Phase Space AI

Valuation

Exelixis [EXEL]

Exelixis, Inc. [EXEL] — Valuation & Criteria

2026-07-29 · spot $56.51 (2026-07-28 close) · Criteria framework 2026-07-29

Two outputs are required on every name and both are here: the implied-path test (5 years, the Valuation Criteria) and the 12-month target (what the book trades on). Neither replaces the other. Sensitivity is run over the exit multiple, never over scenario probabilities.

---

1. Verified inputs

The screen's net cash of $1,029.692m is wrong by $396.659m (27.8%). Every multiple below is recomputed on the verified figure. Where the diluted share count changes an answer materially, both are shown.

---

2. The implied-path test — Valuation Criteria: INDETERMINATE

2.1 The framework run, done exactly as specified

assets/reverse_dcf.py --spot 56.51 --shares 251.355 --net-cash 1426.4 --revenue 2375.5 --years 5 --wacc 0.10 --solve cagr --terminal-margin 0.376 --exit-multiple 20.6

• Parameter solved for: revenue CAGR.
• Parameters held fixed, named: terminal operating margin 37.6% (Exelixis' own FY2025 figure, which is what the screen used), exit multiple 20.6x EBIT, WACC 10.0%, horizon 5 years, net cash $1,426.4m, share count 251.355m.
• EV implied by today's price: $12,778m (5.38x current revenue).
• >>> THE MARKET REQUIRES: revenue CAGR of 2.3%
• Margin (demonstrated − required) = 12.9% − 2.3% = +10.6pp.
• Implied compression from today's trading multiple: today EXEL trades at 13.6x EV/EBIT; the path was solved at a 20.6x exit. That is not compression — it is an implied multiple EXPANSION of +7.0 turns, +51%. The screen's headline result depends on the terminal multiple being half again what the name trades at now.

On the diluted share count the same run gives a required CAGR of 3.7% (+9.2pp margin).

Taken at face value this is a strong PASS. It should not be taken at face value, for two reasons.

2.2 Reason one — the sensitivity, run on the parameter that decides the answer

The framework is explicit that sensitivity belongs on the exit multiple, and that terminal value carrying

60% of EV makes the reverse DCF the primary instrument. Here terminal value is ~100% of EV by construction of the tool. So the exit multiple is not a detail; it is the answer.

Required revenue CAGR by exit multiple (terminal margin fixed at 37.6%, WACC 10%, 5 years):

At the TTM margin of 39.4% the flip point moves to 12.0x exactly (required CAGR 12.9%, margin 0.0pp).

Inverted: at the demonstrated 12.93% CAGR the price implies an exit multiple of 12.6x EBIT. At the ex-milestone 13.4% it implies 12.3x. At the FY2026 guided growth of 11.0% it implies 13.7x. At zero growth it implies 23.0x.

The verdict is 100% determined by a parameter that swings the required CAGR from 30.9% to −0.8%. The screen recorded this name twice on the same night with opposite verdicts — scan_all_v2 gave valuation: PASS, margin +10.0pp at a 37.6% terminal margin; scan_final gave valuation: FAIL, margin −0.7pp after capping the terminal margin at the industry 75th percentile of 22.9%. Neither run touched the exit multiple, which is where the answer actually lives. Both are artifacts of an unexamined anchor.

2.3 Reason two — the exit multiple is UNIDENTIFIED, and that is dispositive

The rule: "An exit multiple may only be drawn from a comparator set whose growth brackets the subject's growth at the exit year. If no such comparator exists, the multiple is UNIDENTIFIED and must be declared so."

The screen's 20.6x came from a GROWTH_MATCHED set of 542 peers, matched on Exelixis' trailing 12.9% growth. Ask what Exelixis' growth rate is at the exit year, mid-2031:

• The composition-of-matter patent expires 14 August 2026.
• The Delaware judgment of 23 October 2024 puts the earliest possible FDA approval of MSN's generic CABOMETYX at 15 January 2030 — eighteen months before the exit year.
• In the exit year, cabozantinib — ~99% of revenue today — is therefore either in generic decline (growth deeply negative) or, if the Federal Circuit reverses, has already been in decline for longer.
• The only path in which the exit year shows positive growth is one in which zanzalintinib, which has zero approved revenue today, has replaced the entire franchise.

