Exelixis has been a one-drug company for a decade, and its share price now assumes it will not be one after 2030. When generic cabozantinib arrives, zanzalintinib has to carry the franchise. On the evidence so far, it cannot carry that much: our $41.57 fair value is ~40% below the $69.39 price. The disagreement is about duration, how much of Cabometyx’s earning power survives its loss of exclusivity, not about whether the near-term catalysts succeed.

Cabometyx, Exelixis’s (EXEL) kinase inhibitor first approved in kidney cancer, will earn about $2.3B of US product revenue this year, and generic competition is scheduled to start between late 2030 and early 2031 (Section 2). The company’s answer is zanzalintinib, a related molecule in Ph3 in colorectal cancer (CRC), non-clear-cell kidney cancer (nccRCC) and neuroendocrine tumors (NET), with further kidney and CRC studies behind them.

That price is paid for earnings our model says peak in 2030. The stock trades at 15.5x trailing EBIT, ~19% above its own 24-month median of 13.0x, yet in our conditional case zanza’s 2032 sales replace only 23% of the $2.02B of Cabo revenue lost after 2030 (8% counting only zanza’s growth after 2030, since $0.29B of it is already in the 2030 base).

Exhibit 1Zanza’s 2032 sales equal 23% of the $2B Cabo revenue lost

Company revenue is $2.95B in 2030. We model $1.05B in 2032, about a third of the ~$2.96B in a representative sell-side long-dated forecast, which has 2032 revenue back at our 2030 level.

Company revenue by source, $B

$1B$2B$3BCabo US $2.48BZanza $0.29BRoyalties $0.18B$2.95BCabo US $0.46BZanza $0.46BRoyalties $0.13B$1.05BRepresentative sell-sidelong-dated view~$2.96B$1.91B gapours is 65% below2030Cabo’s last full year2032our model2032Street view
23% zanza vs Cabo loss
zanza’s 2032 US sales ($0.46B, conditional) against the $2.02B of Cabo US revenue lost from 2030 to 2032 (growth since 2030: 8%)
$1.0B to $1.6B Visible Alpha zanza view
Visible Alpha consensus for zanza (global, risk-adjusted, 2031 and 2035; 2025 snapshot) against $0.26B and $0.37B in our model (US, risk-adjusted)
$5B by 2033 management goal
management’s zanza goal (stated Feb 2025, US, unadjusted, basis unspecified) against $0.53B of conditional US revenue in 2033 in our model

Our figures are the conditional case, with every modeled zanza indication approved and no probability weighting; royalties are net ex-U.S. royalties. The sell-side bar is a representative long-dated forecast, unattributed (ChartMill shows $2.95B for 2032). Visible Alpha zanza figures are a 2025 snapshot, may since have moved, and are not like-for-like with our U.S.-only figures. Source: Clinaptis Analytical Model (modeled, not observed data); ChartMill and a representative sell-side long-dated forecast as of 1 Nov 2026; Visible Alpha (2025); Exelixis.

What changed in 2026 was mostly duration. On Aug 31 the Federal Circuit upheld three crystalline-salt patents against MSN, setting its earliest US launch at Jan 15, 2030. STELLAR-303 gave zanza a positive Ph3 survival result in CRC, and Q3 US product revenue grew 11.4% Y/Y. The stock’s 58% YTD return reflects both. Neither tells us how large zanza becomes. Appeal ruling

The near-term catalysts are likely to be positive, and that is the problem. We expect FDA to approve the CRC regimen at or near the Mar 3, 2027 action date, and STELLAR-304 to beat sunitinib in Q1 2027 without establishing an edge over existing nccRCC regimens. Both outcomes sit inside our model, and neither moves value much (Sections 4–5). The ’039 patent trial, which began Nov 2, is downside protection rather than upside (Section 2).

The first pivotal readout, STELLAR-304, comes in Q1 2027. Most of the other events arrive between 2029 and 2032, as Cabo erosion begins, and we credit little for them (Exhibit 2).

