Phase Space AI

Trade Construction

Hims & Hers Health [HIMS]

HIMS — Trade Construction, Sizing and Risk

This memo issues no position verdict — no Long, Short, Watchlist or Avoid. It scores Criteria and outputs an analysis. What follows is the construction work the book needs if it elects exposure, plus the explicit sizing view the brief asked for.

Spot $24.68 · shares 231,457,695 · market cap $5,712m · net debt $223.2m · EV $5,936m

---

1. The scored position

Two BINDING failures. On the long-only absolute-return strategy that consumes this memo, that is dispositive without any judgement from me — and it is not a data artifact. Every input is filed, present and cross-checked (net income ÷ shares reconciles to filed EPS in both periods tested).

---

2. The sizing question the brief posed

Given the reference book holds it tiny at 97% vol — should size be capped irrespective of the margin?

Yes. Capped hard, at ≤0.5% of book, and for three reasons that have nothing to do with volatility.

(1) Inverse-volatility sizing is necessary but not sufficient here. The interim control in the framework is inverse-vol sizing, and it works because a fat-left-tail name is usually a high-vol name and gets sized down automatically. HIMS at 89.6% realised / ~93% ATM implied is already 7x SPY vol, so the mechanical sizer will produce something small. But inverse-vol sizing prices dispersion, and the dominant risk here is not dispersion — it is a discrete, non-diversifiable, non-continuous legal event. A DOJ action, an FDA API restriction, or a refiled Novo injunction to December 2031 does not arrive as a 3-sigma move in a lognormal process; it arrives as a step change in what the business is legally permitted to sell. Volatility-scaled sizing systematically under-caps that. The cap must be a separate, explicit instruction, not an output of the vol sizer.

(2) The central question is INDETERMINATE by the company's own choice, and the disclosure is narrowing. GLP-1 revenue has never been disclosed in any period. The only bound is "a majority of US revenue came from non-GLP-1 offerings" — one-directional, unquantified, unchanged for two reporting periods. Retention and churn are not disclosed at all ("churn" appears zero times in ten consecutive quarterly releases). And in the same release that reported the US decline, the margin collapse, the first operating loss in two years, a $15m litigation accrual and a $33.5m restructuring charge, the company announced it is retiring the quarterly shareholder letter in favour of an annual one. A position cannot be monitored on a metric that is not published. The invalidation conditions in §5 are all second-order proxies because the first-order metric does not exist. That, independently of price, caps size: you cannot size to a thesis you cannot mark.

(3) The failure mode is correlated with exactly the book's other GLP-1 exposure, and the M&A makes it worse. The natural instinct is that international acquisitions diversify. They do not: Eucalyptus, agreed February 2026 for up to $1.15bn, is Australian GLP-1 telehealth. Zava is European telehealth. The company is using its balance sheet — now $223m net debt, from $323m net cash four quarters ago — to buy more of the same regulatory exposure in more jurisdictions. Any book that also holds GLP-1 manufacturers, obesity plays, or DTC pharmacy has a concentration here that a single-name sizer will not see.

On the reference book's own numbers, note this reading. The professional holds HIMS at 0.8% of book, 97% implied vol, +88% target, Sharpe 0.90. The +88% target and the 0.8% size are the same statement: high convexity at low confidence. That is not a conviction long dressed small; it is a lottery ticket sized like one. The Sharpe of 0.90 is a ratio computed on a very small denominator and should not be read as second-highest-conviction. My +37% target at ≤0.5% of book is a slightly more conservative version of the same position, not a disagreement with it. Where I would depart from the reference: the +88% requires the multiple to expand to the p90 of the post-compounding trading range (3.47x EV/sales), which requires the enforcement overhang to lift — and it did the opposite today.

Cap: ≤0.5% of book, hard, irrespective of the valuation margin. No averaging down. No option overlay financed by short puts.

---

3. If the book elects exposure — the vehicle, from the actual chain

Chain pulled 2026-07-29 (data/chain.json, Alpaca options snapshots with Greeks). No structure below was assumed; every strike shown has quoted two-sided size and real open interest.

3.1 Preferred: Jan-2027 $30 / $40 call spread

• Debit at the offer/bid (worst fill): $2.50. Mid-to-mid: $2.06.
• Max value $10.00 → max profit $7.50, 3.0:1 at the worst fill; 3.85:1 at mids.
• Breakeven $32.50 (+31.7%). Max payoff requires $40.00 (+62%).
• Both legs carry >6,000 OI and 100–1,000-lot quoted size. Fillable at institutional size — this is the antithesis of the HCA precedent (18 contracts of max OI across an entire chain).

Why a spread and not stock or calls outright. ATM IV is ~98–99% on the January expiry against 89.6% realised — buying a naked call pays ~10 vol points of premium and takes the full enforcement tail. The spread nets most of the vol (98.8 vs 97.7), caps the loss at a known $2.50, and — critically — caps the loss at the moment the thesis is wrong for a reason no stop can front-run. A DOJ complaint or a refiled Novo injunction does not gap through a stop in an orderly way. Defined risk is the point, not the convenience.

