One question, start to finish.
Ask what if; Henselian runs the model on every record and every path; you read what then.
- What if
What if the wind drops for a week?
- Run
Power book and hedge: a year day by day
A wind PPA, fixed-price supply and a Cal forward, every delivery day simulated with persistent wind and its effect on price.
- What then
Expected shortfall with and without the Cal base hedge: whether the hedge earns its cost.
Start from a model that already asks.
Each is a working model in the app's Model builder, written in Hensel and run on the GPU. They start from a synthetic book; your data replaces it, record by record.
2 models in the Model builder
Trading and origination1
Power book and hedge: a year day by day
A wind PPA, fixed-price supply and a Cal forward, each delivery day simulated with persistent wind and its price effect: expected shortfall with and without a Cal base hedge.
Risk1
Options at risk over the month: full revaluation
A cap, a sold PPA floor and a Cal forward, revalued on every path a month out by their own simulated delivery years: the 1-in-100 loss, against the delta shortcut.
Fast enough to ask again.
for 1.5 trillion member-paths: a 2.5-million-member UK buy-in, end to end on one graphics card
faster than hand-optimised C on all 32 threads of the same machine
engine, one language, one app, for insurance, banking and energy alike
These were measured on an insurance run, a UK buy-in. The energy models above run on the same engine and language; a benchmark of their own is next.
One invitation
Come and challenge us.
If you work in insurance, banking, energy, or anywhere uncertainty drives important decisions, bring us the simulation problem your organisation wishes it could solve faster. Let's find out what's possible.