1. Cabometyx built the company. Its expansion record sets the base rate.

Cabometyx made Exelixis a profitable oncology company. Repeating that success has proved harder. The company spent its first two decades as a partnered kinase-discovery platform, and the stock traded at $1.26 in Dec 2014. Cabo became a standard of care in RCC, including first line in combination with nivolumab (CheckMate 9ER), and added labels in HCC, thyroid cancer and NET. What matters for valuation is what happened next, as the company tried to carry its kinase-inhibitor franchise into other tumors and treatment settings.

Those attempts have produced a succession of disappointing survival results. Cabo plus atezolizumab failed to improve overall survival in HCC (COSMIC-312), in post-IO NSCLC (CONTACT-01, 10.7 versus 10.5 months, HR 0.88) and in post-IO RCC (CONTACT-03). CONTACT-02 in mCRPC missed its final OS analysis, and Ipsen dropped the indication in Europe. COSMIC-313 showed that even a PFS win need not reach survival: adding Cabo to nivolumab/ipilimumab in first-line RCC met its PFS endpoint, but final OS was 41.9 versus 42.0 months (HR 1.02), and the company did not file. Among these expansion attempts, STELLAR-303 is the first randomized TKI-plus-IO success, and a modest one: a 1.5-month median OS difference against regorafenib. That record informs our conservative treatment of two zanzalintinib expansion trials: LITESPARK-034 in post-IO/TKI clear-cell RCC carries a 30% working probability, while LITESPARK-033 in post-adjuvant recurrent RCC receives no separately underwritten value. COSMIC-313

The board has also changed how it allocates capital, with consequences for both the downside case and the reported earnings. After Farallon’s 2023 proxy contest, two activist nominees joined the board and it formed a Capital Allocation Committee; in Jan 2024 the company cut ~13% of headcount and added a $450M repurchase to the $550M completed in 2023. The post-LOE cost response we model in Section 6 is therefore behavior this board has already shown under pressure. The buybacks carry a less flattering implication. About 40% of Q3’s EPS increase, from $0.69 to $0.85, came from an 8.2% smaller share count, and that should not be mistaken for franchise momentum. Restructuring 8-K

2. Cabo can finance the transition. Competition may shorten its runway before the patents do.

Cabo’s problem is duration, not growth. Q3 US product revenue rose 11.4% to $604.9M on 6% unit growth and 5% net price. Momentum is softening at the margin: management trimmed the revenue midpoint by $50M in August, and guidance implies Q4 below Q3. But the end date matters more to value than the slope. Cabo growth 3 points a year above ours through 2030 adds $1.75/sh; extrapolating that growth past exclusivity is the error to avoid. Q3 release

The litigation decides when erosion starts, not whether it does. We model US Cabo revenue reaching ~$2.49B in 2029, with erosion starting in Dec 2030 under Azurity’s licensed entry; settled ANDA filers hold Jan 1, 2031 dates. MSN’s ANDA can launch from Jan 15, 2030 unless Exelixis prevails on the ’039 low-impurity patent, which went to trial on Nov 2. A win protects our working case but cannot extend exclusivity past the Azurity and settlement dates. A loss brings erosion forward by about 10.5 months and costs about $3.6/sh on our Cabo component DCF, before any cost response. The timing of a ruling is uncertain, and an appeal is possible. Q3 10-Q

Exhibit 2Zanza’s nearest RCC readout comes almost four years before Cabo’s US exclusivity ends
Dated by company, FDA or court scheduleRegistry estimate, not a guaranteed dateLarger, bold: higher thesis weightSmaller, faded: lowerExposed if the ’039 patent is lost
2027 2028 2029 2030 2031 2032 2026 Cabo US protection Protected on every route Feb 2032: ’039 patent expires Nov 2, 2026: ’039 patent trial began Jan 15, 2030: earliest generic entry (MSN) if ’039 is lost Dec 1, 2030: Azurity licensed entry Our working-case erosion start Jan 1, 2031: settled ANDAs Zanza RCC trials Q1 2027: STELLAR-304 topline, first-line nccRCC (company guidance) Nov 2030: LITESPARK-034, 2L+ RCC; our working probability 30% Feb 2032: LITESPARK-033, 1L vs Cabo after adjuvant IO; no separate value About 3.8 years from STELLAR-304 topline to base-case generic entry Zanza CRC FDA Mar 3, 2027: CRC FDA action date Zanza NET trial Jun 2029: STELLAR-311 NET US generic entry range

Registry primary-completion estimates for STELLAR-311, LITESPARK-034 and LITESPARK-033 are not guaranteed readout dates; STELLAR-304 timing is company guidance. The 30% and no-value treatments of LITESPARK-034 and -033 are Clinaptis working assumptions; “higher thesis weight” is Clinaptis judgment.
Source: Exelixis Q3 2026 10-Q and Aug 31, 2026 release; FDA; Handa Pharmaceuticals; ClinicalTrials.gov; Clinaptis analysis.

