# Where should I host an AI app in the UK? Source: https://dijituldns.co.uk/guides/where-to-host-ai-app/ Updated: 2026-10-08 > Where to host an AI app depends on whether it calls a hosted model API or runs its own model. API-based chatbots and tools run well on a VPS or managed server; self-hosted models need GPU servers. dijitul quotes UK-managed hosting for AI apps, LLM tools, chatbots and GPU workloads, built around your data and budget. ## Key facts - Most AI apps call a model through an API and do not need a GPU on their own server. - Running an open-weight model yourself usually needs a GPU with enough video memory (VRAM) for the model. - AI app backends are commonly written in Python or Node.js and run behind a reverse proxy. - Vector databases or search indexes are often used to give a model access to your own documents. - Under UK GDPR you need to know where personal data is processed, including by any AI API provider. - dijitul quotes AI, LLM app and GPU hosting rather than offering a fixed plan. ## First question: API or self-hosted model? An AI app is any website or tool that uses a machine learning model, usually a large language model (LLM), to answer questions, write text, classify documents or chat with customers. Where to host it depends on one decision: - **API-based:** your app sends requests to a hosted model from an AI provider. Your server runs the app, stores data and handles users. No GPU is needed. - **Self-hosted model:** you run an open-weight model on your own hardware. This needs a GPU server with enough video memory for the model you choose. Most small business AI projects start API-based, because it is quicker and cheaper to build. Self-hosting makes sense for data control, predictable costs at high volume, or custom models. ## What an API-based AI app needs - **A VPS or managed server** with enough memory for your app and database. - **A runtime** such as Python (FastAPI, Django, Flask) or Node.js, kept running by a process manager. - **A reverse proxy** such as nginx, with SSL. - **A database** such as MySQL, MariaDB or PostgreSQL, and often Redis for queues and caching. - **A vector store or search index** if the app answers questions from your own documents. - **Safe key storage**, so API keys sit in environment variables, never in code. - **Background workers and cron** for document processing and scheduled jobs. See [LLM app hosting](https://dijituldns.co.uk/llm-app-hosting/) and [AI chatbot hosting](https://dijituldns.co.uk/ai-chatbot-hosting/) for more detail. ## What a self-hosted model needs - **A GPU server** with enough VRAM for the model size and the number of users at once. - **Drivers and an inference server** to load the model and answer requests. - **Plenty of fast disk**, as model files are large. - **Monitoring** for GPU memory, temperature and response times. GPU servers cost far more than standard hosting, so size them carefully. A smaller model on modest hardware is often good enough for a focused business task. See [GPU servers](https://dijituldns.co.uk/gpu-servers/). ## Data protection and UK hosting If your AI app handles personal data, UK GDPR applies. Before you build, write down: - What personal data goes into the app, and why. - Where it is stored, including logs and backups. - Whether it is sent to an AI API provider, and where that provider processes it. - How long it is kept, and how a person can ask for it to be deleted. Hosting the app and its database in the UK, with a UK company managing it, keeps the main data store close to home. Read our guide on [GDPR and web hosting](https://dijituldns.co.uk/guides/gdpr-web-hosting-uk/) and our [GDPR-compliant hosting](https://dijituldns.co.uk/gdpr-compliant-hosting/) page. ## Questions to answer before choosing - Will the app call an AI API, run its own model, or both? - How many people will use it at the same time, and at what times of day? - Does it handle personal or confidential data? - Does it need to read your own documents, and how many? - Who will maintain the server, updates and security after launch? Clear answers make it much easier to size the server and avoid paying for GPU power you do not need. ## How dijitul helps AI hosting is not one size fits all, so dijitul **quotes** each project. We look at your app, expected users, data and budget, then recommend a managed VPS, a dedicated server or a GPU server. Our UK team sets up the runtime, SSL, database, backups and monitoring, and keeps it maintained. We can also host the website around your AI tool, and handle [custom software hosting](https://dijituldns.co.uk/custom-software-hosting/) and [API hosting](https://dijituldns.co.uk/api-hosting/). Start at [AI hosting](https://dijituldns.co.uk/ai-hosting/), or call 01623 650333 or email info@dijitul.uk to request a quote. ## FAQs ### Do I need a GPU to host an AI chatbot? Not usually. Most AI chatbots send each question to a hosted model through an API, so the heavy work happens on the provider's hardware. Your own server only runs the app, database and website, which a VPS or managed server handles easily. You only need a GPU server if you want to run the model yourself. ### Can I host an AI app on shared hosting? Usually not well. AI apps often need long-running Python or Node.js processes, background workers, queues and custom packages, which many shared hosting plans do not allow. A VPS or managed server gives you that control. A simple PHP website that calls an AI API for one feature can sometimes run on standard managed hosting. ### Is it GDPR compliant to use an AI API with customer data? It can be, if you do it properly. You need a lawful basis, a data processing agreement with the AI provider, clear privacy information, and to know where data is processed and how long it is kept. Avoid sending personal data the task does not need. Take advice for sensitive data such as health or legal records. ### How much does it cost to host an AI app? It varies widely. An API-based app on a VPS costs far less than a self-hosted model on a GPU server, but you also pay the AI provider per use. Costs depend on users, data volume and model size. Get a quote based on your actual app, and estimate API usage before you launch. ### Does dijitul host AI apps in the UK? Yes. dijitul quotes managed hosting for AI apps, LLM tools, chatbots and GPU workloads, managed by our UK team at DIJITUL LTD, trading since 2006. We recommend a VPS, dedicated server or GPU server based on your app and budget. Call 01623 650333 or email info@dijitul.uk to request a quote. ## Pricing and ordering This is quoted to fit. Request a quote at https://dijituldns.co.uk/quote/ or call 01623 650333.