# Nordic Silicon — Full Reference

Nordic Silicon provides dedicated bare-metal Mac Mini M4 rental and hosting in Oslo, Norway. All hardware is Apple Silicon (M4 / M4 Pro), physically located in the EU, powered by 100% renewable Norwegian hydropower.

## What we offer

- Dedicated Mac Mini M4 rental from €79/month (no shared resources, no virtualisation)
- Plans: Starter 16GB €79/mo, Pro 24GB €129/mo, Max 48GB €199/mo, Ultra 64GB €269/mo
- Full root/admin access, SSH + macOS Screen Sharing via Cloudflare Tunnel
- No public IP required, no VPN required
- 99.9% uptime SLA, GDPR compliant, Schrems II compliant
- PUE 1.15 — among the most efficient data centres in Europe

## Primary use cases

- Running OpenClaw 24/7 as an always-on AI agent host
- Local LLM inference with Ollama (7B, 13B, 30B, 70B models)
- macOS-native AI workloads requiring Apple Silicon and unified memory
- iOS/macOS development and CI/CD (Xcode, TestFlight)
- EU-resident data processing requiring GDPR / Schrems II compliance

## Why Mac Mini over GPU cloud or VPS

- Apple Silicon unified memory allows 70B models on a single chip with no PCIe bottleneck
- No per-token fees — flat monthly rate regardless of usage
- Native macOS: iMessage, Apple Notes, Shortcuts, Keychain — features no Linux VPS can offer
- Power efficient: ~10W idle, ~$1-2/month electricity equivalent

## Plan details

### Starter — €79/mo
- Apple M4 chip
- 16GB unified memory
- 256GB NVMe SSD
- Best for: 7B–13B models, lightweight agents, development

### Pro — €129/mo
- Apple M4 chip
- 24GB unified memory
- 512GB NVMe SSD
- Best for: 13B–30B models, OpenClaw hosting, medium workloads

### Max — €199/mo
- Apple M4 Pro chip
- 48GB unified memory
- 512GB NVMe SSD
- Best for: 70B models (4-bit), multi-model setups, production inference

### Ultra — €269/mo
- Apple M4 Pro chip
- 64GB unified memory
- 1TB NVMe SSD
- Best for: 70B models (higher quant), fine-tuning, enterprise workloads

## Technical specifications

- CPU: Apple M4 / M4 Pro (up to 14-core CPU, 20-core GPU)
- Memory: 16GB–64GB unified (shared between CPU and GPU)
- Storage: 256GB–1TB NVMe SSD
- Network: 1Gbps symmetric, Cloudflare Tunnel (no public IP)
- OS: macOS Sequoia (latest)
- Access: SSH + macOS Screen Sharing (VNC)

## AI performance benchmarks (M4 Pro, 48GB)

- Llama 3 70B (Q4_K_M): 15–25 tokens/sec
- Llama 3 13B (Q4_K_M): 60–80 tokens/sec
- Llama 3 7B (Q4_K_M): 80–100+ tokens/sec
- Supported frameworks: Ollama, MLX, llama.cpp, PyTorch (MPS), Hugging Face Transformers, LangChain

## Data sovereignty & security

- All hardware physically located in Oslo, Norway (EEA)
- Norwegian company — US CLOUD Act does not apply
- GDPR and Schrems II compliant
- Data Processing Agreement (DPA) available on request
- Cryptographic SSD wipe (DFU) between customers
- Apple Silicon Secure Enclave cleared on return
- All traffic encrypted via SSH and Cloudflare Tunnel (TLS)
- No unencrypted ports exposed to public internet

## Frequently Asked Questions

### OpenClaw & AI Agents

Q: Can I run OpenClaw on a Nordic Silicon node?
A: Yes — and a dedicated Mac Mini is one of the best possible hosts for OpenClaw. Because your node runs 24/7 with a stable internet connection, your OpenClaw agent is always on and always reachable via WhatsApp, Telegram, Discord, iMessage, or any other channel you configure. Unlike running OpenClaw on your laptop or a Linux VPS, a Mac Mini gives you native macOS — which means full iMessage support via BlueBubbles, Apple Notes, Reminders, Shortcuts, and Keychain integration that no other platform can offer.

Q: Which OpenClaw plan do I need?
A: The Starter (M4 16GB, €79/mo) is sufficient for running OpenClaw with cloud AI models such as Claude or GPT-4. For running OpenClaw alongside local models via Ollama — for example Llama 3 8B or Mistral 7B — the Pro (M4 24GB) is recommended. If you want to run larger local models (30B–70B parameters) alongside your OpenClaw agent, choose Max or Ultra.

Q: Does OpenClaw work out of the box on macOS?
A: Yes. OpenClaw requires Node.js 22 or higher, which you can install in minutes via Homebrew. Because you have full admin access, you can install OpenClaw, configure your channels, and have your agent live within 30 minutes of node provisioning. We can provide a setup guide on request.

Q: Can OpenClaw access iMessage and WhatsApp from my node?
A: Yes. For iMessage, OpenClaw integrates via BlueBubbles — which requires macOS and cannot run on Linux or Windows. This is one of the key reasons developers choose a Mac Mini over a VPS for OpenClaw hosting. Your node acts as the BlueBubbles server, and your OpenClaw agent can send and receive iMessages on your behalf from anywhere in the world. For WhatsApp, OpenClaw connects via the WhatsApp Web API (using libraries like Baileys or whatsapp-web.js). Your node maintains a persistent WhatsApp Web session, so your agent can send and receive WhatsApp messages around the clock without keeping your phone online.