The more immediate threat comes from Merck, and it arrives before LOE. In LITESPARK-011, belzutifan plus lenvatinib beat Cabo head-to-head after PD-(L)1 therapy on progression-free survival (14.8 versus 10.7 months, HR 0.70), and FDA approved it on Sep 24. The trial targets Cabo’s strongest position: on the triangulation in our companion brief, Cabo holds roughly a third of second-line ccRCC starts, more than any other single brand. Overall survival was not statistically significant (HR 0.85), which argues for erosion rather than collapse. The case for a muted impact rests on tolerability and a revenue mix weighted to first line, where Cabo looks better protected after LITESPARK-012 missed on Apr 21. FDA approval

That argument holds for first line but understates the second-line exposure. If second line and later accounts for 20–35% of Cabo’s ~$2.1B of 2028 RCC revenue (the company does not disclose the split; 25% is our working case, and we apply the brief’s second-line share path to the whole slice), the brief’s base-case decline in share from 37% to 27% would put ~$110–190M of annual revenue at risk, worth ~$0.9–1.6/sh in our SOTP (Exhibit A8). Our base forecast includes neither reduction, which makes this incremental downside and the largest competitive risk outside our forecast. The value is modest. The point is that Cabo’s protected base can start eroding before the patent cliff.

The pool is also smaller than it was. In a US Flatiron cohort of metastatic ccRCC, 57.9% of patients starting first-line therapy before Apr 2018 went on to a second line; after the IO combinations arrived, 37.6% did. Defending share in a smaller pool is not the same as defending revenue. We treat that contraction as already in the revenue base; a further 2% a year, driven by continued first-line gains, would add only ~$0.2/sh of downside.

Handa is a slower risk, because its product cannot be swapped at the pharmacy. OMCAZIO, a cabozantinib capsule taken with or without food, received final FDA approval on Sep 11, but its boxed warning says it is not substitutable milligram-for-milligram with other cabozantinib products. Switching needs a new prescription at a different strength. Handa has not announced US launch timing. FDA label

Cash is not the constraint; the return on spending is. Under our conditional operating case, which assumes modeled zanza launches without probability weighting, we project $4.13B of cumulative FCFF in 2027–30, alongside $1.40B of cash and securities. Q3 SG&A rose 25%, compared with 11% product-revenue growth. Exelixis has the balance sheet to fund its next franchise. Whether that spending earns comparable economics is the investment question that follows.

Exhibit 3Zanza grows into a much smaller revenue base after Cabo erosion
0$1B$2B$3B2.5220262.6320272.7520282.8720292.9520301.9720311.0520320.8420330.8220340.8220350.822036Cabometyx USNet collaborationZanzalintinib, conditionalLabels: total, $B
View data
$M2026E2028E2030E2032E2036E
Cabo US product revenue2,3252,4872,478456113
Net collaboration revenue20019717612766
Zanza, conditional—66292464636
Company revenue, conditional2,5252,7502,9461,048816
Company EBIT margin, conditional38.9%42.3%42.7%3.3%40.0%

Conditional case; zanza is not probability-weighted here. Cabo includes the small COMETRIQ contribution; net collaboration is not identical to gross royalties. 2030 Cabo revenue (~$2.48B) sits slightly below 2029 (~$2.49B) because it carries one month of erosion from Dec 1, 2030 and modeled ncc displacement. Margins include modeled cost responses, not management guidance.
Source: Clinaptis Analytical Model (modeled, not observed data).

