NVIDIA Hopper

Rent H200 servers, reserved from data centers

H200 is a Hopper GPU with 141 GB of faster HBM3e, against 80 GB on the H100. The extra memory and bandwidth mostly help inference: bigger models and longer contexts fit on fewer GPUs.
Memory per GPU
141 GB HBM3e
Architecture
Hopper
Listings
On request
From
Ask

H200 capacity listed now

No H200 capacity is listed right now. Send a request and we will source it from our owners.

Tell us what you need

What H200 is good for

  • Serving 70B to 400B parameter models with long context
  • Fine-tuning that runs out of memory on H100

What to ask before you reserve H200 capacity

  • SXM (HGX, NVLink between the eight GPUs) or NVL (PCIe cards); they are different machines.
  • How the servers are interconnected (InfiniBand or RoCE, and at what speed per GPU) if you will train across nodes.
  • Whether the price covers storage, egress and support, or only the GPU-hours.
  • The uptime they commit to, and what you get back when they miss it.

Questions

How much does it cost to rent a H200?
Owners set their own price per GPU-hour, which depends on the number of servers, the term and the start date. Send a request with what you need and we come back with offers.
What is the minimum term for reserved H200 capacity?
Usually one month or more, set by each owner. Longer terms usually get a lower price per GPU-hour.
How much memory does a H200 have?
141 GB HBM3e per GPU (NVIDIA Hopper).
Can I rent a single H200 server?
Yes, where the owner allows it: each listing shows how many servers it has and you can ask for one or for all of them.

Tell us what you need

Sign in to send a request. It opens a conversation with the owners and the Paralon team, and you get an email for every reply.

Sign in to request

Free, with Google, GitHub or an email code. No commitment.