NVIDIA · Partner Business
NVIDIA Products
As an NVIDIA Authorized Distributor, Azwell AI supplies products across the entire AI infrastructure stack.
NVIDIA Architecture Generations
From today's Blackwell to next-generation Vera Rubin — the AI computing roadmap ushering in the trillion-parameter era
2024 – 2025 · Current generation
Blackwell
Superchips pairing Grace CPUs with Blackwell GPUs, plus the rack-scale NVL72, accelerate training and inference for trillion-parameter models.
- GB200 Grace Blackwell Superchip
- GB200 NVL72 rack scale (72 GPUs + 36 Grace CPUs)
- 5th-gen NVLink · 2nd-gen Transformer Engine · FP4
2025
Blackwell Ultra
An enhanced Blackwell generation with expanded HBM3e memory and inference performance, aimed at large-scale inference and agentic AI.
- GB300 · DGX B300
- GB300 NVL72 rack scale
- Increased memory capacity and inference throughput
2026+ · Next generation
Vera Rubin
An architecture pairing the next-generation Vera CPU with the Rubin GPU, taking AI factory performance to the next level with HBM4 and NVLink 6.
- Vera CPU + Rubin GPU
- Rubin NVL144 rack scale
- HBM4 · NVLink 6 · followed by Rubin Ultra NVL576 (2027)