3. Zanza has to earn a premium over generic cabozantinib.

Zanzalintinib is, by design, a shorter-acting cabozantinib. Exelixis kept the same target profile (VEGFR, MET, AXL, MER) and cut the exposure: Ph1 dose escalation showed a half-life of 16–22 hours, against a predicted terminal half-life of ~99 hours on the Cabometyx label. A drug that clears in days rather than weeks is a plausible advantage when toxicity forces dose interruptions, particularly alongside a checkpoint inhibitor. It is not evidence of a better drug. XL092 data

The difficulty is that the shared biology will soon be cheap. Cabo’s efficacy in second-line RCC is reproducible: its control-arm PFS was 10.8 months in CONTACT-03 and 10.7 in LITESPARK-011. What zanza adds is pharmacokinetic, and a shorter half-life is a development rationale, not a commercial moat.

Nor has the tolerability case been tested. No randomized trial compares zanza with cabozantinib. STELLAR-303 reported grade 3 or worse treatment-related adverse events in 60% of patients on zanza plus atezolizumab versus 37% on regorafenib, although any-grade hand-foot reactions were lower (16% versus 50%). With a different comparator, tumor type and checkpoint-inhibitor partner from Cabo’s trials, it can neither confirm nor refute the half-life argument.

That leaves a payer question zanza cannot yet answer. After 2030, wherever the indications overlap, zanza will face a cheaper drug built around much the same target biology. Its shorter half-life may offer a tolerability advantage, but without comparative evidence, that is a difficult basis for defending branded pricing. Our model therefore assigns no clear-cell RCC revenue from spillover of a nccRCC label, and values clear-cell use only through LITESPARK-034, where zanza is paired with belzutifan and tested in its own randomized trial.

The successor franchise is a set of labels, not a demonstrated better molecule. Generic cabozantinib will be a formidable comparator wherever the indications overlap, including nccRCC, where Cabo’s RCC label is not restricted by histology and PAPMET provided randomized Phase 2 data (Section 5). Zanza has a more defensible opportunity in later-line CRC, where Cabo holds no label, and in partnered regimens such as LITESPARK-034. The expansion trials in Section 7 are therefore the real test.

4. CRC approval is likely. Its place in the sequence decides the revenue.

We expect approval. STELLAR-303 met its overall-survival endpoint against regorafenib: 10.9 versus 9.4 months (HR 0.80) in 901 patients with non-MSI-H metastatic CRC. A separately tested non-liver-metastases hypothesis missed at final analysis, but that does not undermine the intent-to-treat result. Patients with liver metastases did at least as well (HR 0.78, nominal and post hoc), which gives little reason to expect a label restricted to those without liver disease. FDA moved the action date to Mar 3, 2027 after a major amendment. We assign 80% to a commercially usable label by end-2028, covering approval, label breadth and launch timing: 65% for a 2027 launch and 15% for a 2028 delay. STELLAR-303

STELLAR-303 establishes that the regimen improves survival against regorafenib. It does not establish why physicians should prescribe it ahead of TAS-102 plus bevacizumab. The trial excluded patients previously treated with TAS-102, regorafenib or PD-(L)1 therapy, so its strongest evidence applies before that regimen, which many physicians already prefer: in SUNLIGHT it improved survival to 10.8 from 7.5 months against TAS-102 alone (HR 0.61). Cross-trial medians cannot rank the two. The label may permit third-line use; prescribing practice may push zanza to fourth line, where patient flow is smaller, the evidence extrapolates beyond the pivotal population and fruquintinib already has randomized data after TAS-102 and/or regorafenib. SUNLIGHT

Exhibit 4CRC efficacy is established against regorafenib; exposure and sequencing limit the commercial extrapolation
STELLAR-303Zanza + atezolizumab10.9Regorafenib9.4SUNLIGHTTAS-102 + bevacizumab10.8TAS-1027.5FRESCO-2Fruquintinib7.4Placebo4.8CORRECTRegorafenib6.4Placebo5.0
STELLAR-303Zanza + atezolizumabRegorafenibRead
ITT median OS10.9 months9.4 monthsHR 0.80; positive
Final NLM median OS15.9 months12.7 monthsHR 0.83; failed
Objective response~4%~1%Disease control, not shrinkage
Grade ≥3 treatment-related AEs60%37%Meaningful burden
Median treatment duration3.4 months2.2 monthsNot mean paid months

Sources: Primary paper, corrected appendix, final NLM update. Final NLM analysis (Jun 2026) shown; it supersedes the paper’s interim NLM estimate of HR 0.79 (0.61–1.03). PFS and response are descriptive; no cross-trial ranking is implied. Chart: medians as published; blue is the active regimen, grey the comparator; cross-trial comparison is descriptive only. SUNLIGHT, FRESCO-2 and CORRECT from their primary publications.

