TSXV: UCU · Critical minerals
A rare-earth processor, not a miner: it is building a separation plant in Louisiana, USA, after a demonstration plant in Kingston, Canada. No published plant economics yet.
Expensive: even certain success is worth about 1.05× · low confidence
Changed on 16 Nov: the chance it works is now the analyst's own figure from the deep dive, 25% (before, 30%, the figure the rankings page used). Before: 0.64×, expensive. Now: 0.61×, expensive. Re-run by the analyst and checked by an independent re-run.
Confidence: low. Low. There is no company study; the value if it works rests on one comparable company (Lynas, an integrated miner; enterprise value dated 27 Oct 2026); cash is an estimate until the Q3 statements; the start close could not be re-read on the web. The label itself is robust: certain success is worth only 1.05×, so it stays expensive at any chance; with more late builds (5 in 10 misses) the expected value is 0.66×, on the rankings page's own target 0.65×, and at NdPr (the rare-earth oxide used in magnets) oxide US$180/kg 0.94×.
What one share should be worth in Nov 2030, by the analyst's estimate:
Works means Machine A runs on commercial feed by 31 Dec 2027 and a binding offtake is signed.
Fails means That has not happened by 31 Dec 2027: the plant runs about 18 months late (4 in 10 of these outcomes) or stalls (6 in 10).
Per share, after the new shares each path needs, so each value compares directly with today's price. € at the ECB reference rate of 13 Nov 2026 (1.5758 C$ per €); the rating itself is worked out in Canadian dollars, and future exchange rates will differ.
Expensive at 1.10× or less, cheap at 2.00× or more. Expensive also needs a written reason; cheap also needs a written reason, a dated catalyst and no tripped thesis breaker.
Latest close, Mon 16 Nov 2026: C$2.02 (€1.28), in the expensive range. Same range as the rating.
The new shares and options each path needs are re-priced at each level.
No cheap price. Price alone won't make this share cheap; it would take good news (see “What would change the rating”).
A share counts as cheap here if, on average, it should be worth at least 2.00 times its price by Nov 2030. Worked out on paper, that would happen at C$0.44 a share. But at that price a share would still be worth C$0.52 if the project stalls, because that value is fixed at today's prices (what the market pays today for this company or for similar stalled projects) and does not fall with the share price. The sum would then say you make money even if everything goes wrong, which is not believable: if the share really fell that far, the value if it stalls would fall with it. So the page shows no cheap price instead of a misleading number.
At the rankings' own target: 0.65×. The rankings page's own target for a working Machine A (a range of about 0.25 – 2.65 times today's price, centre 1.18×) gives an expected value of 0.65× with the same value if it fails: still expensive. Shown beside the rating, never counted.
At the top metal price on the rankings: If NdPr oxide (the main rare-earth product used in magnets) were US$180/kg, the top regular price on the rankings page (not a forecast): value if it works 1.72×, expected value 0.94×, still expensive. Shown beside the rating, never counted.
The value at metal futures prices about three years out is pending for all 14 (exchange data on hold).
Expensive. At C$2.18 (close of Fri 13 Nov 2026) the expected value of a share in four years is 0.61 times today's price (about -11% a year), under the 1.1× line below which a share counts as expensive. The expected value weighs what a share is worth if the plan works and if it fails by their chances. The plan: Machine A (600 tonnes a year of rare-earth oxide) runs on commercial feed with a binding offtake (a contract to sell the output) by 31 Dec 2027. Even certain success is worth only 1.05 times today's price, because the market already pays about US$300k for each tonne a year of Machine A's capacity, 69% of what it pays for Lynas's operating, mine-fed output, before the machine is installed or any binding contract is signed. If the plan fails, a share is worth 0.47 times today's price. The analyst's chance that it works (a judgement): 25%.
Price alone won't make this cheap. What could raise the value: factory-acceptance results for Machine A (expected by 31 Dec 2026, marked overdue in the research records); its installation and commissioning (H1 2027); and a first binding feed or offtake contract (required by 31 Dec 2027). A binding feed and offtake contract with committed outside money for Line 1 would let the full 3,600 tonnes a year be valued like an operating producer. Canada's contribution agreement for Kingston has no company date.
Already expensive. It stays so at any price above about C$1.16 with nothing else changed, and it gets worse if Machine A's commissioning is re-dated past 30 Jun 2027, or if equity is raised below C$1.50 a share or cash falls below C$40M before Machine A is commissioned (these are thesis breakers: dated tests, set in advance, that would mean the rating must be redone).
Set in advance: any one of these would mean the plan has failed or the rating must be redone.
| When | What |
|---|---|
| by 30 Nov 2026 | Q3 2026 statements: cash, share count, Machine A spending |
| by 31 Dec 2026 | Factory-acceptance results for Machine A (overdue since September) |
| 10 Jan 2027 | China's suspension of its October 2025 export controls expires |
| H1 2027 | Machine A installed and commissioned in Louisiana (company guidance) |
| 31 Dec 2027 | Works date: Machine A on commercial feed and a binding offtake |
Move a slider, and the value, the rating and the two price levels are worked out again with the same arithmetic as the rating above. The gold mark on each scale is the desk's setting; tap a chip to jump to a value.
