Cloud-only LLMs break on jurisdiction, logging, and control. DWS ships the NVIDIA stack on-site, with customer-owned hardware, bridged inference, and auditable handoff.
When your team needs a sovereign AI brain that lives in your own building — not in someone else's cloud. One Acer Veriton GN100, NVIDIA model stack installed on-site, configured for how you actually work. Built for places like VPK (vapaapalokunta) where decisions can't wait for a cloud round-trip.
NVIDIA is the credibility layer — production silicon and models customers already trust. The actual moat is what surrounds the box: jurisdiction, ownership, bridged architecture, auditable handoff, and a Finnish service team. Common LLM vendors can copy model access. They cannot easily copy the jurisdiction, the install, the compliance trail, or the service wrapper.
OpenAI, Anthropic, Google Vertex, Mistral cloud — every cloud-only LLM hits the same four hard failures the moment the workload sits inside an EU-regulated industry:
| Failure mode | Why it kills the use case |
|---|---|
| CLOUD Act exposure | US-incorporated cloud means US law can compel access to your customer data. Defence, healthcare, dual-use, critical infrastructure: instant deal-breaker. |
| GDPR Article 5(f) & Schrems II | Customer data crossing to a third-country processor without an adequate lawful basis is automatic GDPR liability. |
| NIS2 — 24 h incident report | Critical entities (energy, water, transport, finance, health, digital, public admin) must report incidents in 24 h. You can't reconstruct what the model did when the logs live in someone else's cloud. |
| EU AI Act Article 12 logging | Required immutable record-keeping for high-risk AI. Cloud LLMs don't surface prompt / response / decision logs at the granularity an inspector demands. |
NVIDIA does the credibility work. Buying this stack is buying the same NVIDIA components NVIDIA itself ships — not a research project, not a custom-silicon bet, not vendor risk. Customers, inspectors, and procurement teams already trust the NVIDIA name. That's half the sales conversation gone before we walk in.
Itemised. No black box. Customer signs off on what arrived, what is installed, and what gets pushed each month.
| Item | What it is | When |
|---|---|---|
| NVIDIA hardware | Acer Veriton GN100 (or DGX Spark / Veriton 2000 / customer-supplied NVIDIA-capable) — pre-imaged, customer-owned | Day 0 |
| On-site install | Lifetime engineer: validates CUDA drivers, loads NVIDIA NIM bundle, brings up local MCP endpoint, sanity-tests first inference end-to-end | Day 0–1 |
| Operator training drill | 30-min scenario rehearsal with your team · signed install protocol · runbook handover | Day 1 |
| Article 12 immutable log | Live from minute one · 7-year retention · cryptographic hash chain · sealable for inspector handover | Day 1 |
| Monthly NVIDIA refresh | New Llama Nemotron NIMs, Cosmos checkpoints, Nemotron Speech, TensorRT-LLM kernels, GB10-tuned quantization | Monthly |
| Weekly security backports | CUDA driver patches, container CVE fixes, OpenSandbox + Ubuntu 24.04 LTS updates | Weekly |
| Daily hotfixes (when needed) | Zero-day CVE patches, NVIDIA driver hotfixes, model-side jailbreak mitigations — pushed within 24 h | As needed |
| 24/7 technical support | Named Finnish engineer · SLA sized to your operation · production incidents covered | Ongoing |
| Warranty servicing | We handle hardware warranty for you (Acer GN100 RMA paperwork, swap-out coordination, redeploy) | Ongoing |
| Remote configuration | Bundle switches, patches, OS updates over your secure tunnel — sovereign always | Ongoing |
You're not buying a sealed appliance. The GN100 ships with a full Ubuntu 24.04 LTS stack you own and extend. Cron jobs, systemd timers, Docker compose stacks, custom Python services, REST gateways — anything that runs on Linux runs here, alongside the NVIDIA model stack.
