Training, inference, fine-tuning
AI Compute
GPU-dense capacity for AI training, fine-tuning and inference, engineered around the power source rather than around a grid connection.
What a buyer qualifies us on
Specification
Economics
All-in delivered power cost.
Evidence tier: Modelled
$0.042/kWh
Exusia Digital at Ughelli, on associated gas: fuel, O&M and generation capex recovery included.
As at August 2026. Financial model v3| Source | Range | US$ per kWh | Evidence tier |
|---|---|---|---|
| Exusia Digital, Ughelli | 0.042 | Evidence tier: Modelled | |
| Nigerian grid, industrial | 0.09 to 0.13 | Evidence tier: Illustrative | |
| Diesel generation, Lagos | 0.28 to 0.35 | Evidence tier: Illustrative | |
| US merchant power, ERCOT | 0.03 to 0.05 | Evidence tier: Illustrative |
US dollars per kWh. Comparators are illustrative ranges. We publish the comparison least flattering to us (ERCOT) because an investor will find it anyway.
Next step
Request compute capacity
Tell us what you need and a named engineer replies within 2 business days.