It starts with a relentless drive to cut costs with custom-built servers, giving us access to reliable yet affordable high-performance computing power.
To handle customer needs during surges, we partner with other cloud providers, where we get discounted pricing due to our massive scale.
Our low prices mean that we don't need to spend that much on advertising, which means we can cut prices further. Neat, right?
Yes, seriously. With over 1,000 GPUs in-stock. Get the right GPU model for your workload. Only need FP32 performance? Skip the A100 and get the A40. Need best price-to-performance? Skip the A40 and get two A5000s. We have GPUs optimized for every use case:
Deep learning training
VFX / Rendering
Built from scratch: metadata/restrictions, stock, deploy, start, stop, delete, list, get, plans — it's everything you need! And we're still building more!View documentation
We use a super duper fast network-based storage system. It's triple-replicated, so you'll never lose your data (still take backups, though!). It's Ceph-based, so the technology is tried and reliable. And, it's fast. Over 1 GB/s for our NVMe-based network SSD block storage! Learn more.
TensorML is our own Ubuntu 18.04-based OS template with Jupyter Notebook, Jupyter Lab, TensorFlow, Keras, PyTorch, MxNet, Numpy, Pandas, NVIDIA drivers, CUDA, cuDNN, CUDA Toolkit, Docker, and NVIDIA Docker2. Phew! That's a lot! And it means that you can get started without wasting any time installing your own software.