/ › expansion › 05-simulator-agi-horizon › live

Simulator (live) — Anchor 0 Fase 1

Slider-driven sister of /simulador. Where /simulador selects from curated presets, this surface exposes the six highest-leverage knobs (PRCC ranked) and rebuilds the KPI distribution on every input. Pipe the resulting parameter set straight into a one-page recipient-tailored brief at /brief. Same Anchor 0 Fase 1 financial structure (blueprint/14-paraguay-anchor-0.md §5).

HS.offtake.tariff USD / GPU-hr 1.50

REG.friction_multiplier ×schedule 1.20

CX.HPC.greenfield USD M / MW 8.0

FM.private_match.realization share 0.75

GATE.anexoC.signing_per_Q prob/Q 0.060

FM.cost_of_debt.bond_A % nominal 7.0%

Generate decision brief → Switch to presets (/simulador) →

Live KPIs (1,000 runs, stratified G/M/B)

DSCR (year-5)

— · — · —

P10 · P50 · P90 · target ≥1.5

NPV (USD M @ 8% real)

— · — · —

target P50 > 0

Financing gap (USD M)

— · — · —

target P50 → 0

P(KPI 2032 met)

—

DSCR ≥ 1.5 ∧ NPV > 0 ∧ gates clear

capex_total (USD M)

— · — · —

chap 14 §5.1 design 2,253–2,995

P(Anexo C signed by COD)

—

21 quarters × per-Q signing prob

Distribution charts

NPV distribution (USD M @ 8% real)

DSCR distribution (year-5 ratio)

Capital stack composition

P(Anexo C signed by quarter)

NPV: amber < 0, green > 0; vertical line at 0. DSCR: amber < 1.5× (sub-bankable), green ≥ 1.5×. Cap-stack shows manifest centroid (pre-realization). Anexo C curve is cumulative probability over 21 quarters at the slider's per-Q rate.

Sensitivity tornado — what shifts NPV most

Bar = (P50 NPV when knob at P90 input) − (P50 NPV when knob at P10 input). Six knobs swept ± across their range. Sign = direction.

Anchor 0 cash-flow shape (year-by-year, P50)

Construction-year capex outflows (amber, 2028–2030) → COD 2031 → revenue ramp (green) − opex (gray) per year. P50 trajectory at current slider settings.

Provenance

engine js-port v0.1 of simulator.py

scope anchor-0-fase-1

manifest hash —

backing python run db114bb35ecd4d4c… (10K runs, byte-identical when knobs match defaults)

tags VERIFIED=26 · ASSUMPTION=2

Method (click to expand)

Same financial logic as tools/simulator/simulator.py ported to vanilla JS: capex = (HPC + subestation + cable + cooling + fixed_other) × (1 + 0.5 × (friction − 1)); revenue per MW per year = utilization × 8,760 × tariff × 250 / 10⁶; opex = energy MWh × energy cost + security + opex_pct × capex; DSCR = (revenue − opex) / debt_service at year 5; NPV at 8% real over 2026–2036, capex spread evenly across construction years.

Six sliders override the manifest's mode for that parameter. The remaining 22 parameters sample from manifest v0 ranges (triangular / beta / lognormal). Scenario stratification (G 0.31 / M 0.45 / B 0.24) taken from the multi-source posterior in chapter 06; B-conditional capex mode is 9.5 USD M / MW. 1,000 samples per render — fast enough for ~30 fps slider drag on a laptop.

Anexo C signing is treated as a per-quarter Bernoulli over 21 quarters (2026-Q1 → 2031-Q1); energy cost is the probability-weighted mix of ANDE retail (USD 38–66 / MWh) and PPA directo (USD 18–34 / MWh). The displayed P(Anexo C by COD) is the simulator's posterior, not a forecast.

For the full 28-knob taxonomy, correlation blocks, and engine spec see state/expansion/05-simulator-agi-horizon/01-parameter-taxonomy.md and 02-engine-spec.md. For the back-office 10K Monte Carlo + Sobol indices use the Python engine.