Enterprise data center GPU for AI inference, virtual workstations, and professional graphics

Rent NVIDIA L40 GPUs for production AI workloads, virtualized graphics, and enterprise applications. Get the reliability and features of data center GPUs with flexible on-demand pricing.
| L40 | RTX 4090 | % Diff | |
|---|---|---|---|
| Architecture | Ada Lovelace | Ada Lovelace | N/A |
| CUDA Cores | 18 176 | 16 384 | +10.9% |
| Tensor Cores | 568 (4th gen) | 512 (4th gen) | +10.9% |
| RT Cores | 142 (3rd gen) | 128 (3rd gen) | +10.9% |
| Memory Type | GDDR6 ECC | GDDR6X | N/A |
| VRAM | 48 GB | 24 GB | +100% |
| Bus Width | 384-bit | 384-bit | 0% |
| Bandwidth | 864 GB/s | 1 010 GB/s | −14.5% |
| FP32 Performance | ~90.5 TFLOPS | ~82.6 TFLOPS | +9.6% |
| TDP | 300 W | 450 W | −33.3% |
| PCIe | PCIe 4.0 ×16 | PCIe 4.0 ×16 | N/A |
| Form Factor | Dual-slot | Triple-slot | N/A |
| Cooling | Passive | Active | N/A |
| Display Outputs | 4× DP 1.4a | 3× DP 1.4a | N/A |
| vGPU Support | Yes | No | N/A |
Optimized for AI inference and training with 18,176 CUDA cores delivering ~90.5 TFLOPS FP32 performance. Perfect for production-scale AI deployments.
Built for multi-tenant cloud environments with NVIDIA vGPU support, enabling secure workstation virtualization and remote graphics workloads.
48GB ECC memory and advanced encoding (3x NVENC/NVDEC with AV1) enable real-time ray tracing, 8K video workflows, and complex 3D scene rendering.
Common questions about renting NVIDIA L40 GPUs
We're here to support your compute and AI needs. Let us know if you're looking to: