TensorDock is active here, founded in 2021, headquarters Boston. The tagline is an affordable GPU cloud with bare-metal VMs. Billing is per hour. Payment methods on file are credit card and crypto. The published catalogue includes A100 SXM and PCIe, H100 PCIe, L40, L4, V100, RTX 4090, RTX 3090, RTX 5090, RTX 6000 Ada, and RTX A6000, A5000, and A4000.
Why TensorDock stands out
Our data records persistent storage, multi-GPU, an API, Jupyter, Docker, spot, reserved instances, and SOC 2. Kubernetes is the feature that is not set.
Pros
- Wide published mix of A100, H100 PCIe, and GeForce cards
- Per-hour billing, credit card and crypto
- Jupyter, Docker, an API, spot, reserved instances, and persistent storage recorded
- Marked SOC 2 compliant, Boston, founded 2021
Cons
- No Kubernetes in our data
- Pricing-competitiveness score is only a little above the middle. “Affordable” is the tagline, not a measured ranking at the top of this directory
Getting started
- Open TensorDock and choose spot or reserved if you need one of those. Both are recorded.
- Pay by credit card or crypto.
- Use Jupyter or Docker. Kubernetes is not recorded.
Best for: Hourly bare-metal style A100, H100 PCIe, or 4090 capacity with Jupyter and spot, and no need for Kubernetes.