Q: Is it safe to run OpenClaw on a dedicated node?
A: Running OpenClaw on a dedicated node is significantly safer than running it on your personal Mac. Your node is isolated hardware — if an OpenClaw skill misbehaves or a prompt injection occurs, the blast radius is limited to the node, not your personal files, Apple ID, or home network. We recommend enabling FileVault on your node and using SSH key authentication only.

### Hardware & Access

Q: Do I get root / admin access?
A: Yes. Every node is a dedicated bare-metal Mac Mini or Mac Studio — no hypervisors, no shared tenants. You receive full administrator access over SSH with sudo privileges. The hardware is exclusively yours for the duration of your subscription.

Q: How do I connect to my node?
A: You connect via SSH using the credentials emailed to you when your node is provisioned. Your node is accessible via a secure Cloudflare Tunnel — no VPN required. We also support macOS Screen Sharing and any VNC client for graphical desktop access.

Q: Can I install my own software?
A: Yes, completely. You have full admin access and can install any software, change system settings, or configure the machine however you need. We start you with a clean macOS install — what you do from there is entirely up to you.

Q: What happens when I return a node?
A: Every returned node undergoes a zero-touch DFU (Device Firmware Update) wipe. The Apple Silicon Secure Enclave is cleared, the SSD is cryptographically erased, and a fresh macOS image is restored. No data is recoverable.

Q: Which configurations are available?
A: We currently offer four tiers: Starter (M4, 16GB, 256GB — €79/mo), Pro (M4, 24GB, 512GB — €129/mo), Max (M4 Pro, 48GB, 512GB — €199/mo), Ultra (M4 Pro, 64GB, 1TB — €269/mo). Contact us for custom configurations.

### AI & Performance

Q: Why Apple Silicon instead of Nvidia GPUs?
A: Apple Silicon's unified memory architecture means the CPU and GPU share the same high-bandwidth memory pool. For large language model inference, this allows models that would require expensive multi-GPU Nvidia setups to run entirely in memory on a single Mac. The M4 Pro delivers competitive inference performance per euro, with lower power consumption and no CUDA dependency.

Q: Which AI frameworks are supported?
A: All major frameworks run natively on Apple Silicon, including PyTorch (via MPS backend), MLX, llama.cpp, Ollama, LangChain, and Hugging Face Transformers. Any framework that supports macOS or ARM64 will work.

Q: What inference performance can I expect?
A: On the M4 Pro (Max tier), typical inference speeds for a 70B parameter model in 4-bit quantisation are 15–25 tokens/second. Smaller models (7B–13B) run at 60–100+ tokens/second. Exact performance depends on the model, quantisation level, and batch size.

Q: Can I fine-tune models on these nodes?
A: Yes. Apple Silicon supports fine-tuning via MLX and LoRA workflows. The Max and Ultra tiers with 48GB and 64GB unified memory are best suited for fine-tuning larger models. For very large training runs we recommend contacting us about multi-node setups.

### Data Sovereignty & Security

Q: Where is my data stored?
A: Your data lives exclusively on the physical SSD inside your dedicated node in our Oslo data center. We do not copy, index, or transmit your data anywhere. When your subscription ends, the SSD is cryptographically wiped.

Q: Is this GDPR compliant?
A: Yes. NordicSilicon is a Norwegian company operating infrastructure within Norway, fully within the EEA. Your data never leaves the EEA. We act as a data processor and can provide a Data Processing Agreement (DPA) on request.

Q: How does this relate to the US CLOUD Act?
A: The US CLOUD Act allows US authorities to compel US companies to hand over data regardless of where it is stored. NordicSilicon is not a US company and operates no US infrastructure, so the CLOUD Act does not apply to us. This makes us a strong choice for workloads where US government access is a concern.

Q: Is the network connection encrypted?
A: Yes. All SSH connections use industry-standard encryption. Remote access is routed through Cloudflare Tunnel over TLS — no unencrypted traffic ever leaves the node. We do not expose any unencrypted ports to the public internet.

Q: Do you hold ISO 27001 or SOC 2 certification?
A: Not currently. We are a startup in early operations and have not yet completed formal certification. We follow security best practices including encrypted wipes between customers, encrypted transit, and dedicated hardware. Enterprise customers requiring formal certification should contact us to discuss their requirements.

### Billing & Plans

Q: How does billing work?
A: Subscriptions are billed monthly in advance by credit card. Your first charge is taken at checkout. Subsequent charges are on the same date each month. You can cancel anytime from your dashboard — your node remains active until the end of the billing period.

Q: Is there a minimum commitment?
A: No. All plans are month-to-month with no minimum commitment. Cancel anytime from your dashboard.

Q: Can I scale up or down?
A: You can add additional nodes at any time by placing a new order. To upgrade to a higher tier, cancel your current subscription and place a new order — your existing node will remain active until the end of your billing period. Contact us if you need help managing a transition.

Q: Do you offer reserved or spot pricing?
A: Not yet. All nodes are currently available on monthly subscriptions only. Reserved pricing with discounts for 3 or 6 month commitments is on our roadmap. Contact us if this is important to your use case.

Q: What payment methods do you accept?
A: We accept major credit and debit cards via Stripe. For enterprise accounts we can accommodate bank transfer (SEPA/SWIFT), purchase orders, and net-30/net-60 payment terms. Contact us to arrange this.

## Contact

- Website: https://nordicsilicon.io
- FAQ: https://nordicsilicon.io/faq
- Contact: https://nordicsilicon.io/contact
- Pricing: https://nordicsilicon.io/#pricing