Actionable insights — 50 new hedge fund pitches (Q2-2026 fund-letter compilation)

1. Read a fund-letter compilation for clusters, not for names

1. Do not treat a compilation as a buy list. Fifty pitches from fifty managers is a survey of positioning, not fifty recommendations that survived one analyst's judgement. Its usefulness is aggregate.
2. Bucket the pitches by theme before reading any single one and count. A theme carrying eight or nine of fifty names is telling you where professional money already sits — which is information about crowding, and therefore about who has to sell if the theme breaks.
3. Flag duplicates explicitly. A name pitched twice by unrelated funds in the same quarter is either an unusually clear opportunity or an unusually consensus one. Either way, treat the second pitch as a warning that you are not early.
4. Hunt for the internal dissent. The most valuable page in a fifty-idea compilation is usually the one that argues against the other forty-nine, because it comes from inside the same information set.
5. Note what is absent. A survey with no gold, no uranium, no rates and almost no energy is telling you which trades bottom-up managers are not doing — which is where crowding risk is lowest.
6. Cross-check every overlap against your own book and, when a pitch contradicts a position you hold, write down which specific facts each side disputes. Usually they dispute none, and the disagreement is about price and timing.

• Whether next quarter's compilation shows the same cluster (persistent positioning) or a rotation out of it (the crowd moving first); a duplicate name appearing a third time; and the dissenting manager either capitulating or being proved right — the single cheapest read on whether a consensus theme is turning.

2. Buy the special situation after a short-seller attack, once the specific allegations are disproven

1. Use the attack as the screen, not the verdict. A published short campaign creates a forced-sale price in a business whose operations may be untouched. That is an entry, provided you can test the claims.
2. Wait for the operating data that the thesis said would not appear. Do not argue with the report; let two or three quarters of reported numbers settle it.
3. Look for third-party validation from the party with most to lose. The strongest disproof of "this business is a house of cards taking share dishonestly" is the incumbent it takes share from signing a commercial agreement with it.
4. Re-underwrite the business as a compounder, not as a rescue. If it clears, you are no longer holding a special situation, and the position should be sized and held accordingly — say so explicitly, because the temptation is to keep trading it like the trade it used to be.
5. Watch for the model changing shape. A capital-heavy business migrating to fees (managing other people's capital against the same assets) changes the multiple it deserves once the fee engine flips from investment to harvest.
6. Price the new optionality at a floor, not a forecast — one signed contract with a number attached beats an addressable-market slide.

• The specific falsifiable claim in the original short report and whether reported data has now addressed it; any OEM or regulator action that would confirm rather than refute the thesis; SCI II fundraising progress (the fee engine); and, for the power leg, whether the second Mod-1 order arrives — one contract is a proof of concept, two is a business.

3. Orphaned equity: buy the capital-structure event, not the operating story

1. Screen for businesses whose problem is the balance sheet rather than the business — legacy loss-making contracts completed, negative book equity, a discount to pro-forma tangible book after an announced transaction. The operating fix is already done; only the financing is unresolved.
2. Read the terms of the fix, not the headline. A rights offering priced far below the market, with no over-subscription facility, is designed to punish passive holders and reward participants — that asymmetry is the opportunity, and it is disclosed in the documents.
3. Identify who backstops it. Unexercised rights going to a small group of large existing holders tells you those holders want more of the company at that price. Follow the money that is already inside.
4. Commit to participating in full before you size the position, because a position you cannot fund through the offering is a position that will be diluted out of existence.
5. Value the post-event entity, not the current one. Compute the multiple on forward EBITDA or earnings after the capital comes in and the bidding constraint is lifted.
6. Find the historical template. Rights offerings with identical structures have identical outcomes often enough to be worth naming — and naming the analogue forces you to price the timeline too (in the template case, years of sideways price action before the payoff).
7. For the pure merger version: when the shareholder vote is done and only a regulator remains, the residual risk is timing, and the second leg of the return is what management does with the discount afterwards (buybacks below tangible book).

• The record date and subscription window (missing it is the whole risk); whether the backstop group increases its commitment; the first post-offering quarter, where the bidding capacity thesis either shows up in new awards or does not; and, for the merger case, the regulator's calendar and the first buyback authorisation after close.

