AI-powered multi-agent system that analyzes climate, infrastructure, risk, policy & sustainability to identify the world's best data center locations.
Inspired by the global race to build AI infrastructure — from Google's data centers in India to hyperscale deployments across Southeast Asia and Europe — this tool gives founders, infrastructure teams, and researchers a structured AI engine to evaluate and compare data center sites worldwide.
You input a country (or multiple countries to compare), and 7 specialized AI agents run a deep analysis across every dimension that matters for data center siting — from natural cooling potential and renewable energy availability to seismic risk, government incentives, and carbon neutrality timelines.
The result: a ranked, scored report with an interactive map, detailed site breakdowns, and a strategic recommendation — all generated in under 60 seconds.
Single country mode → AI discovers top 4 regions, scores them, maps them.
Compare mode → Pit India vs Germany vs Singapore and get a global ranking.
| Agent | Role |
|---|---|
| 🌍 Region Discovery | Identifies the top 4 strategic regions within a country based on tech infrastructure, climate zones, and connectivity |
| 🌡️ Climate Analysis | Evaluates natural cooling potential, avg temperature, humidity, PUE estimate, and renewable energy (solar/wind/hydro) |
| Scores seismic, flood, cyclone, extreme heat, water scarcity, and political stability risks with mitigations | |
| ⚡ Infrastructure | Analyzes power grid reliability, fiber connectivity, land cost, workforce, IXP proximity, and latency to major hubs |
| 📜 Policy & Incentives | Researches government initiatives, tax incentives, data localization laws, and foreign investment openness |
| 🌱 Sustainability | Computes carbon footprint, ESG potential, water efficiency strategy, carbon neutrality timeline, and green certifications |
| 🏆 Final Ranking | Synthesizes all agent outputs into a strategic recommendation with ranked sites and investment guidance |
Suitability Score =
Climate Score × 0.25
+ (100 − Risk Score) × 0.25
+ Infrastructure Score × 0.25
+ Policy Score × 0.15
+ Sustainability Score × 0.10
| Grade | Range | Label |
|---|---|---|
| S | 85–100 | World-class — Pursue immediately |
| A | 70–84 | Excellent — Strong investment case |
| B | 55–69 | Good — Manageable trade-offs |
| C | 40–54 | Viable — Significant improvements needed |
| D | 0–39 | Not recommended |
- 🌐 Single Country — Enter any country, AI discovers and scores the top 4 best regions automatically
- 📍 Country + Region — Specify your own region of interest alongside AI-discovered alternatives
- ⚖️ Compare Countries — Enter 2–3 countries, AI picks the top 2 regions per country and ranks them globally
git clone https://github.com/yourusername/datacenter-intelligence.git
cd datacenter-intelligencepip install -r requirements.txtSign up at https://console.groq.com — no credit card required.
Open utils/groq_client.py and paste your key:
GROQ_API_KEY = "gsk_your_key_here"python app.pyhttp://localhost:5000
datacenter-intelligence/
│
├── app.py ← Flask app + route orchestration
│
├── agents/
│ └── agents.py ← All 7 AI agents + orchestration logic
│
├── scoring/
│ └── scoring_engine.py ← Weighted suitability score formula
│
├── utils/
│ ├── groq_client.py ← Groq API wrapper with JSON retry logic
│ └── prompt_templates.py ← Structured prompts for each agent
│
├── templates/
│ ├── index.html ← Input form (single / compare modes)
│ └── result.html ← Results: map + ranked cards + reports
│
└── requirements.txt
| Layer | Technology |
|---|---|
| Backend | Python + Flask |
| LLM | Groq API (Llama 3.3 70B) — free |
| Frontend | HTML + CSS + Vanilla JS (Jinja2 templates) |
| Map | Leaflet.js with CartoDB dark tiles |
| Styling | Custom dark UI, IBM Plex Mono + Space Grotesk |
India→ Discovers Rajasthan (solar), Maharashtra (infra), Andhra Pradesh (coast), Himachal Pradesh (cooling)Germany→ Discovers Frankfurt, Berlin, Munich, Hamburg with EU policy analysisCompare: India, Singapore, Germany→ Global cross-country ranking with head-to-head scores
The app also exposes a REST endpoint for programmatic access:
curl -X POST http://localhost:5000/api/analyze \
-H "Content-Type: application/json" \
-d '{
"country": "India",
"datacenter_type": "hyperscale",
"scale": "large-scale"
}'- Each analysis runs 7 sequential Groq API calls — expect ~45–90 seconds total
- All agents enforce strict JSON output with automatic retry on parse failure
- Model:
llama-3.3-70b-versatilevia Groq (completely free tier) - Latency and coordinate data is AI-estimated — validate before production use
- Real-time climate data integration (Open-Meteo API)
- Power grid data from national energy APIs
- PDF report export
- Historical site comparison saved to local DB
- Docker Compose deployment
MIT License — free to use, modify, and deploy.
Built with ⚡ Groq · 🌍 Leaflet · 🐍 Flask · 🤖 Llama 3.3