AI servers: your own private AI, on hardware you own

CloudHosting's private GPT processes invoices, contracts, HR files and medical records on hardware you own, in your office or in its Riga data centre, so no personal data goes to a third-party AI provider. If the unit stands in your office, no processing agreement is needed for the model itself; if it is colocated in our Riga data centre, we act as your processor for hosting it and sign a data processing agreement with you. Either way the processing stays inside the EU and your NIS2 supplier list stays short. You are still the data controller, so internal access rules still matter: the setup includes API keys, an audit log and rate limits, so you can control and document who used the model and when.

A CloudHosting AI server with 128 GB of unified memory runs Llama 3.3 70B quantized to FP8, which keeps answers close to the original model, or to 4-bit for faster answers and more room for long documents. The full-precision weights of a 70B model take about 141 GB and do not fit on one unit. Smaller local models such as Qwen or Mistral fit with plenty of room and answer noticeably faster, so they are the usual choice when many people use the box at once. Models run on vLLM or llama.cpp, so swapping one for another is routine, and under the managed plan we refresh them quarterly as releases like Llama 4 or Qwen 3 ship. If you are unsure which model fits, we will size it with you.

The CloudHosting private GPT exposes the standard OpenAI chat-completions API on your LAN or VPN, so libraries, SDKs and plugins built for OpenAI endpoints work after you change only the base URL and the API key. Existing integrations keep running without code changes, which is the point of standardising on this interface. The endpoint is reachable only inside your network, not from the public internet, and each application can get its own key. Alongside the API you get a chat UI for staff who simply want to ask questions in the browser. If a specific tool in your stack is critical, name it to support and we will verify it before deployment.

CloudHosting's optional RAG setup lets the private model answer from your internal documents: sources such as SharePoint are indexed in Qdrant or pgvector on the same machine, so the index never leaves your network either. When someone asks a question, the relevant passages are retrieved and the model answers from them instead of guessing. Under the managed plan the index is rebuilt every night, so yesterday's edits show up in answers, and we tune retrieval as your content grows. Answer quality depends on how clean the source documents are, which is honest to say upfront. List your sources when you request a quote and we will confirm what can be connected.

A CloudHosting AI server lives either on your premises, reached over the LAN, or in the Tier 3+ data centre CloudHosting runs itself in Riga, where your team connects over VPN. On your premises, inference does not depend on an internet connection. In both cases the hardware is your property, and every byte on it, models, indexes and logs included, belongs to you. Because you own the box, you can start in the office and move it to the data centre later, or the other way around. Delivery within Riga is included with the purchase; delivery elsewhere is arranged individually, so mention your location when asking for a quote.

CloudHosting's managed plan for AI servers is optional and starts at 500 EUR a month; setup alone already hands you a complete, hardened service with a backup and disaster-recovery plan that you can run yourself from day one. Be honest about what that means: someone on your side then watches GPU load and latency, applies OS and CUDA patches, and swaps in new model releases. The managed plan takes exactly that off your plate and adds 4 hours of engineer time every month for changes and requests. You can also start on your own and add the plan later if maintenance turns out to be a burden; write to support and it is switched on.

A private GPT server from CloudHosting costs three parts: the hardware as a one-time purchase at the price shown on this page, setup from 5000 EUR one-time, and an optional managed plan from 500 EUR a month, all without VAT. Setup covers deployment, hardening and the API and chat UI going live. There are no per-token fees, so the monthly cost does not grow when you onboard more users or your team starts using it heavily. If you already own a suitable server, an accelerator card alone can be the cheaper route. For an exact figure for your case, request a quote and we will itemise it.

One CloudHosting AI server on the NVIDIA GB10 chip with 128 GB of memory serves a small group on a 70B model, or more people at once on a mid-size model, and the NVIDIA GB10 unit can be linked with a second one when the team outgrows it. There is no single fixed number, because it depends on the model, the context length and how your team works. Concurrent means people sending requests at the same moment, and in a typical office only a fraction of staff is chatting at once, so the total number of accounts can be much larger. A 70B model suits a group that values answer depth over speed, while a mid-size model answers faster. Tell us your headcount and expected usage during the sizing call and we will recommend a model and configuration.

The two AI servers CloudHosting sells, NVIDIA GB10 and ASUS Ascent GX10, run models identically on the same GB10 Blackwell chip with 128 GB of unified memory, and differ in storage, 4 TB against 1 TB of NVMe, and in cluster support. Inference speed and the models they fit are the same. The NVIDIA unit supports clustering; the ASUS unit runs single-node only. In practice storage matters once you keep several models and a growing RAG index on the box, and cluster support matters only if you expect to add nodes later. A stable small team is well served by the ASUS unit. Setup and the managed plan are the same for both, and delivery within Riga is included either way.

Your contract for an AI server is with CloudHosting SIA, a Latvian IT infrastructure provider operating since 2016, which runs its Tier 3+ data centre in Riga itself. The hardware you buy becomes your property, and setup and the managed plan are invoiced by the same company. Invoices are issued and paid in EUR: companies in Latvia pay 21% VAT, EU companies with a valid VAT number are invoiced under reverse charge, and companies outside the EU are invoiced without VAT. If the unit is hosted in our data centre, we sign a data processing agreement with you as your processor. Latvian law applies.

CloudHosting engineers answer in English, Latvian and Russian on working days from 9:00 to 17:00 Riga time, and the managed plan adds 24/7 monitoring of GPU load, VRAM, queue and latency. Setup, the sizing call and questions after go-live are handled by the same team, by ticket, email or phone. Without the managed plan the unit is yours to run: we answer questions in working hours, and your side watches the box. With the managed plan we also patch the OS and CUDA in a monthly window, refresh models quarterly, re-index RAG every night and give you 4 hours of engineer time every month.

With a CloudHosting AI server you own the hardware outright, so leaving means taking the unit with you: the models, indexes, logs and documents on it stay yours. There are no per-token fees and no subscription for the box itself. The managed plan ends when you stop paying its next invoice; to end it early, inside a period you have already paid for, write to info@cloudhosting.lv 20 days in advance. If the unit is colocated in our Riga data centre, we hand it back to you by arrangement. Because the service runs on open tools, vLLM or llama.cpp, Qdrant or pgvector and an OpenAI-compatible API, anyone you choose can take over running it.