LiveGPU prices tracked across 12 neoclouds

Rent GPUs from the neocloud market. One console.

Browse live GPU offers from marketplace capacity, cheapest first. Launch a plain VM or your container and get SSH — billed from one prepaid balance. No contracts, no vendor lock-in.

  • Prepaid credits
  • Pay only while running
  • Terminate anytime

Early access: we're onboarding teams in waves.

Neocloud list prices we track

LambdaNebiusRunPodHyperstackFluidstackPaperspaceScalewayHyperbolicCUDODigitalOceanSeewebVerdaLambdaNebiusRunPodHyperstackFluidstackPaperspaceScalewayHyperbolicCUDODigitalOceanSeewebVerda

GPU prices

What a GPU hour costs right now

The lowest list price per GPU-hour across the neoclouds we track, refreshed hourly. Your exact price is shown before you launch.

GPUFromAvailabilityTrend
H100 SXM$1.29 /GPU-hrWideLaunch
H200$2.60 /GPU-hrLimitedLaunch
B200$6.69 /GPU-hrLimitedLaunch
A100 80GB$0.99 /GPU-hrWideLaunch
A100 40GB$1.99 /GPU-hrScarceLaunch
L40S$0.85 /GPU-hrLimitedLaunch
GH200$2.29 /GPU-hrScarceLaunch
RTX 6000 Ada$0.84 /GPU-hrScarceLaunch
MI300X$2.39 /GPU-hrScarceLaunch

Source: the public SkyPilot catalog of neocloud list prices. What you can launch comes from live supplier inventory, so the GPUs, regions and prices on offer can differ from this table.

How it works

From sign-in to a GPU shell in minutes

  1. 01

    Sign in & add credit

    Create an account and top up a prepaid balance with a card. No sales call, no commitment.

  2. 02

    Pick a GPU

    Browse the offers available right now, sorted by price, and filter by GPU, count and region. Which GPUs are on offer depends on live supply. CPU machines are listed too.

  3. 03

    Run your image, SSH in

    Bring a public Docker image or take the plain VM. We provision the machine and show you an SSH command. Terminate it when you're done and billing stops.

Platform

One place to buy, run and account for GPU compute

Vendor-neutral by design: standard VMs and Docker images, so your workloads stay portable. Tags mark what's in preview and what isn't built yet.

Live GPU offers

Live offers from marketplace capacity, sorted cheapest first. A price guard cancels the launch if the price rises above the one you confirmed.

One prepaid balance

Every machine draws from one balance, with usage shown per instance. You get a low-balance warning and a grace period before anything is stopped.

Your container, your keys

Run a public Docker image or the plain VM, and log in with your own SSH key. There's nothing proprietary to adopt.

Built for teams

Owner, admin, member and viewer roles. Everyone sees the org's fleet; members manage their own machines.

CLI, SDK & API

The matmul CLI, a Python SDK and a REST API cover what the console does, with per-person API keys.

Preview

MCP server

Find and launch GPUs from Claude Code, Cursor or another MCP client. Every launch waits for you to OK a price quote.

Soon

Managed jobs

matmul run --gpu H100 --max-price 3.50 -- python train.py runs your command on the cheapest matching GPU, streams the logs and shuts the machine down when it exits.

Bring your SkyPilot

Already run SkyPilot? Attach its dashboard to MatMul and share it with your team, with no VPN and no inbound ports.

Soon

Inference endpoints

Deploy a model behind a stable endpoint that runs on the cheapest capacity available. Not built yet.

Early access

Get GPUs by the hour, on one balance.

We're onboarding teams in waves. Tell us what you run and we'll get you access. Already approved? Sign in and launch.

Request access

Workload