.png)
.png)
.png)
%20(2).png)
%20(3).png)
.png)
.svg)
.svg)
Obelisk Portfolio
Highlights
NVIDIA H100
80 GB HBM3 · the gold standard for training and serving frontier-scale models. Available on demand, billed by the hour.
NVIDIA A100
80 GB HBM2e · workhorse for production training and large-model inference. Reliable supply at half the cost of H100.
RTX 4090
24 GB · best $/TFLOP for inference, fine-tuning, and rendering. Ready in under a minute.
L40S
48 GB · datacenter card built for AI inference at scale, with excellent throughput per dollar.
RTX A6000
48 GB · pro-grade for memory-heavy workloads, multi-model serving, and large-batch fine-tuning.
A40
48 GB · cost-efficient entry into datacenter-class accelerators — perfect for development and batch jobs.
Strategic Exposure to the AI Compute Stack
Real GPUs rented by the hour. Pay with Solana, connect in 60 seconds. No contracts, no minimums.
60-Second Setup
Pick a GPU, pay, get a live pod in under a minute. No tickets, no waiting list.
Pay With Solana
Send Solana directly from your wallet. No KYC, no card on file, no monthly minimums.
Pre-Built Images
Start with PyTorch, CUDA, or your own Docker image. Ships ready for AI workloads out of the box.
Hourly Billing
Rent by the hour. Stop the pod when you’re done — billing stops the moment it does.
SSH & Jupyter
Every rental ships with SSH and Jupyter access on a public endpoint. Full root.
Global Regions
Capacity in US East/West, EU, and APAC — pick the region closest to your data.
Real Capacity
Pricing in USD, settled in Solana. What you see at quote time is what you pay — no surge pricing, no hidden fees.