The regimen’s burden gives physicians reasons to wait. Median treatment lasted 3.4 months, toxicity ran higher than on regorafenib (Section 3), and atezolizumab adds infusion logistics for a contribution that is uncertain: in the small STELLAR-001 RAS wild-type cohort, OS did not separate (Exhibit A9). Recent GI key-opinion-leader commentary favored later use and questioned both the full-dose regorafenib control and atezolizumab’s role: selected opinion, not an adoption survey, but pointing the same way.

Our $288M is not a conservative floor. It requires both third-line adoption and a distinct later-line patient pool. At maturity we model ~3,199 unique annual starts, 4.5 paid months and $20,000 net per paid month: $288M of conditional sales, and $230M after the 80% weight, which is the figure that enters our valuation. Third line supplies 1,620 starts, the base case of our CRC landscape brief (9% of ~18,000 non-MSI-H third-line patients; bear 5%, bull 16%). Fourth and fifth line supply 1,579 from patients never exposed to zanza.

The two pools use different denominators, which flatters the total modestly. Later-line starts follow the company-anchored 22,080 cohort through observed line-to-line attrition, after removing anyone who has already started zanza. No patient is counted twice, but the cohort feeding later lines is ~23% larger than the third-line base; on a common 18,000 base, conditional sales are ~$260M (Exhibit A9).

Three assumptions carry the figure (Exhibit 5): third-line uptake ($228M to $393M across the brief’s bear and bull cases), later-line capture, which supplies half the starts ($146M without it), and paid duration ($218M at the trial’s 3.4-month median). Price sits on top of all three. The $20,000 is our assumption, not a disclosed price: a ~33% premium to the $15,000 we use for Cabo as the counterfactual, and revenue scales in proportion to it (Exhibit A10). The premium does not contradict Section 3. In later-line CRC, zanza competes with branded regimens such as fruquintinib, not with generic cabozantinib, which has no CRC label.

The nearest commercial benchmark frames the figure without validating it. Takeda reported $203M of Fruzaqla net sales in the first nine months of 2024, its launch year after US approval in Nov 2023 and before most ex-China launches. That is launch-year revenue across Takeda’s territories, not a mature US figure.

The 80% weight covers approval, label and timing, not adoption; adoption risk sits in the share assumptions, and we apply no further Phase 3 discount. The evidence supports approval. It does not yet support more than the brief’s base-case share of third-line patients. CRC materials

Exhibit 5Zanza CRC: two patient pathways give 3,199 annual starts and $288M of US sales

Path A: third line

~18,000

non-MSI-H patients, Clinaptis CRC landscape brief

9% assumed third-line penetration (bear 5%, bull 16%)

1,620 starts

Path B: fourth and fifth line

22,080

non-MSI-H patients, from the company-anchored 23,000 reference; ~23% larger than Path A (dashed line: 18,000)

Line-to-line attrition, later-line capture, no earlier zanza start

1,579 starts

3,199 unique annual starts

1,620 + 1,579. Two distinct cohorts; no patient is counted as a start twice. On a common 18,000 base: 2,884 starts, ~$260M conditional.

3,199unique annual starts
×
4.5mean paid months
×
$20,000net revenue per paid month
=
$288Mconditional on a usable label
$230Mafter 80% PoS

How far the levers move it: conditional US sales by third-line uptake, $M

Sales at the trial median, 3.4 paid monthsAdded by our 4.5 paid months (bar end = our assumption)
$100M$200M$300M$400MBear5% uptake900 third-line starts · 2,535 in all$172M$228MBase9% uptake1,620 third-line starts · 3,199 in all$218M$288MBull16% uptake2,880 third-line starts · 4,362 in all$297M$393MStressNo later linesbrief count read as all lines · 1,620 starts$110M$146M