Neither of those two states is bracketed by a 542-name comparator set assembled around +12.9% trailing growth. The set has no growth dispersion in the relevant region and it was extrapolated across a discontinuity. This is the same defect the framework already documented (a 1.0–7.5% anchor set used to value 15–39% growers), inverted.

Therefore: exit multiple = UNIDENTIFIED. No peer median is substituted.

2.4 The deeper problem — the instrument itself is not identified for this name

The reverse DCF computes revenue₀ × (1+g)⁵ × terminal margin × exit multiple / (1+WACC)⁵. It assumes revenue compounds smoothly for five years. Exelixis' own legal disclosure says it does not. There is a dated, litigated, adjudicated discontinuity inside the window. Compounding TTM revenue at a constant rate through 15 January 2030 is not conservative or aggressive — it is modelling a company that does not exist.

VALUATION CRITERIA = INDETERMINATE. The required parameter cannot be solved because the exit multiple is unidentified and the five-year instrument is mis-specified for a business with a dated LOE inside the horizon. Per the framework, a missing or unidentifiable input is INDETERMINATE, never FAIL. The long-horizon answer is supplied instead by §3, which is identified.

---

3. The instrument that IS identified — LOE-aware decomposition

Same discipline as a reverse DCF: solve for what the price requires, do not assert a value. Here the unknown is not a growth rate but how much of the enterprise value is being paid for something other than cabozantinib.

Method. Value the cabozantinib franchise explicitly to its dated LOE and let it erode; subtract from EV; the residual is what today's price requires zanzalintinib and the pipeline to be worth.

Anchors, all named: - FY2026E revenue $2,575m — company guidance midpoint, not a house forecast. - Growth to LOE 4% / 6% / 8% — run as a range. FY2026 guided growth is 11.0%; the range is deliberately below it because the NET launch that carries FY2026 does not repeat. - LOE 15 January 2030 — the entered Delaware judgment. - First-full-year revenue retention post-LOE 10% / 20% / 35%, decaying 15%/yr thereafter — the ordinary range for a multi-source oral small molecule. This is an assumption, not evidence, which is why it is run as a range. - Tax 22% (guidance 21–23%), WACC 10%, 15-year horizon, mid-year discounting.

Two margin treatments, because the choice is contestable and the answer should not depend on hiding it:

Case 1 — cabozantinib carries all company opex (38.1%, the FY2026 guidance-derived margin):

Case 2 — cabozantinib standalone (COGS 4%, SG&A $600m, maintenance R&D $150m → 67.2% margin; the $750m/yr of non-maintenance R&D is charged to the pipeline instead, where it belongs):

Case 2's residual must then fund the pipeline R&D that Case 1 charged to cabozantinib: $1,945m in PV terms (4 years of $750m after tax at 10%). Netting it, the g=6%/retain=20% residual is $5,130m.

The result

Between 40% and 77% of Exelixis' $12.8bn enterprise value — call it $5.1bn to $9.9bn — is being paid for assets with zero approved revenue. The central estimates cluster at $7bn–9.5bn.

What that residual requires, at conventional multiples of peak sales for an approved oncology asset:

For scale: CABOMETYX itself did $2,113m of U.S. net product revenue in FY2025, after ten years on market and across six approved indications. Today's price requires zanzalintinib to become a franchise the size of cabozantinib or larger — on the strength of one met phase 3 endpoint, one missed phase 3 endpoint (22 June 2026: OS in patients without liver metastases, HR 0.83, p = 0.1185), and a PDUFA that has not yet occurred.

That is the honest statement of what is being underwritten. It may well be right — seven pivotal studies are running, Merck is funding two of them, and the NET and non-clear-cell RCC settings are real. But it is a late-stage biotech option funded by a melting annuity, not a compounder trading at a −10.6pp discount to its demonstrated growth. The screen's framing and the actual structure of this security are different things.

