Phase Space AI

Trade Construction

Exelixis [EXEL]

Exelixis, Inc. [EXEL] — Trade Construction & Risk

2026-07-29 · spot $56.51 · NO POSITION VERDICT IS ISSUED. This document sizes, structures and bounds a position if the book chooses to take one, and reports the liquidity facts that constrain it. The book decides.

---

1. Liquidity Criteria — PASS on equity, CONSTRAINED on options

1.1 Equity

At $145m/day this supports any size this book would take. A $10m position is ~7% of one day's volume — a fraction of a day to build or exit. Equity liquidity is not a constraint.

1.2 Options — the actual chain, pulled before anything is proposed

The Criteria require the real chain with open interest and quoted size for the specific strikes and expiry — after HCA, where the maximum OI across an entire chain was 18 contracts and nothing tested for it.

Alpaca options contracts + snapshots, retrieved 2026-07-29; open interest as of 2026-07-27 (one session stale, stated).

January 2027 expiry (spans the PDUFA and a plausible CAFC decision window):

December 2026 expiry (brackets the 3 December PDUFA by 15 days — the catalyst-matched expiry):

What the chain actually permits:

• ❌ The default defined-risk structure is uninvestable. A December $60/$70 call spread faces open interest of 216 on the long leg and 2 contracts on the short leg. It cannot be filled at any size. This is the HCA failure exactly, and it is caught here before being proposed.
• ⚠️ A December $60/$65 call spread (OI 216 / 670) is fillable in small size only, and the $60 leg's 58%-of-mid bid-ask means the round trip costs more than the spread's edge unless held to expiry.
• ✅ The only genuinely liquid structure is January 2027: the $50 call at 2,417 OI and the $55 call at 866. A Jan-2027 $50/$60 call spread (2,417 / 343) is the one multi-leg trade with real depth on both legs.
• 📌 Note the $50 December put at 3,600 open interest — by a wide margin the largest single line in the chain, and it is downside protection struck 11.5% below spot expiring 15 days after the PDUFA. Somebody is already hedging exactly the event this memo identifies.

Conclusion: the equity is the vehicle. Options are a supplement in small size at Jan-2027 $50/$55/$60 only. Any structure quoted in this memo names the open interest that supports it.

---

2. Downside Criteria — MEASURED, blocks nothing

Volatility is not the risk. Permanent impairment is. The realistic permanent-loss scenario, with its named cause:

The United States Court of Appeals for the Federal Circuit reverses or vacates the Delaware District Court's 23 October 2024 final judgment in MSN II, removing the 15 January 2030 floor on generic CABOMETYX entry. Appeal No. 2025-1236. Briefing completed 1 August 2025. Oral argument held 4 June 2026. Decision pending as of 2026-07-29.

Why this is the permanent-loss case and not a drawdown: cabozantinib is ~99% of revenue and 100% of profit. The 15 January 2030 date is what supports the ~3.5 remaining years of the annuity, and the annuity is what funds the $900m/yr of R&D on which the entire residual value depends. Remove the date and both legs move at once.

Why the bear case is −40% and not −80%: a reversal would not clear MSN to launch. The '039 patent (expiry 2032) is still headed to trial in the Consolidated Litigation on 2 November 2026, and the '342 formulation patent (2033) was never asserted against MSN. What a reversal destroys is certainty, not the whole estate. The market re-prices the terminal multiple, not the terminal cash flows, inside twelve months. §4.3 of the valuation puts that at 11x trailing on ~$3.10 = $34.

The bear case is logged to the ledger with its probability and its cause, per the Criteria.

Secondary risks, ranked

1. A Complete Response Letter on 3 December 2026. The NDA rests on the ITT overall-survival endpoint; the second dual primary endpoint missed on 22 June 2026 (HR 0.83, p = 0.1185). A dual-primary design in which one endpoint fails is a live review risk. Standard review, no advisory committee announced.
2. Guidance shortfall at the 5 August print. FY2026 product guidance requires +13.1% in Q2–Q4 against Q1's +8.1%. Seven days away and not in this analysis.
3. Handa's 505(b)(2) capsule. Paragraph III on '473 means Handa is prepared to enter on the August 2026 composition-of-matter expiry. Exelixis argues a different-strength capsule will not be an AB-rated generic equivalent (so no automatic substitution) and says it "continues to evaluate all legal and strategic options" — i.e. it has not sued. A non-substitutable competitor still takes share.
4. Undisclosed Sun settlement entry date. Sun settled in December 2025 on terms that have not been filed. A settled entry date earlier than 2030 would reset the whole analysis, and it is unknowable from the public record.
5. Customer concentration. Five customers are 78% of revenue (Cencora 22%, McKesson 19%, CVS 15%, Accredo 12%, Optum 10%). These are the U.S. specialty channel, not a hidden distributor — but the channel is the channel.
6. 340B chargeback creep. Gross-to-net went 28.2% (FY2022) → 30.2% (Q1 FY2026) and management guides it higher. A 3% list increase net of that is thin.
7. Merck is running two phase 3 trials of zanzalintinib + belzutifan against cabozantinib. Success cannibalises the annuity earlier than the patent does. Strategically correct; economically a real near-term revenue risk.

---

3. Position construction (conditional — the book decides whether to take it)

3.1 Sizing

Inverse-volatility sizing is the framework's active protection and it does the right thing here: realised 252-day volatility of 39.3% is well above a typical large-cap, so the name sizes down automatically.

At a 39.3% volatility and the bear case at −39.7%, a position sized so that the bear case costs 1.5% of book implies a ~3.8% position. Given (a) the binary sitting on an undated docket and (b) the Valuation Criteria returning INDETERMINATE rather than PASS, a half-weight expression is the defensible maximum: ~2% of book, in equity.

3.2 Entry

Momentum is entry timing only, never whether. It is constructive but not extreme: 12-1 of +24.5% at the 60th percentile of 1,975 names, price at the 97th percentile of its own 52-week range, +26.6% over three months. A name at a 52-week high two sessions before a print that must show an acceleration is not an ideal entry.

If entering: stage it around the 5 August print rather than ahead of it. The information asymmetry is one week wide and closes on its own.

3.3 Exit and invalidation — dated and mechanical

3.4 Vehicle

• Equity — primary and recommended. $145m/day, no borrow question on a long, no spread cost.
• Options — supplement only, small size, Jan-2027 only. The one structure with real depth on both legs is the Jan-2027 $50/$60 call spread (OI 2,417 / 343). At the quoted mid prices that is roughly $9.95 − $4.69 ≈ $5.26 debit for a $10 wide spread — a poor risk/reward that requires the stock above $65.26 to profit, versus a base target of $60. On the quoted chain it does not clear its own hurdle and is not recommended.
• The December 2026 catalyst-matched expiry is not investable at scale — the $70 strike carries 2 contracts of open interest.
• Do not use the December $60/$70 call spread, which is the structure a default catalyst-trade template would generate. It cannot be filled.

---

4. What would change this analysis