$1.30 / gpu / hour - 80 RTX PRO 6000s available from 05/29/2026 to 12/31/2026 (Salt Lake City, UT)

80× NVIDIA RTX PRO 6000 Blackwell | 96 GB VRAM | Salt Lake City, UT | Available immediately

We're a GPU infrastructure company with 10 bare-metal Supermicro servers, each with 8× RTX PRO 6000 Blackwell GPUs (96 GB VRAM per card). All 80 GPUs are online and available today — no waitlist.

Why 96 GB matters: Serve 70B parameter models on a single GPU without tensor splitting. Run the largest open-weight image and video diffusion models (FLUX, SD3, CogVideoX) on one card. Fine-tune 34B–70B models on a single 8-GPU node without CPU offloading.

Workload fit: LLM inference & serving, fine-tuning, diffusion image/video/audio generation, 3D rendering, computer vision, synthetic data generation.

Pricing: On-demand available at $1.3/GPU-hr — well under RunPod, CoreWeave, GCP, and AWS. For weekly, monthly, or reserved terms, reach out and we'll work something out.

Minimum booking is one node (8 GPUs). Happy to spin up a free test node for any serious buyer - SSH access within the hour.

We beat RunPod, Hyperstack, CoreWeave, GCP, and AWS on price. Spot and on-demand available with no minimum commitment. Happy to do a free test node for any serious buyer — SSH access within the hour.

Contact: info@helios.co