4. Screen the capital structure for a forced seller — the covenant that realigns a controlling family

1. Start from the debt, not the equity. In a controlled company whose stock is down 90%, the family's incentives and the minority's have usually diverged completely; only a change in the debt can re-converge them.
2. Look for a distressed exchange with a control trigger. The pattern: a fund buys the notes cheaply, offers an exchange at a discount to face that cuts net debt materially, and attaches a shorter maturity plus a clause handing over most of the equity if it is not repaid.
3. Ask what that clause forces management to do. A family facing loss of control must monetise assets — which is precisely the value-realising action minority holders could never previously compel.
4. Value the assets separately from the operating business. Broadcast licences, real estate and other saleable assets are the collateral for the thesis; the declining operating business is not.
5. State the execution risk explicitly in the write-up — timing of sales, execution, market conditions — because an asset-value thesis with a deadline is a race, not a valuation.
6. Enter before the incentive is widely understood and add as the alignment becomes visible in management's actions rather than its statements.

• The new maturity date as the real catalyst calendar; announced station or property sales and the prices achieved; any refinancing that removes the trigger (which would destroy the thesis by restoring the family's freedom to do nothing); and insider purchases or sales around the deadline.

5. The free-cash-flow-yield discount screen — buy the cheap asset whose bear case is already known

1. Rank a sector on free cash flow relative to price, not on earnings multiples, and shortlist the widest discounts to the peer group.
2. Name the bear case out loud and be specific — "shorter Permian resource life", "COVID-era fleet depreciation" — because a discount without an identified cause is usually a discount you do not understand.
3. Ask whether the market is pricing a permanent impairment or a temporary distortion. Accounting distortions (depreciation from an abnormal purchase period) resolve mechanically; resource depletion does not, and must be paid for with optionality elsewhere.
4. Require a balance sheet that lets you wait — investment grade, or debt low enough that a bad two years is survivable.
5. Look for the unpriced call option: exploration acreage in another jurisdiction, an under-monetised second segment, or property carried far below use value.
6. Buy the sell-off caused by something that helps the customer, not the company — an oil price falling on a peace headline reprices the equity without changing the reserves.

• The depreciation or depletion charge normalising in reported results (the mechanical part of the thesis); capex falling as the distortion clears; first cash returns to shareholders; drilling or appraisal results in the optionality jurisdictions; and, for a private-market valuation gap, any transaction in the sector that marks the assets.

6. Test an AI-disruption story against the customer, not the technology

1. When a franchise de-rates on a technological threat, write out the disruption story as its proponents would — in full, without strawmanning. If you cannot, you do not understand the risk you are underwriting.
2. Ask who the actual buyer is and what they are actually buying. Disruption arguments almost always assume a retail-style buyer optimising on price; institutional buyers optimise on settlement certainty, regulatory standing, recourse and bespoke terms.
3. Find the historical precedent where the same disruption already happened — and identify why it worked there. If it worked because the product was commoditised, then the test is whether your product is commoditised. Usually that is the whole answer.
4. Check whether the incumbent can simply launch the disruptive product itself. A threat an incumbent can copy in a quarter, with better clearing and distribution, is a feature, not a competitor.
5. Identify the proprietary data asset. Ask specifically whether it is scrapeable. Decades of private pricing across counterparties is not, and that is a durable barrier in an era when public information has been commoditised.
6. Reframe the threat as an expansion where it applies to a market you do not serve. A new rail that wins in cross-border business payments or remittances is additive if your revenue comes from consumer card volume.
7. Then require a discount, and buy it. The method only produces returns when the fear has actually compressed the multiple.

• Evidence that would break the rebuttal rather than confirm it: a genuine AI-native broker winning a Fortune 500 commercial programme; perpetual futures posting volume from commercial hedgers rather than speculators; a large merchant routing consumer volume off the card rails; and, in every case, the multiple — the trade is over when the discount is gone, regardless of whether the story was right.