Cross-check: what if the appeal is lost?

If the Federal Circuit reverses and generics arrive in H2 2027, cabozantinib plus net cash alone is worth:

This is the floor under a total pipeline write-off, not the bear case — it assigns zanzalintinib a value of zero, which no reversal would justify (the '039 patent, expiring 2032, is still headed to trial on 2 November 2026, and '342 to 2033 was never asserted against MSN). It is reported to establish how much of the current price is not supported by the cash-generating asset.

---

4. The 12-month target — $60 (+6.7%), band $34 to $72

Built per references/valuation.md: near-term estimates plus named product-cycle events, on Exelixis' own multiple history with the percentile stated. Not a DCF. Not a peer median.

4.1 Step 1 — near-term earnings

Consensus was unavailable (Alpha Vantage quota exhausted; the API returned its rate-limit payload). The substitute is company guidance, which for FY2026 is unusually complete. This is the weakest input in the memo and is flagged as such.

Q1 FY2026 actual was $0.79, annualising to $3.16 before buyback accretion, so the FY2026 build is consistent with the print. Trailing EPS at the 12-month point (end-July 2027) ≈ $3.35.

Named product-cycle events inside the window, each dated in the catalyst calendar: Q2 results 5 Aug 2026, STELLAR-304 topline 2H 2026, zanzalintinib PDUFA 3 Dec 2026, Consolidated Litigation trial 2 Nov 2026, CAFC decision (undated).

4.2 Step 2 — the multiple, on Exelixis' own history

Daily trailing multiples computed from closing price × verified shares, less as-known net cash, over as-known TTM earnings stepped in at each filing date.

A regime change is declared and the window is cut for it. Exelixis' operating margin ran 9.3% in FY2023 and 39.4% today. Multiples from the pre-inflection era describe a different company: over the full 2021-01-04 → 2026-07-28 window (n = 1,397 sessions) the median trailing P/E is 28.8x and the maximum 122.1x, because the denominator was near zero. That window is not used.

Window used: 2024-01-02 → 2026-07-28 (n ≈ 640 sessions), the post-inflection profitable era.

(Series computed on basic shares outstanding throughout, so it is internally consistent. On the diluted count the current trailing P/E is 18.1x.)

Base multiple selected: 18.0x — above the current 17.2x, below the post-inflection median of 19.6x. The reason for not mean-reverting to the median is specific and not a haircut for its own sake: over the target horizon the runway to 15 January 2030 shortens from 42 months to 30 months, which mechanically compresses any multiple applied to a terminal-decay asset. The reason for being above spot's multiple is equally specific: the 3 December 2026 PDUFA either converts the second franchise from option to product inside the window, or does not.

4.3 The target

The band is wide because the name is genuinely binary, not because the estimate is loose. The base sits above spot, consistent with the framework's finding (item B16) that a process producing only below-spot targets is expressing a market view rather than valuing companies.

External sanity band: not available. No professional's target could be sourced within this memo's data access, and none is invented.

---

5. Criteria summary

---

6. Required disclosures on this valuation

1. The parameter solved for, and every parameter held fixed, is named in §2.1.
2. The margin demonstrated − required is +10.6pp at the screen's 20.6x anchor, 0.0pp at 12.0x, −10.7pp at 8.0x. It is reported as a function of the exit multiple because reporting the point estimate alone would misrepresent the analysis.
3. Implied compression: none. Implied EXPANSION of +7.0 turns (13.6x today → 20.6x exit, +51%).
4. Sensitivity is run over the exit multiple (§2.2). It is not run over scenario probabilities.
5. Terminal value is ~100% of EV in the reverse-DCF construction, far above the 60% threshold, which is why the reverse DCF is primary and the forward build in §4.1 is used only for the 12-month horizon.
6. The base exit multiple does not sit below every stated anchor: no base exit multiple is asserted at all, because the anchor is declared UNIDENTIFIED.