LowEndBox - Cheap VPS, Hosting and Dedicated Server Deals

Cheap GPU List - Nvidia GPUs for AI, Training LLM Models, and More!

Unless you’re a corporate ghoul with an unlimited AWS budget, getting your hands on cheap GPU horsepower is nigh impossible.  We have a solu…okay, scratch that, the word “cheap” doesn’t pair well with “GPU rental”.

But we’ll try. We’ve scoured the digital wasteland to compile a list of cheap GPU hosting providers for those want to stretch their coin to the absolute limit.  Whether you’re training an AI to predict your doom, mining pretend Internet coins, or just trying to render a video before you hit retirement age, this directory will point you toward smaller, scrappier providers who will rent you GPUs for non-Zuckerburg budgets.

Again, there are no $7/month offers here.  Even the cheapest are not cheap, if you know what I mean.  But we did our best. Pricing is current as of this article.

Please note that this is a very hot market, so you may see systems out of stock.  If something you want isn’t available, either check back or contact the provider to ask for an ETA.

Are you a provider who wants to be listed on this page?  Send us a link and we’ll include you!

CrunchbitsCrunchbits

This community provider offers GPUs on a month-to-month rental basis, paired with a big server (for example, 20GB of RAM, 480GB SSD, 20TB of bandwidth, etc.).  Service is in Spokane, Washington or Allentown, Pennsylvania.

 

Example Pricing:

  • RTX 3070 – $65/mo
  • RTX 3090 – $135/mo
  • RTX A6000 – $325/mo

InterServerInterServer

Another community provider (which is why we list them first!).  InterServer offers a range of hosting services, including typical shared/VPS/dedicated and also dedicated backup VPS.  For GPU servers, they pair the almighty card with a fast Ryzen CPU, a pair of 4TB NVMe drives, 192GB of RAM, and 1Gbps unmetered.

 

Example Pricing:

  • RTX 4090 – $399/mo

TensorDockTensorDock

TensorDock is a little different.  Instead of building out their own fleet of servers and selling them, they instead broker for others.  Different providers sign up to offer GPU-based systems, and TensorDock coordinates the whole relationship, with reputation ratings.  While you lose the “one throat to choke if something goes wrong” advantage, you do get access to a true global network of providers, plus hourly pricing.

Example Pricing:

  • H100 – $2.21/hour
  • RTX 4090 – $0.34/hour
  • RTX 3090 – $0.30/hour
  • RTX 4080 – $0.22/hour
  • RTX 6000 – $.72/hour

VultrVultr

Vultr is well-known and we’re including them here for users who need a little more API but don’t have the massive budget needed for AWS or Azure.  Vultr offers hourly pricing and while their pricing is not rock bottom, it’s cheaper than the big boys.  In theory a big company like this is more stable, though some of our community feel that having a smaller player who’s hungry for your business is a better choice.  Up to you.

Example Pricing:

  • H100 $2.30/hour

PureGPU

PureGPU

PureGPU is a monthly-contract provider, offering GPUs riding on AMD Epyc systems with SSDs.  They offer servers with up to 4 GPUs, so if your budget is large (and I do mean large – we’re talking > $3,000/mo) you could have 192GB of VRAM at your fingertips.  We’ll leave that pricing as an exercise for the reader.  Here’s what pricing on a 4090 looks like for us mere mortals.

Example Pricing:

  • RTX 4090 – $320/mo
  • RTX 4000 – $125/mo

vast.ai

vast.ai

“Vast.ai is the market leader in low-cost cloud GPU rental,” according to themselves.  Compared to others, their pricing is pretty good, and is offered on an hourly basis.  This is particularly true if you can utilize interruptible spot instances.

 

Example Pricing:

  • H100 $2.07/hr (or $1.73/hr on interruptible)
  • A100 $1.538/hr

ScalewayScaleway

Scaleway offers many hosting options, including block storage, which is what is utilized for their GPU offers.  I find their offers strange…for example, their 3090 pricing simply says “16 cores” for CPU but doesn’t tell you the model…I mean, that could be a 15-year-old CPU for all we know.  Ask questions.

 

Example Pricing:

  • 3090 – €0.98/hour
  • H100 €2.73 /hour