| Build capability | What customers actually do with it |
|---|---|
| Ubuntu 24.04 LTS — full Linux platform | Root access. Standard apt / systemd / cron / journald — everything your DevOps already knows. Pre-imaged but never sealed. |
| cron / systemd timer workflows | Schedule nightly ETL into the Article 12 log · 06:00 dispatch-readiness report · weekly Nemotron RAG re-index of your SOPs · monthly CSRD evidence bundle export. Workflow scheduling is built-in. |
| Docker / Docker Compose + OpenSandbox | Drop in your own containers next to the NVIDIA NIMs — internal microservices, custom ETL, in-house ML models. OpenSandbox keeps them isolated from each other and from the NIM stack. |
| Local MCP server | Wire Claude Code / your IDE / your dispatcher app to the box over your internal network. No internet hop. Customer-supplied MCP tools mount alongside ours. |
| Pre-installed toolchain | PyTorch, Jupyter, Ollama, vLLM, TensorRT-LLM, NVIDIA NIMs — ready out of the box. Bring your own libraries via apt / pip / conda. |
| Article 12 log API | HTTP API for read-only access to the immutable log. Wire it to your SIEM, your CSRD platform, your inspector portal. The log is yours; the API is documented. |
| REST / webhook gateway | Expose any agent or model as an internal HTTP endpoint for your ERP, dispatcher, ticketing system — sovereign, on-prem, no third-party API call. |
| Bring-your-own model | Drop a custom fine-tune (Llama / Mistral / Qwen variants) into the NVIDIA NIM runtime alongside the curated bundle. The box doesn't care whether the weights came from NVIDIA or you. |
| Air-gap switchable | Firewall policy can cut external network at any time. Update tunnel opens only when you allow it; otherwise the box runs fully offline for as long as you want. |
Six layers. Each is hard. The combination is harder. This is what the customer is really buying — and why a cloud LLM vendor can't compete on the same axis.
| Moat layer | Why it's hard to copy |
|---|---|
| EU-sovereign jurisdiction | Lifetime Oy is Finnish-registered (Y-tunnus 0772407-9), EU-based. A US or non-EU vendor cannot credibly sell "EU sovereign" to a defence ministry, a NIS2 operator, or a CER critical entity. Jurisdiction can't be cloned. |
| Customer-owned hardware | The customer owns the box. We don't host anything. CLOUD Act exposure and Schrems II concerns die in one move. A SaaS vendor can't replicate this without reinventing themselves as a hardware vendor — and most won't. |
| Bridged inference architecture | The GN100 ConnectX-7 bridge for 405 B-param models, TensorRT-LLM tensor parallel, and Article 12 log replication across nodes — that's months of engineering against a specific hardware topology. Cloud vendors don't optimise for "your customer's two-node cluster". |
| Auditable handoff | Signed install protocol + Article 12 cryptographic log + operator drill = inspector-friendly out of the box. Most AI vendors leave compliance evidence as the customer's homework. We make it the install deliverable. |
| Service wrapper depth | Operator training + 24/7 named-engineer support + warranty servicing + remote configuration — a Finnish team that answers the phone. Copy that without a Finnish team, an EU office, and a regulated-industry customer base. |
| Front-line feedback loop | VPK / industry-specific operator notes → next month's NVIDIA NIM bundle. Field experience in EU-regulated operations compounds month over month. New entrants start at month zero. |
The pattern repeats across every EU-regulated industry. Sovereign on-site NVIDIA stack + service wrapper. Different bundle per industry, same moat.
When sovereign AI matters and the network doesn't always cooperate. Whether you're running the operations bridge of a vapaapalokunta (VPK) on a Saturday alarm, coordinating a CSRD audit week, or just keeping your customer data physically inside your own walls — Control Room gives you the brain on the desk, not in the cloud.
Two line items make up the €5,294 turnkey price. Both ship together; you can also buy the install separately if you already own NVIDIA-capable hardware.
€3,999 · NVIDIA GB10 Grace Blackwell SoC · 20 cores · 120 GiB unified memory · Ubuntu 24.04 LTS pre-imaged · 230 V EU plug.
The hardware substrate. EU air-gapped, customer-owned, zero CLOUD Act exposure.
€995 · On-site engineer · CUDA + NVIDIA drivers validated · NVIDIA NIM microservices · model bundle of your choice loaded and sanity-tested · MCP endpoint live · Article 12 log started.
The service that turns the box into something your operator can actually use on day one. Included here; also sold standalone for adding nodes later.
The Acer Veriton GN100 has the headroom to run a quantized 72B-class language model alongside specialist models for voice, vision, and embeddings. Bundle picks are tuned to the operational reality of a control room.
| Model | Role in the control room | Footprint (quantized) |
|---|---|---|
| Qwen 2.5 72B (QLoRA) | Primary reasoning model — incident summaries, regulatory Q&A, multi-step decisions. Domain-fine-tuned on Finnish + EU regulatory corpus. | ~40 GiB RAM |
| Llama 3.1 8B Instruct | Fast triage and structured-output tasks (form-filling, JSON extraction, classification). | ~6 GiB RAM |
| Whisper-large-v3 (FI/SV) | Voice → text for radio chatter, dispatcher calls, debriefs. Finnish & Swedish first-class. | ~3 GiB RAM |
| BGE-M3 embeddings | Local RAG over your SOPs, runbooks, regulatory PDFs — no document ever leaves the building. | ~1 GiB RAM |
| Optional: YOLOv8-s (vision) | Light vision model for camera feeds (e.g. VPK garage door, perimeter, weather cam). | ~0.5 GiB RAM |
Here's how Control Room covers it without leaving the building:
One device, one install fee — but you choose which models we tune for. Pick at order time, switch later (subscription required).
