Actionable insights — Weekly SSR: the capex-burn diagnosis, Micron's take-or-pay durability test, the holdco-NAV stub screen, the preferred-cushion check, the SOTP on Sea

1. Diagnose the capex-burn de-rate — separate multiple compression from earnings, then sort the spenders by who can fund it

1. When a whole cohort sells off, decompose the move: is it earnings falling, or is it valuation multiple + free-cash-flow falling? If EPS is fine and the multiple compressed (here Mag7 32× → 22×), the market is repricing the business model, not the quarter — specifically punishing capex with unprovable ROI.
2. Quantify the burn directly: track aggregate free cash flow of the spenders (here ~⅓-trillion → ~zero) and the capex trajectory (10% → 36% → ~90% growth, heading to ~$1T by 2028). Rising GAAP margins don't earn full credit if the balance sheet has to lever up to compete.
3. Rank the names by fundability, not just drawdown: who can raise capital cleanly (negative net debt, a 100-yr bond, an absorbed equity raise) vs who gets hammered on a raise rumor or shows rising CDS. Buy the de-rate only in the fundable ones, and size small because timing the bottom is hard.

• EPS-flat / multiple-down divergence; aggregate-FCF collapse; raise-rumor reactions and CDS moves as the "who can fund it" tell; 2027 capex-growth guidance slowing versus expectations as the trigger for a semis unwind.

2. Re-rate a commodity business off its new contractual floor — the take-or-pay durability test

1. When a historically cyclical, commodity producer signs binding multi-year take-or-pay contracts (customers must buy set volumes or pay penalties), stop valuing it on the old boom-bust cycle — the demand is now contractually guaranteed regardless of spot prices.
2. Test the floor's durability: is the guaranteed minimum price set high enough that even a deep downturn leaves margins above any prior peak? Are customers pre-paying cash deposits to secure allocation (skin in the game)? What share of total revenue do the contracts cover?
3. If the answer is "a high floor + upfront cash + >half of revenue," apply a higher, less-cyclical multiple than the market's commodity-cycle reflex — but keep the bear case live (does the product actually have a moat, or can rivals add capacity over years?).

• Take-or-pay vs soft "letters of intent"; the floor-price level vs prior peak margins; upfront customer deposits; the % of revenue locked; and the moat rebuttal (state-subsidized new entrants).

3. Buy the holding company below the value of its one good asset — the NAV-stub screen

1. Find a parent whose publicly-valued stake in a single subsidiary is worth more than the parent's entire enterprise value — the market is mispricing the parent as its old (out-of-favor) business.
2. Confirm the asset is clean: is the subsidiary debt-free, and does the parent hold a controlling stake? Add any extra parent assets (cash, crypto) on top to widen the margin of safety.
3. Express the gap as a discount-to-NAV and size it to the catalyst: the discount closes as the subsidiary delivers on its growth/EBITDA targets and the market re-attributes the value.

• Sub market cap > parent EV; a debt-free, controlled subsidiary; extra parent cash/securities; a growth catalyst at the sub that forces the re-attribution.

4. Buy the preferred over the common when the yield give-up is tiny and the equity cushion is huge

1. Compare the preferred's yield to the common's: if the common yields only modestly more, the extra income isn't worth the equity risk — the preferred is the better risk-adjusted bet.
2. Measure the cushion: how much common-equity market cap sits below the preferred (i.e., must be wiped out before the pref dividend is impaired)? A multi-billion cushion on a profitable issuer is a thick margin of safety.
3. Prefer prefs trading below par (lifts the yield and adds pull-to-par upside); confirm the issuer's quality (diversified, profitable, low private-credit exposure) and score the risk explicitly.

• Small common-vs-pref yield gap; the size of the equity cushion below the pref; a discount to par; issuer profitability and rate-hedge characteristics.

5. Value a multi-segment conglomerate on sum-of-the-parts — then check the optionality on top

1. Break the business into its distinct segments and value each on the right method (DCF, EV/sales, a comp), rather than slapping one blended multiple on a company that's really three businesses.
2. Apply a holdco discount and divide by the full share count to get a clean per-share fair value vs spot — the headline number.
3. Layer the optionality you're not paying for (a margin-expansion lever, a new product that deepens the moat) as the upside beyond the base case, and stress the downside drivers that caused the selloff.

