SPATIAL AI ROUTING

The spatial router
for AI compute.

One API. Multiple providers. Real-time energy and carbon data.
Your workloads run where the grid is cleanest.

Inference API
OpenAI-compatible chat completions. Drop in your base URL. We route each request to the lowest-carbon provider available for that model, and return the carbon data with every response.
GPU Jobs
Fine-tuning, batch inference, evals — placed on the lowest-carbon datacenter. An H100 in Quebec emits 3g CO₂/kWh. The same H100 in Virginia emits 430g.
Get API Key Try Playground — No Signup
The energy arbitrage

50 TWh of clean energy is wasted every year because the grid can't absorb it.

Renewable energy is curtailed when supply exceeds demand — 50 TWh per year, enough to power Norway. AI workloads are flexible. They can run anywhere. Windfall routes them to where the grid is cleanest, right now.

Same GPU, different carbon footprint
An H100 in Iceland runs on geothermal at 15g CO₂/kWh. The same H100 in Virginia runs on gas at 430g. The hardware is identical. The carbon is not. Windfall routes to the lowest-carbon option available for your model.
6 million GPUs could run on wind power alone
Research shows existing wind farm infrastructure could power 6M+ GPUs — the capacity is already there, just underused. One API call, routed to the cleanest available grid.
The infrastructure already exists
Bitcoin miners spent billions building datacenters next to cheap renewable energy. Now they're converting to AI compute. Windfall routes your workloads to that capacity.
How it works

One endpoint. We handle the routing.

1
Swap your base URL
OpenAI-compatible API. Works with every SDK. Add your Windfall key, keep your existing code. For GPU jobs, specify the model and we handle placement.
2
Windfall checks grid zones in real time
Real-time wholesale energy prices and carbon intensity across multiple grid zones. We score every available provider and pick the cheapest clean option.
3
You get the result + a carbon receipt
Every response includes: provider identity, grid zone, carbon intensity, energy price, and methodology. Audit-ready data for anyone who needs it.
Who uses Windfall

Built for teams that run AI and care what it costs.

ML teams running GPU jobs
Fine-tuning, RLHF, batch inference, evals. Multi-hour jobs where the carbon delta compounds. Same H100, but placed in Quebec (3g CO2) instead of Virginia (424g).
Enterprises & institutions
Scope 2 and Scope 3 carbon reporting per request. CSRD-ready exports, SCI for AI scoring (ISO 21031), EU data residency options. Every request returns provider, grid zone, and carbon intensity — audit-ready for ESG and compliance teams.
Onchain agents
Pay per request with USDC, ETH, or x402. Configure routing preferences via API — set max carbon intensity, preferred zones, and mode. Your agent gets green inference without human intervention.
Integrate

One extra field. Full cost and carbon visibility.

curl
# Route inference through the cheapest clean provider curl -X POST https://windfallrouter.xyz/v1/chat/completions \ -H "Authorization: Bearer $WINDFALL_KEY" \ -d '{ "model": "deepseek-v3", "messages": [{"role": "user", "content": "Hello"}], "mode": "greenest" }'
Every response includes
{ "choices": [...], "windfall": { "verifiedProvider": "Fireworks.ai", "carbonIntensityGCO2": 200, "carbonDelta": +220, // grams saved vs US grid average "energyPricePerKwh": 0.0189, "costBreakdown": { "providerCostUsd": 0.00028, "marginPercent": 15 } } }