Qwen 72B + Whisper FI/SV + BGE-M3 + YOLOv8-s. Built for dispatch, debrief, perimeter camera, and incident reporting in Finnish.
Qwen 72B regulatory-tuned + DocStruct (PDF parsing) + EUR-Lex retrieval. Built for CSRD, CBAM, EU AI Act Article 12 evidence work.
Qwen 72B + SiteSense vision (PPE, crane envelope) + schedule reasoning + BIM-aware embeddings. Built for site office command.
Tell us which models matter to your operation; we scope and bundle during onboarding. Common picks: legal, healthcare triage, energy ops.
Control Room is end-to-end NVIDIA on the inference side. We install and tune the full sovereign stack on-site so you don't have to:
Lifetime Oy is a NVIDIA Inception applicant — sovereign EU edge AI track. Partner enquiries: risto@onelifetime.world.
You pay €5,294 once for the box and the install. Then €299/mo keeps NVIDIA's proprietary model stack rising on the same box — without you changing the hardware.
| NVIDIA model / component shipped each month | Why your token economics rise on the same GN100 |
|---|---|
| Llama Nemotron NIMs Nano (edge / real-time), Super (single-GPU high-accuracy), Ultra (data-center-grade) — packaged, versioned, sovereign | These are NVIDIA's flagship agentic-reasoning open NIMs. Customers can't pip-install them with EU sovereignty intact. We ship them onto your box, refreshed monthly. |
| Cosmos world foundation models Cosmos Reason 2 (spatiotemporal VLM, chain-of-thought), Cosmos Predict 2.5 (future-state generation), Cosmos Transfer 2.5 (photorealistic physics simulation) | The basis for spatial-temporal reasoning. New Cosmos checkpoints land here sovereign the moment they're released on NGC. Cosmos 3 is on the roadmap. |
| Cosmos Nemotron VLMs & Nemotron Speech Vision-language NIMs + leaderboard-topping ASR for FI/SV/EN dispatch | Dispatch transcription, image understanding, multimodal Q&A — all NVIDIA NIMs, all local, all refreshed monthly. |
| Nemotron RAG & Safety NIMs Multimodal retrieval-augmented generation + safety guardrails | Local RAG over your SOPs and runbooks + safety filtering — NVIDIA-curated, sovereign, never queries an external API. |
| NVIDIA Isaac ROS & CV-CUDA Multi-sensor calibration, fusion pipeline, accelerated vision kernels | For control-room camera and sensor pipelines — the NVIDIA-optimised path is faster and cheaper-per-frame than anything we could write ourselves. |
| TensorRT-LLM kernels for GB10 FP4/FP8 paths on Grace Blackwell, speculative decoding, optimised attention | Tokens/sec on the same GN100 trend up every release — sometimes 2–3× on a single jump when a new kernel lands. |
| NVIDIA-tuned quantization for GB10 AWQ / GGUF Q4-K-M weights tuned for the exact Grace Blackwell pair | NVIDIA + Lifetime co-tune for the GB10. The same 70B-class Nemotron runs in less RAM, leaving headroom for a second model alongside. |
| Front-line feedback loop Operator notes from your dispatch desk → next NIM bundle | Your VPK dispatch notes / CSRD audit pain → fed back into the next month's NVIDIA NIM bundle for your operation. The system adapts to you. |
This is also why we sell the box once and the NVIDIA model subscription forever. Hardware depreciates. NVIDIA's NIM catalogue keeps growing. Your token economics keep rising. The customer who renews compounds the gain.
| Cadence | What lands | Why it's not optional |
|---|---|---|
| Monthly | New NIM versions (Llama Nemotron releases, Cosmos checkpoints, Nemotron Speech / RAG / Safety updates), new model bundles, TensorRT-LLM kernel refresh, quantization updates | This is the rising-token-economics curve. Skip a month, your competitor's box is a month faster than yours. |
| Weekly | Security backports (CUDA driver patches, container CVE fixes, OpenSandbox + Ubuntu LTS updates), minor model patches, bug fixes | Edge AI boxes in regulated industry (NIS2 / CER / EU AI Act) need a documented patch cadence. Weekly = inspector-friendly. |
| Daily hotfixes (when needed) | Zero-day CVE patches, NVIDIA driver hotfixes for critical issues, model-side jailbreak / prompt-injection fixes | When NVIDIA ships an emergency CUDA / driver fix, you can't wait for next month's release. We push it within 24 h over your secure tunnel. |
All updates ship over your secure tunnel — sovereign always. Nothing about the update process needs your data to leave the building.
