08 / CONFIGURATOR Back to home
How many Planets, how much intelligence.
Choose energy and site: we calculate how much AI runs day and night, with which models and for how many users. Indicative values, then we verify them with a feasibility study.
02 / RESULTS
What your site can run.
Continuous AI load12.7 kW
PV required195.3 kWp
Storage installed261 kWh
GPU memory768 GB
24/7 solar even in the worst month
Energy
| Constant site load | 15.2 kW |
| Energy per day | 391 kWh |
| PV for worst month | 195.3 kWp |
| Max connectable PV | 250 kWp |
| Ground surface | 1,074 m² |
| Solar coverage, worst month | 100 % |
| Extra daytime power (batch) | 71.6 kW |
| Energy in a year | 142.6 MWh |
Compute
| HGX B300 servers (8 GPU) | 0 |
| 8× RTX PRO 6000 servers | 1 |
| 4× RTX PRO 6000 servers | 0 |
| GPU memory | 768 GB |
| Concurrent users | 40–80 |
| Nominal users | 400–800 |
| Closest Barrel size | D1 |
Recommended models
- DeepSeek V4-Flash (MIT) or GLM-5.2 in FP4
- Specialists: MedGemma, Parakeet/Whisper for voice, PaddleOCR-VL for documents, Qwen3-Embedding for search
Order of magnitude
€ 320,000 – € 510,000
Order of magnitude of supply, VAT and installation excluded. The real price comes out of the feasibility study.
Assumptions: 95% depth of discharge, 90% round-trip efficiency, 0.5 kW auxiliaries per cabinet, PVGIS photovoltaic yields. Users and hardware are estimates from public benchmarks.
03 / GET THE FULL SIZING
Get the full sizing
We will send you the sheet with the numbers of your configuration and a first comment from one of our technicians.