Independent comparison · no paid rankings
Home / Blog / Hosting a language model: start with constraints, not GPUs
Guide

Hosting a language model: start with constraints, not GPUs

Renting a GPU before defining latency, context, and confidentiality leads to GPU bills and a POC that never reaches production. Map usage, data, and budget first.

3 min read Updated Jul 19, 2026

Leadership wants "our own ChatGPT." The data team orders a GPU server before answering: what latency, which documents, who accesses, how many requests per day. Three months later: €2,400/month idle GPU, model never connected to IT, DPO discovering prompts contain HR data.

Hosting a language model is not buying a card. It is a chain: model choice, quantization, inference server, embedding storage, observability, prompt security. GPU comes after that mapping.

Constraints to fix before hardware

QuestionWhy it is priority
Target latencyUnder 1s interactive → GPU; overnight batch → CPU possible
Context size32k tokens multiply video memory
Sensitive dataGDPR, medical secrecy → region, logs, prompt retention
Request volume10/day vs 10,000/day changes architecture
Fine-tuning or RAG onlyRAG = vector DB; fine-tuning = training GPU

Common self-host stack

Open-weights model — verify license and commercial use. Inference server — vLLM, Text Generation Inference, llama.cpp by scale. Optional RAG — PostgreSQL/pgvector, Qdrant, separate embeddings. Reverse proxy and auth — no exposed API without rate limits. Logs — prompts may contain PII; retention policy mandatory.

CPU-only: Ollama or llama.cpp on RAM-heavy server for small teams. GPU: plan context margin; multi-GPU for 70B+ models.

Self-host vs managed API

CriterionSelf-host GPUCloud API
ConfidentialityFull control if well operatedDepends on contract and region
Initial costHigh GPU rentalLow, usage billing
OperationsDrivers, model updatesMinimal
CustomizationFine-tuning, deep RAGLimited to API options

For many SMBs, European API plus self-host RAG — data at home, inference at vendor — is the reasonable compromise when legal review accepts a subprocessed model under DPA.

Before any GPU order, define what happens during a two-week hardware outage: fallback API, degraded mode, or hard stop — if you cannot answer, you are not ready for production self-hosting.

The peak: GPU is the last budget line

Compare region and egress on the directory before signing a year of GPU rental.

Decide and move forward without blind spots

Document usage, data sensitivity, latency targets and daily request volume before any hardware purchase. Run a proof of concept on CPU or small GPU with a quantized model and real metrics — not slide decks alone. Validate legal frame: prompts, logs and subprocessors with your DPO. Scale GPU and inference only when the proof of concept has real users and measured load.

Frequently asked questions

GPU mandatory?

Interactive and medium models: yes. Batch or small quantized model: CPU possible for POC.

VRAM for 7B?

4–6 GB quantized, ~14 GB full precision; target 24 GB in production with context margin.

Self-host or API?

Sensitive data or volume → self-host or hybrid RAG. Time-to-market → European API.

Hosting in Europe?

EU region, DPA, prompt log policy — not flag alone.


Before picking a card, one question: what happens if we have no GPU for two weeks? If you have no answer, you are not ready to host an LLM — only to fund one.

Compare European hosts

Filter by compliance, location and use case — then open the sheets to verify the real scope.

Browse the directory
Blog

Related reading

All articles →