Sizing at the cap. On a $100m book, 0.5% = $500k of risk capital → 200 spreads at $2.50 (2,000 shares of notional delta ~0.18 net per spread, i.e. ~36,000 share-equivalents, ~0.2% of one day's volume). Trivially executable.

3.2 Explicitly rejected

3.3 Entry timing (Momentum Criteria is MEASURED — timing only, never a veto)

12-1 momentum −27.9%, bottom quintile. 37.8% of the 52-week high; −64% from an all-time high set five months ago (2026-02-19); below the 200-day at $30.61; −15.8% today on the FTC filing. Six-month momentum was positive before today and is now roughly flat. Read: the tape is still falling and the news flow is still arriving. Nothing argues for immediacy, and the largest scheduled information event is 12 days away. If exposure is elected, wait for the 10 August print — the option premium being charged for it is precisely the market's estimate of how much you learn, and you can buy the January spread the following morning with the US-revenue and gross-margin questions answered.

---

4. Bear case — logged per Downside Criteria (MEASURED; blocks nothing)

Primary bear: 35% · target ~$18 (−27%) · named cause: the §503A exemption is closed by action, not argument

Any one of three dated, live triggers is sufficient:

1. The DOJ referral produces an action. HHS's General Counsel referred HIMS to the DOJ in February 2026 for potential FDCA and Title 18 violations. Title 18 is the criminal code.
2. The FDA executes the restriction it announced in February 2026, in which HIMS was named directly — restricting GLP-1 active pharmaceutical ingredients intended for non-approved compounded drugs "mass-marketed as similar alternatives to FDA-approved drugs."
3. Novo refiles. The '343 patent suit was dismissed without prejudice on 9 March 2026 with the right to refile expressly reserved. The patent runs to 5 December 2031 and the relief sought was a permanent injunction for its full remaining term.

Arithmetic. The compounded remainder of the "personalized" category (>70% of US revenue, of which the compounded share is undisclosed) goes to zero. The residual is a branded-drug reseller at low-60s gross margin with 36% of revenue in marketing and a 58-month gross-profit payback. On the post-exit minimum own multiple of 1.57x EV/sales applied to ~$2.9bn of acquisition-padded revenue: EV ~$4.55bn, less $400m net debt → equity ~$4.15bn → ~$18/share, −27%. At 1.0x sales — not observed in this name but routine for broken DTC — ~$11/share, −55%.

Second independent downside: 10% · the FTC action attacks the conversion engine, not just the P&L

The FTC, joined by California and Utah, sued in N.D. Cal. on 29 July 2026. Two allegation sets: (i) sensitive health data shared with Meta and Snap through website tracking despite privacy representations; (ii) charging consumers "almost immediately after they submit an intake form," with "virtually no opportunity to review the provider's recommended treatment, much less consent to it," plus obstructed cancellation.

The second is the dangerous one, because that flow is the funnel. Being ordered to insert genuine informed consent between intake and billing attacks conversion at the source — not a fine, a redesign of the product's economics. Against a 4.9-year payback and a marketing line at 36% of revenue, a step-down in conversion is not absorbable.

The accrual is $15.0m. The company itself warned in the Q1 10-Q that "there is a reasonable possibility that the final liability could be materially higher than the amount accrued," and that it had "submitted a substantive offer to the FTC to settle this matter." The FTC rejected it and sued — nine weeks after the filing that described the negotiations as "active." Two state AGs joined, and states (unlike the FTC under §13(b) post-AMG) can seek civil penalties. The $15m accrual should be treated as a floor, not an estimate.

Going concern: not argued, explicitly

$750.9m of cash and investments; a 0%-coupon convertible not due until 15 May 2030; $162.4m undrawn revolver; positive reported CFO. Liquidity is not the risk and it would be dishonest to imply it is. The risk is whether the business is legally permitted to sell its largest product, which is not a liquidity question and cannot be cured with a financing.

Caveat the framework requires me to state: the Downside Criteria is MEASURED, pending promotion, because the principal does not yet trust bear cases in this system. These probabilities are unscored priors. They are logged so ledger_scorer.py can Brier-score them, not because they are calibrated.

---

5. Invalidation and confirmation — dated where possible

Invalidation (thesis-that-it-is-uninvestable is wrong)

Confirmation (the analysis is right)

Monitoring hazard, stated

I1, I2, C1, C2 and C4 are all readable on 10 August. After that, the quarterly shareholder letter is being retired and the primary disclosure narrows. Set the monitoring cadence to the 10-Q, not the release, and read the receivables and accounts-payable lines before the headline — Q1'26's $89.4m of operating cash flow contained a +$167.6m accounts-payable build, and that is where the next surprise will show up first.

---

6. Exit discipline

• Hard cap ≤0.5% of book. No averaging down under any circumstance — the bear case is a step function, so a lower price is evidence, not opportunity.
• Defined risk only. The maximum loss must be known at inception and paid up front. No short options.
• Nothing before 10 August. The information is 12 days away and the market is charging 113% IV for it.
• Immediate exit on C5 (Novo refiles, or DOJ/FDA acts) at any price. That is the trigger that revalues the business rather than the stock.
• Automatic exit on C6. A third non-GAAP redefinition or a further disclosure retirement means the position can no longer be marked to a thesis. Exit on the disclosure event, not on the price.