Rent an NVIDIA L40S GPU with 48 GB VRAM
Rent L40S cloud GPU servers by the hour on Cloudj: 48 GB VRAM, Ada Lovelace and about 73.3 TFLOPS FP32 for generative AI, rendering and LLM inference. It is a strong pick for creators and AI teams who need a large frame buffer with high compute, at a 350 W power limit and 626 GB/s of bandwidth. Pick a live configuration, start from the Ubuntu 22.04 + CUDA or PyTorch + JupyterLab template and connect over SSH with your key.
L40S specifications
VRAM per GPU
48 GB
Architecture
NVIDIA Ada Lovelace
Compute capability
8.9
FP32 compute
≈ 73.3 TFLOPS
Memory bandwidth
≈ 626 GB/s
Power limit
350 W
Values reported by the infrastructure provider for machines currently listed; real performance varies by machine.
Live L40S configurations and prices
From $0.96/h
incl. 32 GB disk · Live price, updated every 15 minutes
What you can run on the L40S cloud GPU
Flux and video generation
Run Flux in full precision and heavier video workflows in ComfyUI with room for LoRAs, ControlNet and upscalers.
Quantized 30B–34B LLMs
48 GB holds a 4-bit or 8-bit 30B-class model with context headroom for chat, agents or RAG.
GPU path-traced rendering
73.3 TFLOPS FP32 speeds up CUDA renderers, and 48 GB keeps big scenes and textures in memory.
Fine-tuning diffusion models
Train LoRAs or DreamBooth-style fine-tunes for SDXL and Flux without cutting resolution to fit.
Is the L40S the right GPU for your workload?
Choose the L40S when your workload requires 48 GB of VRAM, stepping beyond standard consumer cards. If your models fit into less memory, a smaller workstation or consumer GPU is more economical, whereas massive unquantized foundation models warrant a larger memory datacenter GPU.