The sliders run the analyst's own rating function (translated line by line from Python and tested against it on random inputs), so at the desk's settings they give exactly the card's numbers, and the two price levels are worked out again at each setting. The label here follows the numbers; cheap also needs the cheap line without late builds, and the written tests. Metal prices and costs are held flat to the end of the horizon.
The same function at the rankings page's regular price columns: sensitivities, not forecasts. Works and late EV move in proportion to NdPr oxide; stalled value, funding and shares fixed. Columns are the rankings' NdPr cases. Peer multiples, the stalled value and the cap on the value if it fails are held at the start close; in reality they move with the metal. Columns far above a study's tested range are rough; costs are held flat. Sensitivities, not forecasts.
Swipe the table sideways for more columns →
| Price | Expected value | If it works | If it fails | Cheap below | Expensive above |
|---|---|---|---|---|---|
| NdPr US$75/kg | 0.45× · expensive | 0.72× | 0.37× | none | C$0.82 |
| NdPr US$95/kg | 0.55× · expensive | 0.91× | 0.42× | none | C$1.01 |
| NdPr US$130/kg | 0.71× · expensive | 1.24× | 0.53× | none | C$1.35 |
| NdPr US$180/kg | 0.94× · expensive | 1.72× | 0.68× | C$0.62 | C$1.84 |
Ucore Rare Metals is worth about C$2.29 a share in Nov 2030 if the plan works and C$1.02 if it fails. At a 25% chance the expected value is C$1.34, 0.61× today's C$2.18: expensive.
Each step says first what its numbers are, then what the step does, then shows the sum.
Shares times price is what the whole company costs on the stock market today. Everything below is compared with this price.
If Machine A runs with a binding sales contract by end-2027, it is valued like Lynas per tonne of capacity.
Only a small raise is needed before the works date; the C$150M Line 1 money is not part of this case.
All of the company's value divided by all the shares there will be (152.07 + 9.89 = 161.96M), then compared with today's price.
No rare-earth processing peers exist, so a stalled Ucore is valued on a stand-in.
A plant that runs late is valued like the works case 1.5 years later, with C$30M more overhead to fund.
Each kind of miss counts by its share.
Each future counts by its chance. Over 4.00 years, 1.10× or less is expensive and 2.00× or more is cheap; between is fair. Expensive and cheap also need the written tests.
The share prices at which, everything else as above, the rating would turn.
Where a reasonable person could choose differently, and what that would do. Each result is the model re-run with that one change; the sliders above let you combine them.
The analyst's lines below are for the earlier case (chance 30%, expected value 0.644×). The calculation in brief above uses the case of 16 Nov.
Codes in the analyst's own notes: E = expected value as a multiple of today's price; Mw, Mf = value if it works, if it fails; pw = chance it works; q = share of misses that are late builds; hurdle = 1.5 times in four years (10.7% a year); A1 = valued at what the market pays similar companies today; C4 = new shares for the money still to raise; F1 = the value if it fails is capped at the company's own value today per unit; F2 = a rescue share sale at 80% of value; F3 = a late build valued like success, discounted for the delay; R1 = the written-reason test; X1 = currencies converted at the ECB rate of the start date; yardstick B = the rankings' own target, shown and never counted.
Unchanged from v0.4: Machine A at Lynas's US$433.4k per tpa × 600 = US$260.0M = C$370.8M ÷ 161.96M shares = C$2.29; late ÷ 1.10668^1.5 with C$30M more overhead → C$1.78; stalled C$96.1M with an F2 raise → C$0.52.
| Metal column | E | Works | Fail | Cheap below | Expensive above |
|---|---|---|---|---|---|
| NdPr US$75/kg | 0.47× | 0.72× | 0.37× | no cheap level | C$0.86 |
| NdPr US$95/kg | 0.57× | 0.91× | 0.42× | C$0.52 | C$1.07 |
| NdPr US$130/kg | 0.74× | 1.24× | 0.53× | C$0.72 | C$1.43 |
| NdPr US$180/kg | 0.99× | 1.72× | 0.68× | C$1.01 | C$1.95 |
Method: Works and late EV move in proportion to NdPr oxide (the v0.4 method); stalled value, funding and shares fixed. Columns are the row's NdPr cases. Limits: Peer multiples, the stalled value and the F1 cap are held at the start close; in reality they move with the metal. Columns far above a study's tested range are rough; costs are held flat. Sensitivities, not forecasts.
