
1. Technology overview
Onprem Technologies GmbH, operating as OnPrem AI, is based in Schaffhausen, Switzerland. A partner press release published on parashift.ai states that it has operated since 2023 with customers in the DACH region and Central Europe. The company lets organisations run language models and agents on their own premises instead of a public cloud.
The offer combines on-premise AI servers with a software layer that acts as an operating system for enterprise AI. It manages servers, GPUs, models, APIs, monitoring and updates, and provides a catalogue of language models tested for the installed hardware, an OpenAI-compatible API gateway and an operations dashboard. Services cover use-case guidance, security testing and integrations. The company is an NVIDIA Inception member and a Dell technology partner.
2. Highlighted features & specifications
- Model catalogue: language models tested and optimised for the installed hardware, with a VRAM check before a model goes live.
- OpenAI-compatible API: authenticated gateway supporting SGLang, vLLM and Llama.cpp back ends.
- Operations dashboard: web view of status, alerts and usage.
- Governance: controls mapped to ISO 27001 and SOC 2 with auditable, air-gap-capable updates.
- Hardware: single or cluster servers with NVIDIA RTX PRO 6000, H100, H200 or GH200 and Kubernetes.
- Services: use-case guidance, security tests and audits, and integrations for agents, RAG and MCP.
3. Potential application
A Vietnamese bank could host an internal knowledge assistant on its own servers so customer data never leaves the data centre. A hospital group could run document-reading models for medical records on-premises. A manufacturer could deploy private RAG over engineering documents on one server cluster.
4. Benefits
Documents and prompts stay inside the customer's own data centre, which helps with confidentiality rules. The partner release states that customers own the hardware and pay a flat fee that does not depend on tokens processed. A tested model catalogue and a free-VRAM check before go-live reduce deployment trial and error. Updates are auditable and can be applied air-gapped.
5. Market trends & opportunities
Organisations in regulated sectors increasingly want generative AI without sending data to foreign clouds, and per-token cloud costs push some towards owned hardware. Vietnamese banks, insurers, hospitals and government bodies face data-localisation expectations and may look for packaged on-premise AI instead of building GPU projects themselves.
6. Other information
Onprem Technologies GmbH, Zur Stahlgiesserei 8, 8200 Schaffhausen, Switzerland. Member of the NVIDIA Inception Program and Dell technology partner. Listed partners and customers include Parashift, agorum, Orderfox and Rosenau AG. Founding year (2023) is taken from the Parashift partnership release of 2 November 2026.
One platform connecting startups, corporations, research institutions, regions and investors











Playbook & Handbook · February 6, 2026
Winning Accelerator Playbook 2026 by BambuUP
RETAIL E-COMMERCE: AI AND REAL-TIME DATA ARE TRANSFORMING HOW RETAIL BUSINESSES OPERATE








































