ICHOR · OS
ICHOR / Domains / Critical-minerals royalty
D-06 · Critical-minerals royalty Live
Crucible

Show up already knowing your number.

Crucible is the ICHOR engine for critical-minerals royalty. It runs the deal desk: candidates ranked to a hard mandate, a written call held on every metal, and any royalty valued to NPV in the time it takes to read the deck. It was validated out-of-sample against a specialist's read before any claim was made for it.

The decision problem

Royalty and streaming capital lives or dies on two edges: seeing the right deals before they are crowded, and underwriting them faster and more honestly than the feasibility deck the promoter hands you. Both require holding an entire junior-mining universe in view, reading the NI 43-101 and S-K 1300 technical reports behind each name, and being able to value a royalty stream, whether a net smelter return, a gross royalty, or a stream, on the timeline of a live negotiation.

Done by hand, that is weeks of a specialist's time per candidate, and a directional view on each metal only as good as the last deck you read. Most desks end up reacting to what is brought to them, at the promoter's valuation, on the promoter's clock.

And if you sit on the other side of the table, the same engine tells you what the stream you are about to sell is worth before the buyer does.

Watch it run

A valuation engine at work: option pricing on the position, a royalty DCF that moves as the commodity scenario shifts from bear to base to bull, and a sponsor-credit model across macro regimes. A capture from the live engine, with the specific deal held back.

Crucible · royalty valuation enginerecorded capture · 2026
The counterparty, the specific asset, and the commodity benchmark are redacted. The option pricing, royalty DCF, and regime-credit math are shown as they run.

The method, applied to royalties

The same five-step method behind every ICHOR system, run into the ground truth of mining finance. See the method in full →

01 · Source

Assemble the candidate universe from the authoritative record, the SEDAR+ and EDGAR filings and the technical reports behind each name, and rank it against a hard investment mandate before the deal is shopped.

02 · Resolve

Reconcile each candidate's deck, terms, and commodity assumptions into one clean picture, with the numbers that do not agree flagged and sourced.

03 · Decide

Hold a written call on each metal, set against the price deck the feasibility studies assume, with the divergence stated where there is one, and mandate-rank the pipeline so capital goes to the highest-conviction fits.

04 · Act

Underwrite any royalty to NPV: the per-share impact, and the commodity price the seller's ask implies, so you know what you would be paying for the deck's optimism before you are at the table.

05 · Prove

Every figure traces to standard mining-finance formulas, auditable against your own model. The valuation is fast because it is disciplined, not because it cuts corners.

The out-of-sample test

Before any claim was made for the underwrite, it was put through a pre-registered program built to be falsifiable. The hypothesis and the scoring were locked before the first evaluation. It was then run on transactions that closed after the model's knowledge cutoff, so no capable model could have memorized the outcome, and scored against a specialist's independent read of the same deals.

12
transactions, all closed after the model's knowledge cutoff, priced cheap to expensive
12 of 12
matched or exceeded the specialist's read on risk-adjusted return against a hurdle
Locked
hypothesis and scoring fixed before the first deal was seen
Pre-registeredPost-cutoffLeakage-controlled, 0 breachesCalibrated firstRisk-adjusted vs hurdle

The figure, not the model

This is the part a mining committee cares about most, and the part most AI vendors cannot answer.

No language model computes any valuation.Every valuation engine is deterministic. Identical inputs return identical figures, every run records a hash of its code and its inputs, and any number in a memo reproduces on demand.
The goal was parity, not a beat.The engine was scored against a specialist's independent underwrite and required to match it. Beating a good underwriter can only be proven forward, with capital and time, so we do not claim it.
Every figure traces to a primary filing.A net asset value that cannot be traced back to a filing is not one a committee should have to rely on.
The full discipline: read the method note →

By the numbers

It runs on the primary filings across the whole sector, at a scale and speed a deal team cannot match by hand. Every figure it returns traces to its source filing and reproduces on demand.

$150M+
in candidate deal flow, sourced and ranked from the filings, and since closed
400+
operators monitored across the junior-mining universe
11
of the largest public royalty and streaming companies' filings, polled daily
40+
precedent deals in the comp set, every cell source-cited
Minutes
to read a several-hundred-page technical report into the structured inputs a NAV is built on
~30 min
for a full independent underwrite of a candidate

The desk that reads the sector every morning

Crucible values a deal and keeps reading. Before your desk opens, it has already gone through every new royalty and streaming transaction filed across the peer set, pulled the terms and the multiples, and refreshed the comparable set the next underwrite draws on. The picture is current when you arrive, not a week behind.

peer filings polled11 firms · daily
new transactions readevery royalty / stream filed
terms & multiples extractedP/NAV · GEO-normalized · cited
comparable setrefreshed automatically
desk briefone page, sourced, before open
Illustrative of the daily run. Specific transactions are not shown.

See it run on your mandate.

ICHOR deploys one industry at a time, forward-deployed and tied to outcomes. Every inquiry is read personally.

Request access