# Ucore Rare Metals (TSXV: UCU) - value rating v0.4.1 (16 Nov 2026). C$M; share values C$. Unchanged rules.
# Works: Machine A (600 tpa) running with a binding offtake by 31 Dec 2027, at Lynas's EV per tpa of REO output
# (US$433.4k; one name). Fail: q = 0.4 late (works EV discounted 1.5 years at 10.668%, 30 more of overhead) and a
# stall (Element 25's EV + half Machine A's capex = C$96.1M, F2 raise).
USD = 1.1052 / 1.5758 # US$ per C$
DEFAULTS = dict(
P=2.18, pw=0.30, q=0.4, delay=1.5,
ev_per_tpa=433.4, tpa=600, # US$k per tpa (Lynas, 27 Oct 2026); Machine A capacity
ndpr=110.0, # US$/kg; works and late EV move in proportion (estimate)
uses=55 / USD + 10 / USD + 25.0, # Machine A remaining + first fill + 5 quarters of overhead (estimates)
cash=75.0, milestones=21.1 / USD, late_extra=30.0,
stalled=96.1, stalled_s=13.4,
years=4.0,
)
SHARES = 152.070251
UNITS = 0.640 + 0.025710
INSTR = [(0.364511, 0.75), (0.501113, 1.75), (3.103334, 0.8787), (1.2, 1.3748)] # warrants; option buckets (strike
# table not available: buckets rebuilt from v0.4's counts and proceeds at C$2.18, C$1.22 and C$0.61)
R = 1.5 ** 0.25
HI = 10
def _branches(i, P):
itm = [(n, k) for n, k in INSTR if k < P]
n_i = sum(n for n, _ in itm); money = sum(n * k for n, k in itm)
EVw = i['ev_per_tpa'] * i['tpa'] / 1000 / USD * i['ndpr'] / 110.0
gap = max(0.0, i['uses'] - i['cash'] - i['milestones'] - money)
base = SHARES + UNITS + n_i
Mw_ps = EVw / (base + gap / (0.8 * P))
Ml_ps = EVw / R ** i['delay'] / (base + (gap + i['late_extra']) / (0.8 * P))
n0 = SHARES + UNITS + 0.865624
A, s = i['stalled'], i['stalled_s']; p = 0.8 * (A - s) / n0
Ms_ps = A / (n0 + s / p)
return dict(EV_works=EVw, gap=gap, shares_works=base + gap / (0.8 * P), Mw_ps=Mw_ps, M_late_ps=Ml_ps,
M_stalled_ps=Ms_ps, Mf_ps=i['q'] * Ml_ps + (1 - i['q']) * Ms_ps)
def _E(i, P):
b = _branches(i, P); return (i['pw'] * b['Mw_ps'] + (1 - i['pw']) * b['Mf_ps']) / P, b
def _cross(i, line, hi, lo=0.005, steps=4000):
"""Highest price at or below `hi` where E(P) >= line (scan down on a log grid, then bisect)."""
import math
prev = hi
for k in range(steps + 1):
P = hi * math.exp(math.log(lo / hi) * k / steps)
if _E(i, P)[0] >= line:
a, b = P, prev
for _ in range(60):
mid = (a + b) / 2
a, b = (mid, b) if _E(i, mid)[0] >= line else (a, mid)
return a
prev = P
return None
def levels(inp=None):
i = dict(DEFAULTS, **(inp or {})); y = i['years']; hi = HI * i['P']
exp = _cross(i, 1.1 ** (y / 4), hi); cheap = _cross(i, 2.0 ** (y / 4), hi)
fps = _E(i, cheap)[1]['M_stalled_ps'] if cheap else None
return dict(expensive_above=exp, cheap_below=(cheap if cheap and cheap > fps else None),
cheap_arith=cheap, fail_ps_at_cheap=fps)
def value(inp=None):
i = dict(DEFAULTS, **(inp or {})); P = i['P']; b = _branches(i, P)
E = (i['pw'] * b['Mw_ps'] + (1 - i['pw']) * b['Mf_ps']) / P
Eq0 = (i['pw'] * b['Mw_ps'] + (1 - i['pw']) * b['M_stalled_ps']) / P
return dict(b, Mw=b['Mw_ps'] / P, Mf=b['Mf_ps'] / P, M_stalled=b['M_stalled_ps'] / P, E=E, E_q0=Eq0,
per_year=E ** (1 / i['years']) - 1)
if __name__ == '__main__':
r = value(); l = levels()
print({k: round(v, 4) for k, v in r.items()}); print(l)
# desk, 16 Nov: the counted case
DEFAULTS.update({'pw': 0.25})Fact from a filing, exchange or agency · Estimate the desk's own · Desk choice a judgement you can challenge.
Every number here is computed by the same model that drives the rating card and the sliders, so the three cannot disagree.
How we rate a share, with two worked examples →
Human judgement (Chung Huynh) on what matters. Machine effort on everything else.
Chung Huynh is an independent equity researcher, not a licensed financial adviser.
The author may own shares in companies rated here and may buy or sell them at any time.
A rating is reviewed when the price crosses one of its two levels, after major news or a dated catalyst, and at least every three months.
Ratings use today's metal prices. Every number shows its date and source; figures marked estimate are the desk's own.
Educational research, not personal financial advice.