FluidStack Cloud Pricing

Don't compromise on time to market or profits. Our pricing ensures you get the best of both worlds.

GPU Model
VRAM (GB)
Max vCPUs per GPU
Max RAM (GB) per GPU
GPU On-Demand Price (/hr)
GPU 6 Month Price (/hr)
NVIDIA HGX H100
80
48
256
$4.76
$2.89
NVIDIA H100 PCIe
80
48
256
$4.25
$2.89
A100 80GB NVLINK
80
48
256
$2.21
$1.48
A100 80GB PCIe
80
48
256
$2.13
$1.30
A100 40GB NVLINK
40
32
128
$1.29
$1.29
A100 40GB PCIe
40
32
128
$1.40
$1.38
A6000
48
32
128
$0.79
$0.59

How we compare

We provide the computational power you need to accelerate your computing training to unprecedented levels.

FluidStack.io the cheapest GPU cloud when compared with Amazon Web Services, Google Cloud and Microsoft Azure

Trial & Subscription

Can I get a trial to test a specific machine?

We offer free trials depending on the use-case and for long-term commitments only. If you think this applies to you, please get in touch with sales@fluidstack.io and provider further information on your server requirements and workload.

Otherwise you can spin up instances by the minute directly from our console for as low as $0.5/hr. You can check out V100 pricing and much more directly from the console!

Can I just check the latency of the machine and then rent it?

Yes, just rent the server for a few minutes to check this - for as low as $0.50/hr.

To minimise latency, you will have to pick a server location as close as possible to you. Luckily, our servers span the entire globe, so you will definitely find a machine in your vicinity.

If you have more questions about latency, please get in touch with our support team through our live-chat or by email at support@fluidstack.io, as this can vary from server to server.

Which payment methods do you accept?

We only accept debit or credit card payments for now. For long-term or bulk rent purchases we can also work with direct bank to bank transfers such as ACH. Please get in touch with sales@fluidstack.io if you’d like to explore this option.

How long does it take to get access to the machine I purchased?

If you are renting an hourly machine in the US through our instant deployment portal, you will gain access to the server within seconds. Note that it may take up to 4 minutes in some cases.

For all other locations (including some in the US) or monthly and longer rentals, this will usually take less than 1 hour if purchased between Monday and Saturday 7am-10pm EST. Note that it may take up to 36 hours in some cases, but you will be informed straight away by email by our technical team if that is the case.

Are your machines fully dedicated or are they spot instances?

Machines are fully dedicated 24/7 to you. They will only stop when you spin down the server or finish your rental period.

Data & Security

What happens to my data when I stop my subscription?

When the rental of your server ends, we automatically destroy your VM and any image you ran on. All your data would therefore be automatically erased. Please beware of that and make sure to save and backup your data externally before requesting a server cancellation. FluidStack will not be held accountable for any data loss following a server termination request.

Do you process personal data and/or share it with third parties?

We do not process any of your data under any circumstances. We do not share nor sell any of your data to Third Parties either. You can request the erasure of your data at any time by emailing us at support@fluidstack.io.

Technicals

What is the difference between KVM Virtual Machines and bare metal instances?

Each bare metal machine receives a fully dedicated IP, whereas virtual machines are placed in a firewalled internal network with specific ports forwarded in. There is a minimal performance loss of less than 3% when running a VM compared to a bare metal server. In both cases, the GPU is fully dedicated to you and not shared with any other user.

Can I run crypto-mining software on the machines?

Mining is only allowed on consumer cards (GeForce RTX range) excluding the latest RTX 30xx series range. Mining is NOT allowed on enterprise grade cards such as the Tesla V100, A100, RTX A6000 etc.

Note that if you are caught mining on a server you are not allowed to, you will immediately be removed from that server and we will not refund you for your purchase as this is a serious breach of FluidStack’s policy.

What if I accidentally shut down my machine?

If you accidentally shut down one of the servers you rent, please email our technical team on the support@fluidstack.io email address. Make sure to indicate the VM ID of the server which needs to be rebooted. One of our tech team members will get your server back live as soon as possible. This typically takes less than 15 minutes anytime between Monday and Saturday from 7am to 10pm EST.

I can only see half of the number of CPU threads (vCPUs) I purchased?

Worry not. All FluidStack servers come with Hyper-threading for Intel CPUs or multi-threading for AMD CPUs. In both cases, this allows for 2 threads per physical core (e.g. an 8 core CPU will have 16 threads). If the server you purchased is a VM, the number of vCPUs shown will correspond to the number of physical cores, NOT to the number of threads. So even if the vCPU number displayed on your computer is half of what you expected it to be, you in reality still have access to double that amount of threads. If you are still confused, please email support@fluidstack.io and they’ll be able to better help you.

Providers & Suppliers

How and when do I get paid?

We pay our data center partners in arrears at the end of every of month via bank transfer or PayPal. We can also arrange other payment methods if needed.

What is an example setup I can provide?

An example machine configuration our data centers provide is:
- 4x RTX A6000 GPUs
- 2x Xeon Silver 4208 CPU (32 Threads total)
- 128 GB RAM1.0 TB NVMe + 2.0 TB GB SATA disk space
- 150 Mbps bandwidth up and down

What are the requirements to become a data center provider on FluidStack?

You must host multiple NVIDIA GPU servers in a secure environment. We prefer Tier 2+ data centers with specific industry certifications. Interested in becoming a supplier? Simply email indigo@fluidstack.io. Make sure to let us know the type, quantity and location of the GPUs you have available. Learn more on our data centers page.

How does the application process work?

Simple. Just email us your available machines at indigo@fluidstack.io. If accepted, we’ll advertise your available servers directly onto our website and you'll get customers running on these in no time! This process can also be automated by integrating directly with your own API.

For larger partnerships, we may set up an introductory call to maximise our relationship.

How does bidding work for custom orders?

Whenever a large customer requests a custom server build, you will receive an automated email. This email will include all the details of the deal, including the expected delivery ETA and price points. Simply answer this email and provide a quote for this custom request. We aggregate all the quotes and will then let you know if the customer decided to proceed with your offer. Please note that responses within 24h of the custom request will be prioritized.

HPC & AI experts we work with

Speak with an Engineer