You're not buying a beige box. You're buying a relationship with a Finnish team that keeps the brain alive in your building. Every Control Room ships with full service bundled into the install + subscription:
| Service | What you get | Where it lives |
|---|---|---|
| Operator training | 30-minute install drill, scenario rehearsal (VPK dispatch / CSRD audit / your runbook), remote refresher sessions every quarter, runbook handover. Your operator signs off before we leave. | Included in install (€995) |
| 24/7 technical support | Named Lifetime contact for production incidents. SLAs sized to your operation. No ticketing-system black hole — a Finnish engineer answers. | Sovereign Model Updates Subscription (€299/mo) |
| Warranty servicing | We handle the Acer Veriton GN100 warranty for you — diagnostics, RMA paperwork, swap-out coordination, redeploy. You don't talk to a vendor support queue. | Sovereign Model Updates Subscription (€299/mo) |
| Remote configuration | Model bundle changes, security patches, OS updates, MCP endpoint tweaks — all pushed over your secure tunnel. Sovereign always: nothing leaves your building unless you decide so. | Sovereign Model Updates Subscription (€299/mo) |
If you cancel the subscription, the box still works — your data and audit log stay yours. You just stop getting LLM updates, support, warranty handling, and remote config. The relationship is opt-in, every month.
Verified against Acer's official launch (Oct 2025) and current 2026 stock. Customer-supplied NVIDIA-capable hardware is also supported — see NVIDIA Model Install.
| Specification | Value |
|---|---|
| NVIDIA chip | NVIDIA GB10 Grace Blackwell Superchip — 5th-gen Tensor Cores, next-gen CUDA cores |
| AI performance | Up to 1 PFLOPS FP4 AI performance |
| CPU | 20-core Arm: 10× Cortex-X925 @ 4 GHz + 10× Cortex-A725 @ 2.8 GHz |
| Unified memory | 128 GB LPDDR5x · 273 GB/s bandwidth (CPU+GPU share the same pool) |
| Storage | Up to 4× 4 TB self-encrypting M.2 NVMe SSD — fast model loading, on-device Article 12 log |
| Network | 1× RJ-45 LAN + NVIDIA ConnectX-7 NIC (lets two GN100s be linked for 405 B-param models — that's the Situation Room SKU) |
| Wireless | Wi-Fi 7 · Bluetooth 5.1 |
| I/O | 4× USB 3.2 Type-C · 1× HDMI 2.1b · Kensington lock |
| Form factor | 150 × 150 × 50.5 mm · < 1.5 kg — fits a kalustohalli cabinet or any control room shelf |
| OS / software | NVIDIA AI software stack preloaded · we re-image with Ubuntu 24.04 LTS + DWS IQ 6 at install · supports PyTorch, Jupyter, Ollama, vLLM, TensorRT-LLM out of the box |
| Power | Standard EU 230 V — efficient compared to a rack server, site-deployable |
| Hardware price | €3,999 (Acer launch price, EMEA) — independent of our €995 install service |
The GN100 is our default — but if your workload is different, we can deliver the install service on these alternatives too:
| Requirement | How Control Room solves it |
|---|---|
| Article 12 — Record-keeping | All AI events written to on-device NVMe. Never leave premises. |
| GDPR Article 5(f) | Customer/dispatch data processed in customer-owned hardware. Zero third-party processor. |
| NIS2 — Critical infrastructure | Air-gap switchable. Network-segmented. Incident logs stay on premise. |
| CLOUD Act immunity | Hardware owned by customer in EU jurisdiction. US law cannot compel access. |
| Defense / VPK / healthcare | Sensitive use cases that public cloud can't host — operational because nothing transits external networks. |
Hardware sourced and configured by Lifetime Oy. Delivery lead time: 2–4 weeks within EU. Need redundancy? See the Situation Room (2× bridged). Volume pricing for 3+ nodes: risto@onelifetime.world.