• Segments worth valuing separately; the right method per segment; net-cash as a % of cap; a hidden margin lever; and the named downside drivers (competition, credit costs, regional spending).

6. Re-underwrite a disaster name by calling the sell-side — and price the re-rate clock, not just the value

1. After a disaster-driven plunge, go past the headline: call the analysts who actually cover it and rebuild the NAV asset-by-asset, localizing the damage (which asset, what share of NAV, how reversible).
2. Separate the value question from the timing question: even when the value is clear, ask what it takes to restore market trust — often several clean quarters with no repeat — and don't expect an instant re-rate.
3. Cross-check the relative multiple: a former premium name now trading at a discount to lesser peers (even on multiple metrics) is the mispricing; treat it as a long-term allocation, not a quick trade, when the clock is long.

• The impaired asset's NAV share and reversibility; a former-premium name now at a peer discount; the number of clean quarters required; management's recent guidance-cut history as the trust deficit.

7. Buy the distressed preferred when the issuer can still fund the coupon — and prefer the debt-free structure

1. When crypto- or cycle-linked preferreds gap down well below par, ask the only question that matters: can the issuer keep paying? If it retains an equity-issuance lever to fund the dividend, the selloff is an opportunity, not a default signal.
2. Rank by capital structure: a debt-free issuer with only preferreds is safer than one that still carries senior debt ahead of the pref; and a cumulative dividend is safer than a non-cumulative perp that can suspend without accruing.
3. Size small and stagger entries given the underlying volatility (here crypto + war risk), targeting the pull-to-par plus the coupon as the return.

• An equity-issuance lever to fund the coupon; debt ahead of the pref vs a pure-pref structure; cumulative vs non-cumulative; discount to par; and the volatility of the underlying that justifies small sizing.

8. Track the enterprise pricing-power inflection with market-share data — "model routing"

1. Find a hard dataset that shows customers substituting away from the incumbents' premium product toward cheaper alternatives — here UBS/Bloomberg surveys of enterprise AI-model usage by provenance over time.
2. Quantify the cost gap driving it (a DeepSeek model at <1% the cost of running a frontier model) and the behavior change ("model routing": cheap models for easy tasks, frontier saved for hard reasoning) — that's the mechanism eroding the moat.
3. Translate to the second-order trade: if the model providers' pricing power breaks, question whether the hyperscalers funding the buildout can ever monetize, and lean against the most capex-exposed names (loops back to Insight 1).

• Provenance-by-time market-share surveys; cost-per-token gaps; "model routing" adoption; and per-employee AI bill blowouts (e.g. Uber's quarter-long budget) as the cost trigger.

9. Hedge the whole AI trade with an ultra-long-dated bond — convexity, not a single-name short

1. Instead of shorting a frothy equity outright, find a long-duration bond from a richly-priced issuer in the same theme — the longer the maturity, the more a small spread move swings the price (convexity).
2. Size the asymmetry: with 30+ years of duration, a ~100 bps spread widening ≈ a ~30-point price drop, so a modest credit re-rating delivers an outsized payoff with defined, bond-like downside.
3. Use it as a portfolio hedge to the AI-infrastructure cycle, triggered when a new issue is priced too tight versus governments and trading starts to fade.

• The longest-dated tranche of a theme issuer; spread vs Treasuries at issue; duration-driven price sensitivity; post-issue bond-price fade as the entry.

10. Harvest the hard legal catalyst — the Section 16(b) short-swing-profit windfall

1. Watch for a >10% insider/holder who bought and sold the same stock within six months — Section 16(b) of the 1934 Act forces them to disgorge those "short-swing" profits back to the company, a non-discretionary cash inflow.
2. Size the windfall against the market cap and read the 8-K for the settlement amount and court-approval status; the cash accrues to the company (a hard, event-driven catalyst, not a narrative).
3. Pair it with option structures around the event if the move is binary and the timing is known.

• A >10% holder's buy-then-sell inside six months; the 8-K disgorgement amount vs market cap; court-approval status; and a known event date for option overlays.