Every NVIDIA accelerator, one control plane.
From the Hopper workhorse to rack-scale Blackwell Ultra and next-gen Vera Rubin. Specs are public NVIDIA figures, rendered as instrumentation.
7 of 7 SKUs
H100 SXM
The proven Hopper workhorse for inference and fine-tuning.
- MEM
- 80 GB HBM3
- BW
- 3.35 TB/s
- NVLINK
- Gen 4
H200 SXM
141 GB HBM3e for long-context inference without the spill.
- MEM
- 141 GB HBM3e
- BW
- 4.8 TB/s
- NVLINK
- Gen 4
B200
Blackwell training and inference. Supply-constrained — join the waitlist.
- MEM
- 192 GB HBM3e
- BW
- 8 TB/s
- NVLINK
- Gen 5
GB200 NVL72
Rack-scale frontier training. One 72-GPU NVLink domain.
- MEM
- 13.4 TB HBM3e / rack (72 × 192 GB)
- BW
- 8 TB/s
- NVLINK
- Gen 5
B300
288 GB HBM3e. 15 PFLOPS dense FP4. Built for agentic AI.
- MEM
- 288 GB HBM3e
- BW
- 8 TB/s
- NVLINK
- Gen 5
GB300 NVL72
1.1 ExaFLOPS dense FP4 per rack. The AI-factory standard.
- MEM
- 20.7 TB HBM3e / rack (72 × 288 GB)
- BW
- 8 TB/s
- NVLINK
- Gen 5
VR200 (Rubin R100)
HBM4. NVLink 6. ~3.3× GB300 inference. Reserve H2-2026 capacity.
- MEM
- 288 GB HBM4
- BW
- 22 TB/s
- NVLINK
- Gen 6
Bring your own container or launch from curated images. The full inference and training stack ships ready.
Reserve the silicon your workload needs.
On-demand for bursts, reserved for sustained training, spot for the cheapest throughput.