Underlined inputs are the two levers. Third-line starts adopt the Clinaptis CRC landscape brief (base 9% of ~18,000, bear 5%, bull 16%). Fourth- and fifth-line starts are built on the company-anchored 22,080 and come from patients never exposed to zanza; they move slightly with third-line uptake. Aligning both pathways to 18,000 gives 2,884 starts and ~$260M conditional (Exhibit A9). The stress case reads the brief’s count as covering all lines, so starts are 1,620 and conditional sales $146M. Each share, capture rate, paid duration and price is an analyst assumption; the eligibility screens behind the later-line cohort are analyst estimates, not validated screening data (Exhibit A9). Paid months can run below the 3.4-month trial median duration because of dose holds and dispensing gaps.
Source: Clinaptis Analytical Model and CRC patient funnel (modeled, not observed data); Exelixis CRC materials; STELLAR-303 (Lancet 2025); Clinaptis CRC landscape brief.

5. STELLAR-304 can beat sunitinib without winning the market.

STELLAR-304 could give zanzalintinib its first RCC label without showing that it is better than what clinicians already use. The Phase 3 trial, due to read out in Q1 2027, randomized 317 patients (design target 291) 2:1 to zanza plus nivolumab or sunitinib in papillary, unclassified and translocation-associated RCC, with blinded-review PFS and response rate as dual primary endpoints. It is the next hard test for the molecule, and its comparator sets a low bar. Registry

Sunitinib is a comparator regulators can accept, but it is not the bar clinicians use. In papillary disease, PAPMET showed cabozantinib ahead of sunitinib (PFS 9.0 versus 5.6 months), and single-arm studies of cabozantinib/nivolumab and lenvatinib/pembrolizumab have reported response rates of 48% and 51% (KEYNOTE-B61) in related, though differing, populations. Zanza/nivolumab’s widely quoted 63% response rate comes from 40 untreated clear-cell patients: too few, the wrong histology and single-arm. We found no verified early dataset in STELLAR-304’s histologies, so the evidence for this molecule in this population arrives with the trial itself. PAPMET

We lean toward a statistical win, but the evidence supports the direction of benefit more than its size. PAPMET is the randomized anchor: a VEGF/MET inhibitor beat sunitinib in papillary disease (PFS HR 0.60, with a wide confidence interval). Single-arm IO/TKI activity is supportive but cannot establish a randomized treatment effect, and clear-cell trials overstate what to expect in these histologies. Nor is the move to a Q1 2027 readout a reliable efficacy signal. Without knowing event accrual and population characteristics, slower accrual is as consistent with a stronger control arm or a lower-risk population as with a working drug.

Response rate is the sturdier endpoint. It is tested in the first 222 patients at a two-sided 0.5% level, and a zanza/nivolumab rate near the single-arm 48–51% would clear it comfortably. At 30%, the answer depends on the control rate assumed: power is ~99% if sunitinib repeats PAPMET’s 4%, ~57% at 12% and ~17% at 18%. The 18% is ASPEN’s overall sunitinib response rate (9 of 51, including 24% in papillary disease), but only three of those nine responses were later confirmed, so it illustrates how variable historical control rates are rather than estimating STELLAR-304’s confirmed control rate. PFS is less certain. The event target is not published; if the analysis includes roughly 190–230 events, significance needs an observed hazard ratio of about 0.74–0.76, and curves that separate only after two or three months would materially cut the chance of reaching it. A response-led win may also carry less weight with FDA in first-line RCC than a PFS win. We attach no probability to the statistical outcome.

A positive STELLAR-304 would establish randomized Phase 3 evidence against sunitinib. It would not establish superiority over the cabozantinib-based or IO/TKI regimens clinicians already use, particularly in papillary disease. It would still give zanza a broadly labeled, Phase 3-supported combination in these histologies, and a first label matters. A win would also build prescriber familiarity inside Exelixis’s core RCC franchise and make later clear-cell use more plausible, probably two to three years out. We value that only through LITESPARK-034 (Section 3); it would pay off mostly after generic entry, by converting lost Cabo patients into retained ones.

How we will read the result. The points below are analyst judgment, not trial thresholds.

