NVIDIA Grace Blackwell
NVIDIA Grace Blackwell is a compute platform pairing the Grace CPU (Arm architecture, Arm Neoverse V2 cores) with Blackwell GPUs. Its base unit is the GB200 superchip, which couples one Grace CPU with two Blackwell GPUs over an NVLink-C2C link.
A single GB200 superchip provides 372 GB of HBM3e GPU memory at 16 TB/s and up to 480 GB of LPDDR5X CPU memory. Tensor Core performance reaches 40 PFLOPS (NVFP4, sparse), 20 PFLOPS (FP8/FP6, sparse) and 20 POPS (INT8, sparse). GPUs communicate over fifth-generation NVLink at 1.8 TB/s per GPU.
The platform scales to rack-scale systems: GB200 NVL72 connects 36 Grace CPUs and 72 Blackwell GPUs in a single NVLink domain (13.4 TB HBM3e, 130 TB/s NVLink), while the smaller GB200 NVL4 targets HPC and AI for science. Primary uses are training and inference of large language models (including mixture-of-experts), HPC and data centers. It is also relevant to physical AI and robotics as a compute backbone. Announced at GTC on 18 March 2024.
Classification
AI Accelerator · serves as: AI Inference, High-level compute.
Which group NVIDIA Grace Blackwell belongs to and how it is built
Compute Modules is a subcategory of hardware components that provide processing power for robotic systems. It encompasses onboard computers, single-board computers (SBCs), AI accelerators, embedded processors, GPU/NPU compute modules, and other units responsible for processing sensor data and executing control logic. These modules form the foundation of modern autonomous, humanoid, and perception-capable robots.
An AI Accelerator is a specialized hardware component designed for efficient execution of artificial intelligence computations, particularly neural network inference, computer vision processing, and sensor data analysis. In robotics, AI accelerators are used to run perception models, object recognition, image segmentation, planning, and other tasks that require high computational throughput under constrained power budgets. They may take the form of dedicated NPU, TPU, VPU, or GPU chips, or specialized embedded modules.
Technical specification
Relations
Other hardware parts related to NVIDIA Grace Blackwell

Rack-scale, liquid-cooled computing system built around 72 NVIDIA Blackwell (B200) GPUs and 36 NVIDIA Grace CPUs, designed for training and inference of large AI models.
NVIDIA compute module (superchip) combining 2 Grace CPUs and 4 Blackwell GPUs on a single server board, with up to 1.3 TB coherent memory, for converged HPC and AI workloads.

Rack-scale, liquid-cooled computing system built around 72 NVIDIA Blackwell Ultra (B300) GPUs and 36 NVIDIA Grace CPUs, designed for large AI model inference and training.

NVIDIA's AI GPU architecture (2024) and the B200 chip: 208B transistors, two dies connected at 10 TB/s, TSMC 4NP, 192 GB HBM3e, 2nd-gen Transformer Engine (FP4) and 5th-gen NVLink; the basis of GB200 systems.