Pricing
Bring your own compute, or buy ours.
Connect the servers and cloud accounts you already have for a small Grid fee, or buy compute from us with the Grid fee included. Share it with your team for $10 a person a month.
- Your first 5 servers are free
- Your first 5 users are free
Your Grid
Your computeKinesis compute
One place to run it allDeploy once. Kinesis keeps it running.
Bring your own
- Servers you ownJoin with one install command
- Your cloud accountsAdd an API key; we launch the machines
You pay a small Grid feefrom $0.08 per GPU-hour
Or buy from us
- Data center GPUsA100, H100, and B200
- Kinesis BatchesGeForce, RTX workstation and data center GPUs, at capped prices
- CPU capacityGeneral purpose
Grid fee includedfrom $0.03 per GPU-hour
1Bring your own compute
A small Grid fee for the compute you bring.
Bring servers you own, or let Kinesis launch machines in your cloud accounts with your API keys. Your provider keeps billing you for the machines, as it does today. Kinesis charges only the Grid fee.
Servers you own
Run one install command on any Linux server, on-prem or in colocation, and it joins your Grid in minutes.
Your cloud accounts
Add an API key for any of 20 GPU clouds. Kinesis launches and manages machines in your account as your apps need them.
Ultra high performance data center GPUs$0.27per GPU-hour
B200, B300
Standard data center GPUs$0.16per GPU-hour
H100, H200, A100, AMD MI300X and MI325X
Selected data center, workstation and consumer GPUs$0.08per GPU-hour
RTX 6000 Ada, RTX PRO 6000 Blackwell, RTX 5090, RTX 5080, RTX 4090, RTX 4080, RTX 3090, L40, L4
CPU-only servers and unlisted GPUs$0.03per server-hour
CPU-only servers, and GPU models not listed above
Your first 5 servers are free.The Grid fee starts with your 6th server.
2Or buy compute from us
Rent our GPUs by the hour, Grid fee included.
Add Kinesis GPUs and CPUs to the same Grid when you need them, and release them when you don’t.
More detail
Dedicated machines are yours while you run them. Kinesis Batches runs on spot GPUs, from GeForce cards to data center GPUs, and every Batches price is a cap.
Dedicated
- A1001, 2, 4, or 8 GPUs$1.62 per GPU-hour
- A100 NVLink8-GPU nodes$1.68 per GPU-hour
- A100 SXM8-GPU nodes$1.92 per GPU-hour
- H1001, 2, 4, or 8 GPUsfrom $2.28 per GPU-hour
- H100 NVLink8-GPU nodes$2.34 per GPU-hour
- H100 SXM8-GPU nodes$2.88 per GPU-hour
- B200 SXM8-GPU nodes$4.79 per GPU-hour
- RTX A40001, 2, 4, 8, or 10 GPUs$0.18 per GPU-hour
- General purpose4 GB RAM per vCPU$0.036 per vCPU-hour
Kinesis Batches
Spot GPUs for work that runs in batches. Every price is a cap: the most an hour can cost. Live availability
- B300288 GB288 GB$6.94 per GPU-hour
- B200180 GB180 GB$6.74 per GPU-hour
- H200141 GB141 GB$3.06 per GPU-hour
- H10080 GB80 GB$2.70 per GPU-hour
- A10080 GB80 GB$0.89 per GPU-hour
- L40S48 GB48 GB$0.80 per GPU-hour
- RTX PRO 600096 GB96 GB$1.55 per GPU-hour
- RTX PRO 500048 GB48 GB$0.76 per GPU-hour
- RTX 6000 Ada48 GB48 GB$0.70 per GPU-hour
- RTX 5000 Ada32 GB32 GB$0.49 per GPU-hour
- RTX 4500 Ada24 GB24 GB$0.47 per GPU-hour
- RTX PRO 450032 GB32 GB$0.42 per GPU-hour
- RTX A600048 GB48 GB$0.39 per GPU-hour
- RTX PRO 400024 GB24 GB$0.38 per GPU-hour
- RTX 4000 Ada20 GB20 GB$0.20 per GPU-hour
- RTX A500024 GB24 GB$0.20 per GPU-hour
- RTX A450020 GB20 GB$0.19 per GPU-hour
- RTX A550024 GB24 GB$0.19 per GPU-hour
- Quadro RTX 800048 GB48 GB$0.18 per GPU-hour
