NVIDIA Data Center GPUs

Enterprise AI training and inference GPUs — H100, H200, B200, B300, GB200, L40S, L4.

32 Products32 AI Accelerators

Related manufacturers:NVIDIA products

AI Accelerators (32)

Pre-Order

NVIDIA GB200 Grace Blackwell Superchip

Flagship superchip combining Grace CPU with dual Blackwell GPUs via NVLink 5.0 for trillion-parameter AI workloads.

ArchitectureBlackwell

Memory384GB HBM3e + 480GB LPDDR5X

TDP1000W

Pre-Order

NVIDIA B300 Tensor Core GPU

Next-generation Blackwell GPU with 288GB HBM3e memory and native FP4 tensor core support for generative AI.

ArchitectureBlackwell

Memory288GB HBM3e

TDP900W

Pre-Order

NVIDIA B200 Tensor Core GPU

Flagship Blackwell GPU with 384GB HBM3e memory and 10 TB/s bandwidth for extreme-scale AI.

ArchitectureBlackwell

Memory384GB HBM3e

TDP1000W

In Stock

NVIDIA H200 Tensor Core GPU

Advanced Hopper GPU with 141GB HBM3e memory and 4.8 TB/s bandwidth for large model inference.

ArchitectureHopper

Memory141GB HBM3e

TDP700W

In Stock

NVIDIA H100 Tensor Core GPU

Industry-standard AI training GPU with 80GB HBM3 and 3.35 TB/s bandwidth in SXM5 form factor.

ArchitectureHopper

Memory80GB HBM3

TDP700W

In Stock

NVIDIA L40S GPU

Versatile data center GPU with 48GB GDDR6 ECC memory for AI inference, graphics, and video.

ArchitectureAda Lovelace

Memory48GB GDDR6 ECC

TDP350W

In Stock

NVIDIA L4 GPU

Energy-efficient data center GPU with 24GB GDDR6 for AI inference, edge computing, and video.

ArchitectureAda Lovelace

Memory24GB GDDR6

TDP72W

In Stock

NVIDIA A100 Tensor Core GPU

Industry-defining data center GPU with 80GB HBM2e and 3rd-gen Tensor Cores for AI training and HPC.

ArchitectureAmpere

Memory80GB HBM2e

TDP300W (PCIe) / 400W (SXM4)

In Stock

NVIDIA A30 GPU

Mainstream AI training and inference GPU with 24GB HBM2e and MIG for balanced enterprise deployments.

ArchitectureAmpere

Memory24GB HBM2e

TDP300W

In Stock

NVIDIA A10 GPU

Versatile data center GPU with 24GB GDDR6 for AI inference, graphics, and video analytics.

ArchitectureAmpere

Memory24GB GDDR6

TDP150W

Limited

NVIDIA V100 Tensor Core GPU

Pioneering Tensor Core GPU with 32GB HBM2 memory that revolutionized deep learning training.

ArchitectureVolta

Memory32GB HBM2

TDP300W

In Stock

NVIDIA T4 GPU

Energy-efficient inference GPU with 16GB GDDR6 for ML inference, video transcoding, and VDI workloads.

ArchitectureTuring

Memory16GB GDDR6

TDP70W

Pre-Order

NVIDIA B100 Tensor Core GPU

High-performance Blackwell GPU with 192GB HBM3e memory and FP4 support for efficient AI training.

ArchitectureBlackwell

Memory192GB HBM3e

TDP700W

In Stock

NVIDIA H800 GPU

Hopper-based data center GPU with 80GB HBM3 and reduced interconnect for regulated markets.

ArchitectureHopper

Memory80GB HBM3

TDP700W

In Stock

NVIDIA A800 GPU

Ampere-based GPU with 80GB HBM2e and reduced interconnect for availability in export-controlled regions.

ArchitectureAmpere

Memory80GB HBM2e

TDP400W

In Stock

NVIDIA L40 GPU

Professional-grade data center GPU with 48GB GDDR6 for visual computing, AI inference, and graphics.

ArchitectureAda Lovelace

Memory48GB GDDR6 ECC

TDP300W

In Stock

NVIDIA RTX 4500 Ada Generation

Mid-range professional GPU with 24GB GDDR6 for CAD, 3D rendering, and AI-accelerated workflows.

