The research archive.

1. Earnings bridge: FY27E adjusted EBITDA $507M / $730M / $958M → equity value per share $3.2 / $13.6 / $22.3 at 5.0–5.5x EV/EBITDA; probability-weighted $13.2 (+8.1% capital, +10.2% with the 2.1% dividend).
2. Consensus (7 analysts) averages Hold with targets of $13.1–13.4; the stock has already rallied +23% since BoM's 16 June declaration.
3. Worst historical moves: −21.5% in five days, −32.8% in twenty; expected return is 0.48x the worst week, against 1.0x for CANE.

$3.2 / $13.6 / $22.3 at 25 / 50 / 25%; probability-weighted $13.2 — +8% against the $12.19 price when Issue #8 was published, +26% against the $10.50 average cost the card is scored from.

1. Four strong events: raw sugar rose a median +13% from onset to the ONI peak and fell a median −24% in the six months after; non-El Niño years show −2% in both windows. May→August 2026 (+12%) has already delivered the pre-peak median.
2. India's monsoon is −13% where it matters least and −14% / −24% in Maharashtra and Karnataka; opening stocks ≈ 3.0–3.5 Mt (a decade low); export ban to 30 September and a 1 Mt import quota. Every commercial house now sees a 2026/27 deficit (median −1.3 Mt).
3. Vehicle level: CANE holds deferred contracts that capture 69–80% of the Mar-27 move, so its probability-weighted return is +8.8% net of carry ($12.45 vs $11.46); Adecoagro's unhedged FY27 sugar gives +10% with the dividend but a −74% bear case; Bunge is conditional below $110.

Mar-27 scenarios 15 / 21.5 / 26c at 25 / 50 / 25% → 21.0c probability-weighted vs the 19.05c forward. CANE $9.83 / $12.67 / $14.63; AGRO $3.2 / $13.6 / $22.3; BG $69 / $112 / $155. The illustrative structure the numbers support is a small CANE position with a smaller AGRO overlay, sized jointly as one sugar factor.

1. Bottom-up demand model: 4,000 to 4,526 TWh by 2030 (2.0% CAGR) with AI adding ~200 TWh, EVs +149, industrial +99, heat pumps +51.
2. Ranked longs: GEV (lowest bull-bear spread), HUBB (28x NTM, transmission hedge), ETN (Q4 orders +200% vs consensus 2027 +9%), EME (31x vs FIX 44x for the same end-markets), AGX (GEV-shadow EPC), ACM (14x NTM).
3. Avoid VRT (51x), FIX (44x) and MTZ (44x): multiples that require flawless execution and bull-case capex. Cleanest pair: long EME / short FIX.

Sizing proposed in the note: GEV 5%, HUBB 4%, ETN 4%, EME 3%, AGX 3%, ACM 2%.

1. FY27E EPS $7.30 / $9.76 / $12.44 at 9.5x / 11.5x / 12.5x → $69 / $112 / $155; 9.5x is what the market paid for peak earnings in 2022-23, 11.5x the five-year median forward multiple, 12.5x today's on raised guidance.
2. Consensus target $141 (9 analysts); forward P/E 12.2x; 1-year volatility 32%; worst five-day move −14.1%.
3. $125.32 on 8 September after +5.4% on Viterra-integration commentary — above our probability-weighted $112.

$69 / $112 / $155; probability-weighted $112, −11% vs $125.32 (+2.4% dividend).

1. Fully collateralised: 1,005 Mar-27, 892 May-27 and 1,048 Mar-28 contracts against an NAV of $11.44; carry is a wash (+0.5% net) because the 2.48% TER exceeds the 3.7% T-bill yield only slightly over the 0.4-year window.
2. Sugar reversals began at the ONI peak month in three of four events; the one that ran (2015-16) had India as a net importer. Hence a calendar exit by end-January unless net imports reach 2 Mt.
3. Worst historical week −8.8% (12 months to 8 Sep 2026); expected return is 1.0x that week's loss. Daily turnover is a few million dollars, so size must be worked over days.

$9.83 / $12.67 / $14.63 at 25 / 50 / 25%; probability-weighted $12.45, +8.6% against $11.46.

1. Micron's move from $367 to a $1,213 peak is the demand signal; equipment orders lag pricing by two to four quarters.
2. LRCX is the most memory-levered wafer-fab equipment name (etch and deposition for DRAM and NAND); AMAT is the broadest exposure to the same capex.
3. Both entered Aug 25, 2026 at the close (AMAT about $480, LRCX about $315), sized 10% and 5%.

Multiples at entry not published; the sizing assumes flat-to-up 2027 WFE guidance from both companies.

