SKIP_TO_MAIN_CONTENT
06.0 // Comparison Apples to apples Same YAML

machine.dev vs
Depot.

Depot pitches "30% faster CPUs at half the cost." machine.dev is 12.5× cheaper on spot and 26% cheaper on on-demand at every Linux CPU size — plus a public GPU catalog Depot doesn't ship outside Business plan.

01 Intro
[REF_001]

Bigger surface, higher rate.

Depot is a broader product than machine.dev. They ship CI runners, Docker builders, container registry, Depot CI, and Agent Sandboxes for Claude Code. Their customer roster includes PostHog, Jane, and People Data Labs. They're the only competitor in the market that bills per-second with no 1-minute minimum.

Their Linux x64 GitHub Actions runners are $0.002/vCPU/min linear — the same flat rate as Blacksmith and WarpBuild managed. machine.dev's effective rate is ~$0.00149/vCPU/min on-demand, which works out to a clean 26% cheaper at every size.

Where the gap really opens is GPU. Depot's GPU runners are Business plan only, customer-managed via an AWS sub-account, with no public price catalog. machine.dev publishes T4G, T4, L4, A10G, L40S, RTX 6000, plus AWS Trainium and Inferentia2 on the public per-minute catalog — all on the same credit balance, all included in the $10 free trial.

02 Cpu pricing
[REF_002] // LINUX_X64

Linear, Both Ways.

Per-minute USD. Depot prices verified from depot.dev/pricing April 9, 2026. Both providers scale linearly per vCPU.

[NOTE_SWEEP] machine.dev wins on both spot and on-demand at every CPU size against Depot. Both sides scale linearly: Depot at $0.002/vCPU/min, machine.dev at ~$0.00149/vCPU/min on-demand. The 26% on-demand gap and the 12.5× spot gap are uniform.

[NOTE_DEV_PLAN] Depot's $20/mo Developer plan includes 2,000 GHA minutes, which lowers the effective rate at very low volume. Above ~5,000 minutes/month the per-minute rate dominates and the 26% gap reasserts itself.

02b Gpu pricing
[REF_003] // NVIDIA

Public catalog vs sales call.

Depot's GPU runners are Business-plan only, customer-managed via AWS sub-account, with no public SKU pricing. machine.dev publishes T4G ARM, T4, L4, A10G, L40S, RTX 6000, plus Trainium and Inferentia2 on the public per-minute catalog.

02c Features
[REF_004]

Beyond price.

SPEC DEPOT MACHINE_DEV
Pricing model Linear $0.002/vCPU/min, per-second billing Per-minute USD, ~$0.00149/vCPU/min OD
Spot tier No Yes (12.5× cheaper than Depot)
GPU runners Business plan only (BYO-AWS sub-account) Yes (T4G, T4, L4, A10G, L40S, RTX 6000)
GPU SKU catalog No public catalog Public, priced per-minute
AWS Trainium / Inferentia2 No Yes
ARM GPU (T4G) No Yes (only provider in market)
CUDA pre-installed No (BYO image) Yes (12.1 + cuDNN 9.2.1)
Per-second billing Yes (only competitor with this) Per-minute
Windows runners Yes (Windows 2022/2025) No (Linux only)
macOS runners Yes (Apple Silicon) No (Linux only)
Docker layer cache Yes (first-class product) actions/cache
Free tier 7-day trial, no card $10 free compute, no card
Setup One runs-on change One runs-on change
03 What you keep
[REF_003] // SCOPE_OF_CHANGE

The only line you touch.

Your workflow files, steps, secrets, caches, and Actions syntax stay exactly as they are. The only change is the runs-on label.

BEFORE.YAML DEFAULT
jobs:
train:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- run: python train.py
AFTER.YAML EXECUTE
jobs:
train:
runs-on: [machine, gpu=T4]
steps:
- uses: actions/checkout@v4
- run: python train.py
04 Where depot wins
[REF_004]

Three strengths.

[REF_010] BILLING // SECOND

Per-second billing

Depot is the only competitor in the market that bills per-second with no 1-minute minimum. For pipelines with lots of small fast jobs, this can be meaningful — though the savings are usually under 5% in practice.

[REF_011] PLATFORM // BREADTH

Broader product surface

Depot is more than CI runners. They ship Docker builders, container registry, Depot CI, and Agent Sandboxes for Claude Code. If you want one vendor for all of those, Depot is consolidated. machine.dev is focused on runners.

[REF_012] PLATFORMS

Windows + Mac coverage

Depot offers Windows 2022/2025 and macOS Apple Silicon runners. machine.dev is Linux-only (x64 + ARM64). If you need cross-platform CI, Depot ships more platforms.

05 Where machine dev wins
[REF_005]

Five strengths.

[REF_020] PRICE // SPOT

12× cheaper on spot at every size

Depot has no spot tier — they run linear $0.002/vCPU/min on a single managed cloud. machine.dev's spot CPUs are 11.4-12.5× cheaper at every size from 2 to 64 vCPU. If your jobs tolerate retries, the math is unambiguous.

[REF_021] PRICE // ON_DEMAND

26% cheaper on every on-demand size

Both providers scale linearly per vCPU. Depot is $0.002/vCPU/min on Linux x64; machine.dev's effective rate is ~$0.00149/vCPU/min. The 26% gap holds at every size from 2 to 64 vCPU on x64 and ARM.

[REF_022] GPU // CATALOG

GPU runners with public pricing

Depot's GPU is Business-plan only, customer-managed via AWS sub-account, no public price catalog. machine.dev publishes T4G, T4, L4, A10G, L40S, RTX 6000, plus AWS Trainium and Inferentia2 — all on the public per-minute catalog, all on the same credit balance, all included in the $10 free trial.

[REF_023] GPU // T4G_ARM

The only ARM GPU runner anywhere

T4G ARM 8 vCPU spot is $0.00277/min — the cheapest GPU runner on the public internet. Depot's GPU stops at customer-managed AWS instance types and doesn't offer ARM GPU at all.

[REF_024] CUDA // PRE_INSTALLED

CUDA pre-installed on every GPU runner

Depot Business plan GPU runners are bring-your-own-AMI. machine.dev ships drivers 555.58, CUDA 12.1, and cuDNN 9.2.1 baked into every GPU image. torch.cuda.is_available() returns True on the first run.

06 Live pricing
[REF_PRICING] // LIVE_DATA

Live per-minute rates

See what you'd actually pay. Compare with GitHub's $0.006/min for 2-core, $0.012/min for 4-core, and $0.052/min for GPU.

07 Initialize
READY
> SYSTEM_READY // AWAITING_INPUT

Get started free.

$10 free compute on signup. Enough to run real workloads and compare cost before committing.