Flexible compute that scales with you

Unblock your team with access to cutting edge H100 GPUs, on terms that match your needs

Availability is first come first serve. Supplies are limited *

Designed for Startups

On Demand

Get access to the most powerful GPU on the market without the contract. On Demand lets you spin up compute anytime to match your team's needs.

Billed by the hour
Node online in minutes
Reserve up to 8 nodes

Simple Pricing

No hidden ingress, egress, or support costs. Pay for compute you use, at fair market rates, with a simple monthly billing break down.

No data ingress or egress fees
No hidden support or service fees
Simply hourly pricing

Peak performance

Provision an 8x H100 bare metal server, without any virtualization, running popular OS's like Ubuntu 20 and 22.

8x H100 SXM5 GPUs with NVLink
19.2TB of local NVMe PCIe storage
1TB DDR5 ECC RAM
2x CPUs with 52-cores each

Startup Spotlight

Atomic AI

Atomic AI raised $50M from Greylock to develop advanced machine learning and structural biology methods, specifically designed for RNA drug discovery.

SF Compute

Evan and Alex, both YC alumnus, are building San Francisco Compute which sells bursty compute designed for training runs to researchers, startups, and grad students.

Genmo AI

Ajay and Paras, both PhDs from UC Berkeley, are training state-of-the-art video models with support for stunning controls as well as text and image input.

Limited Availability

Register now to reserve your spot in line. Nodes will be allocated on a first come, first serve basis, and H100s are in high demand.