1. Realised: CAT +49%, MSFT +21%, APO about +17%, LTM about +17%, EWJV +11%, FCX +4%; losses PBR −8%, MTZ −6%, BABA −3%; LKNCY flat. Exit prices are approximate and marked as such in the closed-trade table.
2. Concentration is deliberate: the top two positions are 60% of the book, semiconductors are 45%, and gold is the 30% non-tech leg. Top five = 90%.
3. This supersedes two published calls, 'avoid VRT at 51x' (Issue #7, Apr 24) and 'buy the hyperscalers, don't chase the chips' (Issue #5, Apr 10). VRT is 21% below its Apr 24 price; the reasoning for the semiconductor concentration will be written up in a later issue. Until then, treat the new positions as positioning, not published research.

The pipeline is now focused on opportunities outside technology to rebalance the book over the next quarter.

1. First position +40.9% (entered $110, exited $155); re-entered Apr 3, 2026 at $366.79 after the stock had run past $362.
2. From the re-entry the stock ran to a $1,213 close on Jun 25, retraced to $739 on Jul 29 and trades near $930; the position was added through May and the late-July dip to an average cost of about $700.
3. HBM economics scale with each GPU generation rather than with the commodity DRAM cycle; the equipment names (AMAT, LRCX) were added Aug 25 as the second leg of the same view.

Re-entered Apr 3 at $366.79 and added through September; the holdings table carries the broker average cost. No target published; the position is sized on the view that HBM pricing holds through the next GPU generation.

1. Indonesia ≈ 59% and Malaysia ≈ 24% of world exports — 83% from the two dry-side countries; yields fall with a 6–24-month lag.
2. CPO RM4,977 (+11% y/y) despite 2.63 Mt of Malaysian stocks; GAPKI sees 2027 Indonesian output at 56.8 vs 58.5 Mt; planters guide RM5,200 for Q1-27.
3. Activation conditions: (i) December CPC confirmation of the ONI peak; (ii) MPOB stocks turning down from the Q4 seasonal high with fresh-fruit-bunch yields falling year on year; (iii) CPO holding above RM4,500 through the production peak.

Not valued yet. A first-half-2027 note would take the planters through the same bridge as AGRO: CPO price → realised price after levies → EBITDA per hectare → multiple. Stocks above 3.0 Mt with CPO below RM4,200 would say the 2027 narrative is early.

1. Issue #7 (Apr 24, 2026) rated VRT 'avoid' at 51x NTM; the stock closed at $323.46 that day and at $255.75 on Aug 25.
2. Power and cooling are the binding constraints on data-center capacity; VRT sells into both.
3. Entered Aug 25, 2026 at the close, sized 10%.

Entry about $256. The multiple at entry is not published; the thesis is that the de-rating has removed the objection raised in April.

1. Sized at 30% as a deliberate hedge rather than a trade: it is the only position in the book that does not depend on memory pricing or data-center capex.
2. Cross-market divergence from Issues #3 and #4: bonds most fearful, commodities in the middle, equities most complacent; gold has re-rated through every equity drawdown this year.
3. Rebuilt in August at an average cost of about $410 after the earlier 6% position; GLD trades near $428.

No target: the position is sized for its correlation, not its upside.

1. Bought at the May 22 low of about $46 after a 15% slide from $54 at the start of the month; the stock trades near $46 today.
2. Outbound travel volumes continue to recover while inbound is only beginning; both run through the same platform.
3. The one non-technology equity position in the book; sized 10% and the template for the non-tech pipeline.

Multiple at pitch not published.

1. A platform with 150M+ monthly users is a demand-aggregation moat no AV player can replicate cheaply.
2. The core rideshare business is generating free cash flow while the market prices disruption.
3. Waymo partnership as the template for robotaxi distribution.

Thesis only; not held and no target published.

1. GTA VI is the most anticipated software release in history; pre-launch monetisation from in-game purchases, GTA+ subscriptions and Shark Cards builds a multi-year cash-flow engine.
2. Valuation should rest on the normalised free-cash-flow engine after launch, not on current losses.
3. Generative AI lowers content cost while deepening game worlds.

Thesis only; not held and no target published.

1. Cross-market divergence: Treasuries most fearful, commodities in the middle, equities most complacent.
2. Barrick lagged the gold price on cost-inflation fears; miners historically show 1.5–2x leverage to gold in trending markets.
3. Number one by proven reserves with 12 Tier-1 assets; technically oversold against gold and its own range.

Expressed in the portfolio through GLD (30%) rather than Barrick equity; no price target published.