24–32 GB class
L4, A10, RTX 4090 or RTX 5090. Consider for smaller models only after measuring total memory, including runtime and cache.
AI infrastructure
AI infrastructure
Plan infrastructure for model hosting, inference, training and fine-tuning with explicit memory and workload requirements.
Choose a configuration in Console to review its resources and estimated cost.
What to plan for
An AI workload is more than a GPU: model weights, cache, datasets, checkpoints and an application endpoint all consume different resources.
Choose the exact model revision, precision, context length and concurrency before sizing. Validate throughput with your own workload.
Configuration choices
48–80 GB class
L40S, A40, A100 or H100. More memory may accommodate larger working sets; it does not guarantee concurrency or token speed.
141 GB and multi-GPU
H200, B200 or B300 and explicitly sized multi-GPU systems. Confirm exact model, precision, topology and serving framework.
Weight-only memory ≈ parameters × bytes per parameter. Runtime, KV cache, activations and optimizer state are additional. This is a sizing method, not a deployment recommendation.
Smart Infra Finder
Your workload, your requirementsExplore the catalog
LLM hosting
LLM hosting: model selection, memory requirements, storage and deployment choices without invented performance claims.
Explore →AI inference
AI inference: model selection, memory requirements, storage and deployment choices without invented performance claims.
Explore →AI training
AI training: model selection, memory requirements, storage and deployment choices without invented performance claims.
Explore →Fine-tuning
Fine-tuning: model selection, memory requirements, storage and deployment choices without invented performance claims.
Explore →Infrastructure for DeepSeek
Infrastructure for DeepSeek: model selection, memory requirements, storage and deployment choices without invented performance claims.
Explore →Infrastructure for Llama
Infrastructure for Llama: model selection, memory requirements, storage and deployment choices without invented performance claims.
Explore →Infrastructure for Mistral
Infrastructure for Mistral: model selection, memory requirements, storage and deployment choices without invented performance claims.
Explore →Embedding infrastructure
Embedding infrastructure: model selection, memory requirements, storage and deployment choices without invented performance claims.
Explore →Image generation infrastructure
Image generation infrastructure: model selection, memory requirements, storage and deployment choices without invented performance claims.
Explore →Speech AI infrastructure
Speech AI infrastructure: model selection, memory requirements, storage and deployment choices without invented performance claims.
Explore →Check availability: AI infrastructure
Check the current price and configuration in Console. Your choice is saved through registration. The current lookup uses mock adapters; no real capacity is reserved.
Related products and guides
GPU cloud
Compare GPU memory and architectures, from RTX to data-center accelerators, then check configurations in one console.
Explore →LLM hosting
LLM hosting: model selection, memory requirements, storage and deployment choices without invented performance claims.
Explore →Cloud servers
Cloud servers: configuration choices, sizing requirements and availability checks in GlobalAtlas AI.
Explore →