ArchitectureAda Lovelace

Memory24GB GDDR6 ECC

TDP200W

In Stock

NVIDIA A100 SXM Tensor Core GPU

Flagship SXM variant of the A100 with 80GB HBM2e memory and 400W TDP for AI training and HPC.

ArchitectureAmpere

Memory80GB HBM2e

TDP400W

In Stock

NVIDIA A100 PCIe Tensor Core GPU

PCIe variant of the A100 with 80GB HBM2e and 300W TDP for broad server compatibility.

ArchitectureAmpere

Memory80GB HBM2e

TDP300W

In Stock

NVIDIA RTX 4000 Ada Generation

Entry-level professional Ada GPU with 20GB GDDR6 ECC for CAD, rendering, and AI inference.

ArchitectureAda Lovelace

Memory20GB GDDR6 ECC

TDP130W

In Stock

NVIDIA H100 NVL Tensor Core GPU

Dual-GPU Hopper accelerator with 188GB HBM3 (2x 94GB) for power-efficient LLM inference and training.

ArchitectureHopper

Memory188GB HBM3 (2x 94GB)

TDP600W

In Stock

NVIDIA H100 NVL PCIe Tensor Core GPU

PCIe dual-GPU Hopper accelerator with 188GB HBM3 for broad server compatibility and LLM inference.

ArchitectureHopper

Memory188GB HBM3 (2x 94GB)

TDP600W

In Stock

NVIDIA RTX 5000 Ada Generation

Data center Ada GPU with 32GB GDDR6 ECC for AI inference, graphics rendering, and virtual workstations.

ArchitectureAda Lovelace

Memory32GB GDDR6 ECC

TDP250W

In Stock

NVIDIA RTX 4000 Ada Generation

Compact data center Ada GPU with 20GB GDDR6 ECC for edge AI inference, video analytics, and VDI.

ArchitectureAda Lovelace

Memory20GB GDDR6 ECC

TDP130W

In Stock

NVIDIA RTX 6000 Ada Generation

Flagship professional GPU with 48GB GDDR6 ECC memory for AI, rendering, and scientific visualization.

ArchitectureAda Lovelace

Memory48GB GDDR6 ECC

TDP300W

In Stock

NVIDIA RTX 5000 Ada Generation

Mid-range professional GPU with 32GB GDDR6 ECC for workstation AI and graphics.

ArchitectureAda Lovelace

Memory32GB GDDR6 ECC

TDP250W

In Stock

NVIDIA GH200 Grace Hopper Superchip

Integrated superchip combining Grace CPU with Hopper GPU via NVLink-C2C for HPC and AI convergence.

ArchitectureHopper + Grace

Memory96GB HBM3 + 528GB LPDDR5X

TDP1000W

In Stock

NVIDIA RTX 6000 Ada Generation

Flagship professional workstation GPU with 48GB GDDR6 ECC for AI development, rendering, and scientific visualization.

ArchitectureAda Lovelace

Memory48GB GDDR6 ECC

TDP300W

In Stock

NVIDIA RTX 5000 Ada Generation

Professional workstation GPU with 32GB GDDR6 ECC for CAD, 3D rendering, and AI-accelerated workflows.

ArchitectureAda Lovelace

Memory32GB GDDR6 ECC

TDP250W

In Stock

NVIDIA RTX 4000 Ada Generation

Mid-range professional workstation GPU with 20GB GDDR6 ECC for CAD, 3D content creation, and AI inference.

ArchitectureAda Lovelace

Memory20GB GDDR6 ECC

TDP130W

In Stock

NVIDIA RTX 3500 Ada Generation

Entry-level professional workstation GPU with 12GB GDDR6 ECC for CAD drafting, visualization, and light AI.

ArchitectureAda Lovelace

Memory12GB GDDR6 ECC

TDP70W

In Stock

NVIDIA RTX 2000 Ada Generation

Compact professional workstation GPU with 16GB GDDR6 ECC for 2D/3D CAD, edge AI, and power-constrained workstations.

ArchitectureAda Lovelace

Memory16GB GDDR6 ECC

TDP70W