How we will read STELLAR-304
What the trial shows
0.90.80.70.60.5not significantsignificant, ≈0.75Observed PFS hazard ratio (lower is better)
Also neededWhat it means for the model
Statistical win, modest efficacy
or response-ledLabel likely; adoption at or below our working case
Supports the working case
response ~45–50%Commercially credibleworking case: ~$198M conditional US sales
Upside
durable responses, clean tolerabilityUpside case (Exhibit A10)
Cap, whatever the headline
inconsistent histology effects, weak OS direction or poor tolerabilityCaps penetration even if the trial is positive

Analyst judgment, not trial thresholds. The ≈0.75 line is the observed hazard ratio needed for significance on the original design arithmetic (0.74–0.76; event count and any amended analysis plan are unpublished). Response-rate and PFS are dual primary endpoints; a response-led win may carry less weight with FDA in first-line RCC than a PFS win. Working-case adoption carries the 50% commercial weight, applied once.
Source: Clinaptis analysis; Pal et al., Future Oncol 2025; ClinicalTrials.gov NCT05678673.

The commercial stake is modest. We model ~$198M of conditional US sales at maturity (Exhibit A2) and apply a 50% working weight, giving ~$99M risk-adjusted; we test 40% and 65% in Exhibit A10. The 50% is our judgment of the chance that the data support the modeled share, not of a statistical win, and it is applied once: the outcome shares describe adoption, not probability. In valuation terms, nccRCC carries ~$177M of risk-adjusted NPV, about $0.70/sh (Exhibit A1). A miss with no approvable result would remove it, offset only by any development costs avoided. A result that supports the working case could justify removing the commercial weight, worth roughly another $0.70/sh on unchanged cash flows, but only if the data also support the modeled share, duration and tolerability. Either way the swing is about $0.70/sh, small against the valuation gap. Published nccRCC forecasts are several times ours; in our reading the difference lies less in the trial outcome than in the share and treatment duration zanza/nivolumab can hold against cabozantinib-based and other IO/TKI regimens.

6. The gap is about duration, not overhead or probability.

At $69.39, the shares price Exelixis as though its current earnings outlast Cabo’s exclusivity. Our selected-asset SOTP is $31.62/sh, so the market pays $37.77/sh, about $9.6B, for value we do not model (Exhibit 6; Exhibit A3). Our whole-company DCF, which funds the full resized budget and values the continuing company, is lower still at $24.84.

Exhibit 6Our SOTP is $31.62; the market pays $37.77 more
$0$20$40$60+24.89+2.91+2.63(1.20)(6.32)+1.70+1.48+5.53$31.62+37.77$69.39Cabo USEx-USroyaltiesZanzaassetsPre-launchcostsSharedcosts, 50%Tax shieldsand tailSBCadd-backNet cashSOTP valueUnexplainedby SOTPShare price30 Oct 2026

Value per diluted share (253.0M valuation shares); 10.5% discount rate. Cabo US is 79% of modeled equity value ($7,999M); stock compensation is added back ($1.48 a share). Rounding affects displayed sums; the unexplained amount is $9,557M on a diluted basis (basic market EV $15.53B, diluted reference EV $16.16B). The unexplained amount is not uniquely identifiable pipeline NPV or a market-implied success probability.
Source: Clinaptis Analytical Model (modeled, not observed data); price as of 30 Oct 2026.

Neither obvious bridge comes close. Removing all retained shared costs lifts the SOTP only to $36.00; taking the three zanza indications to 100% probability adds ~$1.7/sh at unchanged forecasts. Lower costs are not the missing ingredient either. The model already assumes a severe post-exclusivity restructuring, and even with every modeled zanza indication succeeding, EBIT falls from $1.26B in 2030 to $34M in 2032 before recovering to ~$326M by 2036 (Exhibit 3). The ~$16.2B EV is ~50x that recovered figure. Closing the gap needs higher sales for longer. FY27 earnings are real, but capitalizing them as though they were durable is the central valuation risk.

Duration bridges most of the gap. Replacing our operating forecast with a representative sell-side long-dated operating forecast, and holding everything else including the 10.5% discount rate, lifts the DCF from $24.84 to $55.09, about two-thirds of the way from our DCF to the price (Exhibit 7). Revenue does the work. That forecast’s revenue on our cost dollars gives $59.09, an upper bound because it pairs lean costs with higher revenue (~$53 if 30% of the extra revenue went to cost); its costs on our revenue give $19.17 (Exhibit A11). The rest is convention: on that forecast the price implies a discount rate near 8.1%, and alternative terminal treatments give $51 to $66, still short of the price.

