Paperspace
GPU cloud now part of DigitalOcean, pairing hourly instances with Gradient notebooks for people who want a browser and a GPU rather than a cluster.
Pricing Plans
| Plan | Price/mo | CPU | RAM | Storage | Notes |
|---|---|---|---|---|---|
| Free GPU (Gradient) | $0.00 | 2 vCPU | 8GB | — | M4000 tier, queued, session limits apply |
| RTX 4000 | — | 8 vCPU | 30GB | 50GB NVMe | Around $0.56/hr |
| A100 80GB | — | 12 vCPU | 90GB | 250GB NVMe | Around $3.18/hr |
· Updated
About Paperspace
Paperspace was acquired by DigitalOcean in 2023 and now sits inside that product family, though it kept its own console and identity. Its distinguishing feature is that it never assumed you were a cluster operator. You can open a browser, get a Jupyter notebook attached to a GPU, and start working.
Gradient is the notebook and workflow layer. Core is the raw virtual machine product, which includes CPU-only instances and full desktop VMs, so it is closer to a conventional VPS than most GPU hosts.
There is a free GPU tier on Gradient. It is queued, session-limited, and uses older cards, but it is a real way to try something without a card on file.
Pricing
Gradient subscriptions start around $8/month and buy you better free-tier access and longer sessions rather than the GPU itself, which is still hourly. An RTX 4000 runs around $0.56/hr and an A100 80GB around $3.18/hr.
The A100 rate is noticeably above RunPod or Lambda for the same card. You are paying for the managed notebook experience and the polish, which is a reasonable trade if it saves you setup time, and a bad one if you were going to SSH in anyway.
Storage is billed separately and persists between sessions, which is the point but also the surprise on the first invoice.
Service Details
Machines run Ubuntu, with templates for PyTorch, TensorFlow, and plain ML-in-a-box setups. Core VMs can be persistent, so unlike a pure hourly GPU service you can treat one as a long-lived development box.
The integration with DigitalOcean means billing and, increasingly, networking sit alongside their droplets and managed databases. If your application already lives there, keeping the GPU in the same account is genuinely simpler.
Available in: New York, California, Amsterdam
Learn More
For more information about Paperspace, visit their website at https://www.paperspace.com/