7. Reframe the asset: find the software business hiding inside a physical one

1. Ask what a company is being valued as, and whether that classification still describes what it does. Index membership and sector labels are sticky; business models are not.
2. Separate collecting rent from running the operation. A landlord earns a contracted payment; an owner-operator captures the upside from occupancy and pricing. Only the second is exposed to its own operating improvements.
3. Locate the proprietary layer — an operating platform, analytics, a pricing system — and ask whether it drives margin and occupancy, or is merely internal plumbing.
4. Test whether the layer is being monetised, or only used. Early external monetisation is what converts an operating advantage into a re-rating argument.
5. Check the hiring. A property company deliberately recruiting technology executives is telling you which business it thinks it is in.
6. Apply the same lens to conglomerates: if one subsidiary is a genuine high-return compounder inside a cyclical group, the group's cyclical multiple is the mispricing.

• Disclosure that separates the platform's economics from the property's; external customers paying for the software; margin expansion that outruns rent growth; and the negative signal — a reframing that never shows up in reported segment margins, which means the label changed but the business did not.

8. Read supply discipline as a claim to be verified, and price the cycle it eventually creates

1. In a consolidated commodity industry, the variable that matters is behaviour, not demand. Two or three scaled producers who add capacity destroy the return; the same producers who add density or yield instead preserve it.
2. Find the discipline in what management chooses to spend on — areal density rather than unit capacity, AI-grade memory rather than commodity lines, capex tied to customer commitments — and take that as the testable claim.
3. Check the physical constraints that enforce discipline whether or not anyone means it: wafer intensity of the premium product, the impossibility of converting one product line into another, greenfield build times, tool lead times.
4. Look for structural de-commoditisation — long-term agreements, co-designed products, customer-specific engineering — which is what turns a price cycle into a contract business.
5. Then run the counter-argument on the same facts. Record profits fund capacity; capacity arrives with a lag; scarcity was what produced both the earnings and the multiple. Write down when the added capacity lands.
6. Size for the fact that the equity turns before the industry does. "For the companies, such investments may be rational. For the stocks, it could prove more complicated."

• Capex guidance rising faster than customer commitments; equipment orders (the lead indicator that discipline is ending — note SK hynix funding lithography from ADR proceeds); the first quarter in which a producer adds units rather than density; contract-price resets on long-term agreements; and the gap between spot and contract pricing, which is where the turn shows up first.

9. Anchor a valuation to a private-market or peer transaction, not to a multiple you chose

1. Find a recent transaction in the same business — a private-equity deal, a takeover of a close peer, a controlling holder's own bid — and use it as the reference point.
2. Adjust for the specific differences (margin gap, scale, jurisdiction, share class) rather than applying the headline multiple unaltered.
3. Express the upside as a gap to that anchor, so the thesis has a number that a third party has actually paid rather than one you have assumed.
4. Where the anchor is a margin rather than a multiple, treat the gap as the operating plan: identify what the better operator does differently and whether the target can copy it.
5. Prefer anchors that are hard to argue with — a completed transaction beats a broker's target; a controlling holder's own bid is the hardest anchor of all, because it is the buyer's own view of value published under legal obligation.
6. Beware anchors that are themselves inflated by the same cycle that inflated your target.

• Whether the anchoring transaction closes at the announced terms; a second comparable transaction confirming or resetting the level; the margin gap actually narrowing in reported results (an anchor that never converges is a thesis that has failed quietly); and, for spin-offs, the tax and lock-up constraints that can make an obvious acquisition impossible for years.

10. Write the detractor up honestly — and let the write-up carry the falsification test

1. Re-underwrite losing positions in public, at length, with the loss stated first. A letter that only discusses winners is not a research process.
2. Distinguish a timing error from a thesis error explicitly and say which you think it is — the distinction determines whether you add, hold or exit, and it is the only claim that can later be scored.
3. Re-derive the valuation at the new price in absolute terms (EV/EBITDA, P/E, net cash) rather than relative to your cost, so the decision is made on the security in front of you.
4. State the mechanism that would resolve the delay, and what it depends on that is outside your control.
5. Attach an external marker of value that does not depend on your own opinion — a peer transaction, a strategic buyer's interest, an asset sale.
6. Accept the possibility that early and wrong are the same thing, and size accordingly.

• The named mechanism actually turning (project approvals resuming; ARR reaccelerating) within the timeframe implied; a second consecutive quarter of the same excuse, which converts a timing call into a thesis error; and the external marker being withdrawn — a peer deal breaking, or the strategic interest evaporating.