Now In Public Beta

Slash costs by 10x by using
world's cheapest GPU servers

We are re-envisioning the very concept of cloud. We are building a cloud on top of dozens of independent hosts' servers. They compete against each other for pricing, uptime, and internet connectivity. The result:

A huge variety of available locations, the lowest costs, and the best incident response times in the industry.

Deploy a Marketplace GPU

Slash Costs

Vetted hosts will compete against each other on pricing and feature set

Deploy Anywhere

With a global set of hosts, you'll be able to find compute near your end users

Reliable

We penalize hosts for downtime, so they are incentivized to keep servers running

...

How it works

1

Hosts join

We interview prospective hosts to ensure they are reliable and trustworthy. Then, they install our software and set prices.

2

You browse our marketplace

You'll be able to browse all available servers from and compare historical uptime, hardware specifications, and location to select the one that best suits your needs.

3

We orchestrate your VMs

Once you've selected your server, our software will instantly orchestrate your VMs. In our testing scenarios, we've seen deployments as low as 30 seconds. Run your workloads confidently, knowing that you've selected a reliable host.

Pricing from...

Pricing is done a la carte. Simply add up the prices for each resource.

Prices may vary from host to host, and availability is tight

Item Hourly Cost
vCPU $0.003/hr
GB RAM $0.002/hr
GB Storage $0.00005/hr
Graphics Card Hourly Cost
A6000 / A40 $0.40
Quadro RTX 8000 $0.30
GeForce 3090 $0.25
A5000 / A10 $0.21
GeForce 3080 Ti $0.19
GeForce 3080 / Quadro RTX 4000 $0.15
GeForce 3070 Ti $0.14
GeForce 3070 $0.13
GeForce 3060 Ti $0.12
GeForce 3060 / 2080 Ti $0.10
GeForce 1650 $0.05

Frequently Asked Questions:

?

Is marketplace the same as Core Cloud in terms of billing?

The marketplace and Core Cloud platforms run on different databases and different accounts. The balances for the two platforms are not shared and there is no transfer between the two. In order to keep all your servers running, both accounts must maintain a positive balance. To read more on the specific differences between the two, visit our Core Cloud page.

?

Can host look at my data?

No. Hosts need to sign data protection agreements, and given the nature of deployments being virtual machines and not containers, it will be very hard for them to see the actual contents of your virtual machine. We will also eventually add encrypted VM disks, which will make it almost impossible for hosts to look at your data. That being said, if you are concerned about data, grab one of our enterprise-ready Core Cloud servers.

?

How can hosts be so cheap?

Operating GPU servers for TensorDock can be profitable. A 5 kW full rack with 1 gbps (we can make an intro) will cost under $1k/month and as low as $700/month depending on the location. That is to say, operating costs are as low as $0.20/kWh at a data center. A 3090 consumes about 400 watts with the system factored in, or about $60 per month. A 3090 costs about $1k right now. That means, if it is rented for just $160/month after marketplace fees, the host will get their initial investment paid back in just a year!

The math is a bit more complex, but in short, hosting servers for TensorDock can be very profitable, even though ~$200/month 3090s would be a fraction of what you'd pay at other clouds.

?

What if a host goes offline with my data?

We withhold payouts and ban hosts that do not boot back up their machines after a downtime. This is to ensure that hosts will not pull their servers from our marketplace to serve higher-paying workloads that they may find. Given our penalties and the fact that hosts will be banned forever, this is extremely unlikely. Still, this situation would be possible — that's why regularly making backups is so important.

?

Will I be charged the same rate when my virtual machine is off?

Yes, unlike our Core Cloud systems which have reduced storage fees when your virtual machine is turned off, our marketplace platform will charge you the same hourly rate as when the machine is running, regardless of its current state. If you need more flexible deployments, Core Cloud may be able to provide services that better fit your needs.

?

Downtime will happen sooner or later. What happens then?

Hosts with good uptime and no data loss will eventually be given "Top Host" badges so that they'll be featured more prominently. "Top Hosts" will typically have redundant power, a multi-homed network, etc. As such, it will be extremely unlikely that these hosts go down. If you are concerned about uptime, select a host that has proven uptime.

When a server does go down, the host will receive a notification, and we'll be in contact. You can also follow up with us directly via our support channels. You will not be charged for the time the server is down, and you'll receive a bit of account credit as compensation as well. Once the server is back up, you'll be back up too :)

...

You're with a good company.

0+ People

As of July 2022

0-figure

June valuation

$0m+ in GPUs

Deployable

We built the software to sell our own servers. Now, we're building the software to sell others' servers.

As companies scale, marginal costs rise. We're here to unite the independent hosts of the world — the ones who have the lowest costs by sourcing hardware locally and being lean — and place them together on a single, intuitive marketplace to provide the compute that rivals the big clouds but at a fraction of the cost.

We're always receptive to feedback, just contact us!

Get started

Our marketplace will .

Our marketplace will fundamentally change the way computing is done. We'll be the first-ever site to offer automated, virtualized computing from a network of independent hosts. This will bring down costs, allow you to deploy servers in more locations, and power these small businesses to fight back against the monopolistic landscape of the large clouds.

Ask us a question Register
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...