
Hopper
Shown: H100 PCIe
| Model | Memory per GPU | Form factor |
|---|---|---|
| H100 | 80 GB | SXM, PCIe |
| H200 | 141 GB | SXM, NVL |
Applied Work Compute
GPU capacity for evaluation, reinforcement learning, and post-training. We source operator offers around your workload and coordinate the commercial process.
Hopper and Blackwell capacity from operators, matched to your configuration. Specify SXM or PCIe, memory, and interconnect.

Shown: H100 PCIe
| Model | Memory per GPU | Form factor |
|---|---|---|
| H100 | 80 GB | SXM, PCIe |
| H200 | 141 GB | SXM, NVL |

Shown: HGX B200 board
| Model | Memory per GPU | Form factor |
|---|---|---|
| B200 | 180 GB | HGX (SXM) |
| B300 | 288 GB | HGX (SXM) |
Memory per GPU as published by NVIDIA. Availability, rates, and dates are confirmed per request.
Images: H100 PCIe by Geekerwan, CC BY 3.0. HGX B200 by Pokiiri, CC BY-SA 4.0. Both via Wikimedia Commons (cropped, tinted).
Defined requirements. Confirmed offers. Direct contracts.
GPU model, count, region, timing, term, and budget.
Led byYou
We confirm quotes and availability, with referral terms agreed first.
Led byApplied Work
Configuration, delivery, and service terms are confirmed with the operator.
Led byApplied Work and operator
You contract with and pay the operator. They provide and support the capacity.
Led byYou and operator
Configuration, cost, and delivery reviewed together, so operator offers can be compared line by line.
Share the configuration, timing, and constraints. We’ll establish whether a suitable operator offer is available.
Share your configurations, regions, lead times, and minimum terms. We agree referral terms before introducing a buyer.