- RTX 4000 SFF Ada20 GB20 GB$0.18 per GPU-hour
- RTX A200012 GB12 GB$0.15 per GPU-hour
- Quadro RTX 600024 GB24 GB$0.14 per GPU-hour
- RTX A400016 GB16 GB$0.14 per GPU-hour
- TITAN RTX24 GB24 GB$0.14 per GPU-hour
- RTX PRO 200016 GB16 GB$0.10 per GPU-hour
- Quadro RTX 500016 GB16 GB$0.09 per GPU-hour
- RTX 2000 Ada16 GB16 GB$0.08 per GPU-hour
- Quadro RTX 40008 GB8 GB$0.05 per GPU-hour
- RTX 509032 GB32 GB$0.61 per GPU-hour
- RTX 508016 GB16 GB$0.35 per GPU-hour
- RTX 5070 Ti16 GB16 GB$0.27 per GPU-hour
- RTX 507012 GB12 GB$0.17 per GPU-hour
- RTX 5060 Ti16 GB16 GB$0.12 per GPU-hour
- RTX 50608 GB8 GB$0.11 per GPU-hour
- RTX 50508 GB8 GB$0.07 per GPU-hour
- RTX 409024 GB24 GB$0.39 per GPU-hour
- RTX 4080 SUPER16 GB16 GB$0.30 per GPU-hour
- RTX 4070 Ti SUPER16 GB16 GB$0.27 per GPU-hour
- RTX 408016 GB16 GB$0.27 per GPU-hour
- RTX 4070 Ti12 GB12 GB$0.20 per GPU-hour
- RTX 4070 SUPER12 GB12 GB$0.18 per GPU-hour
- RTX 407012 GB12 GB$0.16 per GPU-hour
- RTX 4060 Ti16 GB16 GB$0.13 per GPU-hour
- RTX 40608 GB8 GB$0.08 per GPU-hour
- RTX 3090 Ti24 GB24 GB$0.31 per GPU-hour
- RTX 309024 GB24 GB$0.20 per GPU-hour
- RTX 3070 Ti8 GB8 GB$0.19 per GPU-hour
- RTX 308010 GB10 GB$0.17 per GPU-hour
- RTX 3080 Ti12 GB12 GB$0.17 per GPU-hour
- RTX 30708 GB8 GB$0.13 per GPU-hour
- RTX 3060 Ti8 GB8 GB$0.10 per GPU-hour
- RTX 306012 GB12 GB$0.07 per GPU-hour
- RTX 30508 GB8 GB$0.03 per GPU-hour
3Share with your team
$10 per person a month. Share as many Grids as you like.
You pay for the unique people you share with, not for Grids. Share one Grid with 10 people, or 10 Grids with the same 10 people: it’s the same price. Your first 5 users are free.
1 Grid, 10 people
10 Grids, the same 10 people
Free$10 / month
$50 / month
10 people − 5 free = 5 × $10
In every Grid
Everything else is included.
- Deploy from an image, a Dockerfile or a GitHub repo
- HTTPS endpoints, custom domains, automatic certificates
- Autoscaling, health checks and alerts
- Logs, metrics and a browser terminal
- Fair share, quotas and MIG slicing
- Cost by project and app, invoices, CSV exports
- Roles, MFA and an audit trail
- A REST API and an MCP server
Questions
What people usually ask about pricing.
For compute you bring, your cloud provider keeps billing you for the machines in your account, and your own servers cost what they already cost. Kinesis bills only the Grid fee. Compute you buy from Kinesis is one all-in hourly rate, Grid fee included.
Add an API key for one of the 20 GPU clouds we support, and Kinesis launches and manages machines in that account for you, as your apps need them. They run in your account, under your contracts and credits.
Find the model in the Grid fee table above. GPU models that are not listed are charged as unlisted, at $0.03 per server-hour.
Each unique person you share any of your Grids with, counted once however many Grids they can reach. Your first 5 are free; after that it is $10 per person a month, billed to the owner.
Your first 5 servers and your first 5 users. You can run a real Grid with your team before you pay anything.
Only if you ask for it. Your Grid never places work on community hardware unless you choose Kinesis Batches.
Kinesis Private is quoted per deployment. Talk to us about your environment and requirements.
Your first 5 servers and 5 users are free.
Connect your first cloud account or server in a few minutes and set up your first Grid. You pay only when you grow past the free tier.