Packet AI is an active GPU cloud in this directory, billed per hour, with credit card as the recorded payment method. We do not have a founding year or a headquarters on file. What we do have is a wide scraped catalogue: L4, A30, A40, A6000, A100, H100, H200, B200, and L40S among the published rows.
Why Packet AI stands out
The earlier profile on this page treated Packet as an empty beta. That is out of date. The data card now records an API, Docker, multi-GPU nodes, persistent storage, and reserved instances. Jupyter, Kubernetes, and spot instances are not set, and there is no SOC 2 flag.
Use the price table on this page for numbers. The rows are hourly. A wide catalogue with reserved capacity and no spot market is a different product from a peer-to-peer marketplace.
Pros
- Published NVIDIA range from L4 and A40 through H100, H200, and B200
- Per-hour billing and credit-card payment recorded
- API, Docker, persistent storage, multi-GPU, and reserved instances all set
- Active in the directory, not a beta stub
Cons
- No Jupyter and no Kubernetes in our data
- No spot instances, so there is no preemptible discount path recorded
- Not marked SOC 2 compliant
- No public founding year or headquarters in our data, which makes the company harder to judge
Getting started
- Open Packet AI and compare a card you care about with the hourly row on this page.
- Create an account with a credit card.
- Confirm Docker and API access if you deploy from your own tooling. Do not expect a Jupyter or Kubernetes path from what we have recorded.
- Look at reserved pricing if the workload is steady. Spot is not listed.
Best for: Teams that want an hourly cloud with a broad NVIDIA catalogue, Docker, and an API, and that do not need Jupyter, Kubernetes, or spot capacity.