The best prices around town.

And yes, fair-share bandwidth is included.

On-Demand GPU Servers

Among the most affordable prices in the industry.
Fully customize your compute resources.

Simply select the GPU, RAM, CPU, disk, add the prices, and that's it! Simple.

Save by renting a monthly, quarterly, or yearly subscription. Contact us for more details.



GPUs Price Per Hour, Billed Per-Minute
NVIDIA A100 80GB PCIE $2.43
NVIDIA A100 40GB PCIE $2.27
NVIDIA A100 SXM $2.27
NVIDIA A40 $1.41
NVIDIA A6000 $1.41
NVIDIA V100 SXM $0.88
NVIDIA A5000 $0.85
NVIDIA A4000 $0.68
NVIDIA Quadro RTX 5000 $0.63
NVIDIA V100 PCIE $0.52
NVIDIA Quadro RTX 4000 $0.27
Resources Minimum Price Per Hour, Billed Per-Minute
Reasonable Bandwidth Allowance Included Included
Each vCPU (1 thread, 1/2 physical core) 1 vCPUs ($0.010/hour) $0.01
Each GB of DDR4 RAM 2 GB ($0.010/hour) $0.005
Each GB of block NVMe SSD storage (io1 class) 20 GB ($0.0020/hour) $0.0001
Each GB of block HDD storage (st1 class) 20 GB ($0.0012/hour) $0.00006
Deploy Your GPU Server Now

Compare with other hourly GPU cloud platforms —
TensorDock is consistently the most affordable:

Platform vCPUs RAM GPU Count GPU Type Competitor Price Our Price Savings
AWS 8 61 1x V100 $3.06/hr $0.86/hr 72%
Azure 6 112 1x V100 $3.06/hr $1.09/hr 64%
Oracle 6 90 1x V100 $2.95/hr $1.04/hr 64%
Paperspace 16 30 1x V100 $2.30/hr $0.78/hr 66%
LeaderGPU 8 64 1x A6000 €0.06/min ~€0.03/min 50%

Note: Please conduct your own comparisons for your own use case. For V100 comparisons, competing providers may be using V100 SXM cards, whereas our prices are with V100 PCIE cards, potentially leading to slight performance differences, especially in multi-GPU configurations, which are not compared here. In addition, the data may be inaccurate, as competing services may have adjusted their pricing since this table was made. Features like network speed, bandwidth pricing, and storage costs were not factored into this comparison table.

Deploy Your GPU Server Now

On-Demand CPU-Only Servers

Powered by the latest Intel Xeon and AMD EPYC CPUs.

Our on-demand CPU-only servers come in multiples of 1 vCPU and 4 GB of RAM. Unlike other clouds, our CPUs are dedicated, meaning you can use it 100% 24x7. Storage pricing for on-demand cpu-only servers is the same as listed above.

CPU Class RAM Per vCPU Cost Per vCPU Per Hour, Billed Per-Minute
Intel Xeon v4 4 GB $0.025
AMD EPYC Rome 4 GB $0.032
Intel Xeon Scalable 4 GB $0.038
AMD EPYC Milan 4 GB $0.043
Deploy Your CPU-Only Server Now

Long-Term Subscription Servers

More options, more locations, lower prices
Available on monthly or longer terms

Our on-demand deployment platform requires a high amount of available GPUs to handle demand surges. By renting long-term, you'll be able to get access to more GPUs at more locations whose stock is too low to handle the demand surges of our on-demand platform. When you contact us, we'll be able to customize a configuration that suits your needs, e.g. we can place you on a Ryzen CPU instead of an EPYC CPU if you need higher clock speeds. Lead times start at 24 hours.

Learn more about our locations. Please order a subscription today if you're interested. The following pricing assumes a 12-month contract or 3-month prepayment.

Sample subscriptions:

GPU vCPUs RAM Storage Location Cost per month
NVIDIA 1650 8 14 GB 100 GB NVMe Boston 01/02 $75
NVIDIA 1070 12 14 GB 100 GB SSD Boston 01/02 $125
NVIDIA 2080S 12 14 GB 350 GB SSD Boston 01/02 $150
NVIDIA 1070 6 14 GB 100 GB SSD Singapore 01/02 $175
NVIDIA A2000 24 28 GB 350 GB SSD Boston 01 $175
NVIDIA 2080 Ti 12 28 GB 350 GB NVMe Boston 01/02 $200
NVIDIA 3060 Ti 12 28 GB 350 GB SSD Boston 01 $225
NVIDIA 3070 12 28 GB 350 GB NVMe Boston 01 $250
NVIDIA 3070 Ti 12 28 GB 350 GB NVMe Boston 01 $275
NVIDIA 3080 12 28 GB 350 GB NVMe Boston 02 $300
NVIDIA A10 16 28 GB 500 GB NVMe Boston 02 $325
Order a subscription
Enterprise

High spenders get enterprise accounts.

When you spend more than $10,000/month on TensorDock:

Account Manager

A person to ensure that your needs have been met.

24/7 Priority Support

Your requests will be placed in a priority queue.

Slack and MS Teams

Chat with our team in realtime to resolve issues.

FAQ

Frequently asked billing questions.

Contact us to ask something else!

How does billing work?

We operate on a prepaid model: you deposit money and then provision compute. You must manually add more funds, we do not automatically charge your card.

Can we do net 30/60/90 terms?

No — to keep our prices low, we cannot afford to risk companies defaulting or more often, scammers defrauding us. We cannot afford to provide services where there is a risk that we could not be paid. If you do not want to have a large balance with us, you can make small but frequent deposits.

Do you check 3D Secure?

We accept credit cards via Stripe. We will attempt to do a 3D Secure check on all payments. Credit cards that do not support 3DS are automatically assigned a higher fraud score and are at a higher risk of being placed in a "fraud check" mode.

Do you offer refunds?

Each refund costs us, so we can offer refunds of remaining account balances on a case by case basis. Please start off small with a $5 deposit to test out our services.

What payment methods do you accept?

At the moment, we only accept credit card payments via Stripe. This is because with 3D Secure, we can shift chargeback liability to the issuing bank. We can also manually accept cryptocurrency payments on deposits larger than $100 on request.

What is "fraud check" mode?

We utilize multiple fraud checks. If you are automatically flagged and lose your ability to provision servers, do not panic. Contact us (we will not contact you unless we are deleting already-provisioned compute), and we can work things out.

"These prices are some of the best that I've seen! And the scalability... perfect"

Creavite is a freemium animated graphics creator that uses TensorDock's Windows GPU servers with Adobe software for their rendering needs. Note: Creavite received discounts for placing a "partners" link on their website linking to us.

@hannes

"The server is really fast, you never have to worry about noisy neighbors. Wow!"

Ultramarine Linux is an open-source Fedora-based Linux distribution that uses TensorDock's Linux servers to process hundreds of gigabytes of data. Note: because it is open-source, we sponsor Ultramarine Linux compute servers.

... ... ... ... ... ... ... ... ... ... ... ... ...