server, where code the agent writes can read it, unless you run your own proxy.
  • Fork a running machine in one call. Agents that try five fixes in parallel need five copies of the same machine. On IBM each copy means a snapshot, a new server and a boot, and stopped servers still bill their volumes.

Questions people ask

Is boat cheaper than IBM Cloud VPC?

It depends on the job. One agent that never stops costs $26 a month on boat and $57 on IBM. 50 agents in parallel cost $317 against $673. One agent on workdays costs $20 against $18, and sandboxes for your users cost $200 against $197. For 1,000 short jobs at once, IBM costs $344 and boat $8,800 because of paid extra starts. Every IBM figure uses beta Flex servers that guarantee only 25% of each vCPU, so none fits as asked. Below $20 of use a month, IBM costs less because boat's smallest plan is $20.

Does IBM Cloud VPC support nested virtualization?

Not on virtual server instances: IBM's docs say nested virtualization is not supported there. IBM Bare Metal Servers for VPC let you run your own hypervisor. Every boat sandbox has /dev/kvm, so Android emulators, Firecracker and QEMU run inside it.

Can I use IBM Cloud virtual servers as sandboxes for AI agents?

Yes. Each server is a full VM with its own kernel, so it isolates code well. You set up the sandbox layer yourself: copies from snapshots, HTTPS preview URLs, a desktop, a keys proxy and an API to drive coding agents. On boat those come with every sandbox; the keys proxy is a paid option.

Does boat have GPUs or Windows like IBM Cloud?

No. boat runs Linux VMs on AMD Ryzen hosts with no GPUs and no Windows guests. Windows programs run through Wine on the Linux desktop. If you need GPUs or Windows Server, IBM has both.

Where does my data run?

boat sandboxes run on machines in Germany, Finland and France, and you cannot pick a region. Snapshot copies can be stored outside the EU. IBM Cloud VPC runs in 13 regions, including Frankfurt, London and Madrid in Europe.

Is boat SOC 2 compliant?

SOC 2 is in progress; boat has no SOC 2 report today. Live status is at status.ascii.dev and our data retention policy is at docs.boat.dev/data-retention. IBM Cloud lists SOC 2 among its compliance programs.

Start a sandbox

Install the CLI, then run boat new.

7-day free trial

More comparisons