Run GPT-5-class AI
on hardware
you own.
On-premise LLM appliance for German SMEs. EU AI Act compliant. From €2,950. Pays back in under 18 months.
Your AI. Your data. Your server room.
Built on NVIDIA Blackwell hardware
Built for the right kind of buyer
We've kept the scope tight so the value is honest. If you're outside this profile, we'd rather tell you up front than waste your time.
Built for
Companies with sensitive data and real cloud-AI spend.
- 10–500 employees
- Legal, medical, financial, defense supply chain, R&D-heavy manufacturing
- Currently paying €15,000+ per year for cloud AI
- Or barred from cloud AI for compliance reasons
Not for
If any of these sound like you, we're not a fit — yet.
- Solo consultants and one-person shops
- Companies under €2M annual revenue
- Teams that need fewer than 5 AI users
- Anyone looking for a cheap experiment, not a workhorse
Cloud AI is draining your budget — every single month
Most companies don't realize how fast AI subscription costs compound. Here's what 100 users actually costs you per year.
With the EU AI Act taking effect August 2, 2026, companies using AI for high-risk tasks face mandatory compliance — or fines up to €35M.
Data Leaves Your Building
Unpredictable Costs
Regulatory Risk
Vendor Outages
Enterprise AI that lives in your server room
A compact, GPU-powered appliance pre-loaded with state-of-the-art open-source AI models. Plug in, connect to your network, and start using AI in hours — not weeks.
NVIDIA Professional GPUs
8B – 120B+ Parameter Models
Up to 100 Concurrent Users
RAG-Ready Architecture
Pre-configured & Secure
From order to AI in three steps
No cloud accounts, no complex setup, no ongoing vendor management. We handle the hard parts so your team can focus on using AI.
We Configure
Tell us your team size, use cases, and compliance needs. We select the right GPU, models, and software stack — tailored to your workload.
We Deploy
Your pre-configured appliance ships ready to plug in. We assist with network integration, user onboarding, and initial model tuning — remotely or on-site.
Your Team Uses AI
Your employees access AI through a familiar chat interface — no training needed. Documents stay local, costs stay flat, and you stay in control.
Four reasons to bring AI in-house
Data Sovereignty
Predictable Costs
No Vendor Lock-in
Always Available
240+ pre-built agent templates — ready for every department
Curated automation recipes across 22 industries — all running on-premise, behind your firewall, fully GDPR-compliant.
240+ pre-built workflow templates across 24 industries — customize, combine, or build your own.
Built for real deployments, not slide decks
WerkHub AI is engineered for SMEs that need GPT-5-class capability inside their own network. Early-access partners get hands-on engineering support and pricing locked at today's BOM.
Early-access program now open
First 10 customers receive 12 months of premium support and a free on-site discovery session (up to 500 km from our Haskovo/Sofia office) — no extra cost.
- 12 months premium support included
- Free on-site discovery session
- Pricing locked at current BOM
See a real deployment
Get the detailed PDF — hardware specs, integration approach, QA process, and project timeline from a recent build.
Want the detailed PDF with full specs?
Component list, architecture diagrams, and project timeline.
Choose your configuration
Every appliance ships pre-configured with the full software stack — ready to deploy in hours. Custom configurations available on request.
All prices are estimated BOM + assembly. Final pricing depends on configuration, GPU availability, and current component market conditions.
Includes: hardware, pre-installed software stack, initial model deployment, and setup documentation.
Up to 50% government subsidies
German funding programs can significantly reduce your investment. Here are the key programs at a glance.
Digitalbonus Bayern
PlusGrant covering up to 50% of digitalization project costs (max €60k). Covers the full AI solution — software, deployment, integration, training, and appliance hardware.
Innovationsgutschein
BayernFor custom AI development and innovation. Standard tier: up to €22,800 (60%). Special tier: up to €49,750 (50%). Covers R&D services, not hardware.
Forschungszulage
FederalFederal R&D tax credit. 25% of eligible costs (35% for SMEs), up to €1M/year. No application deadline — claimed with your annual tax return.
Common questions
Everything you need to know about deploying on-premise AI with WerkHub AI.
Every appliance ships with enterprise-grade components rated for 24/7 operation. We offer optional maintenance contracts with next-business-day hardware replacement. The system uses RAID storage, so a single drive failure won't cause data loss. For critical deployments, we recommend a redundant configuration.
You have full control over model updates. When a new open-source model is released, we provide a tested update package that you can apply at your convenience — no forced updates, no surprise changes. Your team can also download and deploy models independently. We offer an optional managed update service if you prefer hands-off maintenance.
Yes. The appliance runs standard inference frameworks (vLLM, Ollama) that support any GGUF or safetensors model. If you have fine-tuned models from your own training pipeline, you can deploy them directly. We also offer fine-tuning services if you want to customize models for your specific domain.
From order to a fully running AI system: typically 2–4 weeks. The appliance ships pre-configured with your chosen models and software stack. On-site setup usually takes less than a day — connect power, connect to your network, and your team can start using AI immediately.
The EU AI Act (effective August 2, 2026) classifies many business AI uses as high-risk, requiring full transparency, audit trails, and data governance. With on-premise AI, your data never leaves your building — eliminating cross-border data transfer risks. You have complete control over model behavior, logging, and compliance documentation. We provide compliance documentation templates as part of the deployment.
Every appliance includes 12 months of technical support covering software configuration, model deployment assistance, and troubleshooting. Extended support contracts are available with guaranteed response times. We also provide initial user training and onboarding documentation for your team.
For a 100-person team, cloud AI subscriptions cost €30,000–72,000+ per year. The WerkHub AI appliance is a one-time investment starting at €5,900 (from €2,950 with Digitalbonus) and scaling to €65,000 for enterprise deployments. Minimal ongoing costs (electricity and optional maintenance). Most customers see full ROI within 12–18 months. Use our ROI calculator for a personalized comparison.
The appliance runs completely offline — no internet connection required for AI inference. This makes it suitable for high-security environments. Internet is only needed if you want to download new models or receive software updates, and even then, models can be transferred via USB for fully air-gapped setups.
Book a free consultation
Tell us about your team and AI needs. We'll follow up with a tailored solution, configuration recommendation, and quote.