The exit multiple makes the same point. We value post-2029 cash flows at about 2.2x 2029 cash EBITDA; the price needs about 14x on the same basis. The disagreement is not whether Cabo generates cash. It is how much of that earning power survives its loss of exclusivity.

Exhibit 7A longer-dated forecast closes $30 of the $45 gap to the price
$0$20$40$60$24.84+$30.2568% of gap+$14.3032% of gap$69.39Our DCFcash value, SBC added backLonger-datedforecastsell-side P&LMarket valuationbasisrate and terminal valueShare price30 Oct 2026

Value per diluted share; modeled, not observed. Longer-dated forecast: a representative sell-side operating forecast run through our model at 10.5% ($55.09); its revenue alone gives $59.09, its costs alone $19.17 (Exhibit A11). Market valuation basis: the remainder, equal to discounting at about 8.1%; alternative terminal treatments give $51 to $66.
Source: Clinaptis Analytical Model; FactSet; ChartMill; unattributed sell-side forecast; price at 30 Oct 2026.

Our fair value is $41.57/sh (range $38.68 to $51.07; Exhibit 8). It blends three intrinsic and two market-multiple methods, an editorial convention rather than an expected value, and the market methods pull it up: EV/EBITDA gives $57 because it capitalizes FY27 earnings that precede the cliff, and only the NBIX-adjusted high cases reach the price. No weighting of these methods reaches $69 on central cases (Exhibit A5). Even with that concession to market multiples, the shares carry ~40% downside.

Exhibit 8Hybrid fair value is $42, about 40% below the price
$15$25$35$45$55$65$75WeightIntrinsic / cash flowrNPV / SOTPCabo, royalties, PoS-weighted zanza; 9–12% rate$30$3325%Company DCFResized budgets, continuing terminal; 9–12% rate$24$2715%DCF to 2029 + exit multiple6x / 8x / 10x 2029 cash EBITDA$40$5520%Market multiplesOncology EV / sales2.3x / 2.6x / 5.3x FY27 revenue$29$6110%EV / EBITDA, FY27INCY-adjusted / own median / NBIX-adjusted$56$7230%Memo, not weightedOncology P / sales (memo)3.8x / 4.4x / 5.3x FY27 revenue$40$55EV / EBIT, FY27 (memo)Same peer anchors as EV/EBITDA$57$75Price $69.39Hybrid $41.57
View data
MethodLowCentralHighWeight
INTRINSIC / CASH FLOW
rNPV / SOTP$30.41$31.62$32.9425%
Company DCF$23.53$24.84$26.5515%
DCF to 2029 + exit multiple$39.60$47.40$55.2120%
MARKET MULTIPLES
Oncology EV / sales$29.47$32.60$60.7010%
EV / EBITDA, FY27$55.59$57.34$72.4830%
Hybrid fair value$38.68$41.57$51.07100%

Weights are a Clinaptis editorial choice. DCF and rNPV ranges use 12% and 9% discount rates (central 10.5%). Oncology multiples are the 25th percentile, median and 75th percentile of 11 established commercial names (ex-EXEL); EV/EBITDA and EV/EBIT use INCY-adjusted, own-median and NBIX-adjusted FY27 anchors on a GAAP basis. Oncology sales multiples ignore EXEL’s ~40% EBIT margin and act as a floor (Exhibit A4). Exit-multiple figures in the text are on cash EBITDA with stock compensation added back (implied 2.2x; price requires ~14x); on GAAP EBITDA they are 2.45x and 15.1x. Modeled; not observed data.
Source: Clinaptis Analytical Model; company filings; price at 30 Oct 2026.

7. The price implies a pipeline Exelixis has not shown.

Translated into sales, the residual is far larger than anything Exelixis has in the clinic. Our three modeled zanza indications total $379M of risk-adjusted US sales at maturity; Exhibit 9 shows what else would have to exist.

Exhibit 9Closing the $9.6B residual needs ~$6B of extra peak sales
First year of extra sales45% cash margin55% cash margin65% cash margin
(model margin)
2029$7.4B$6.0B$5.0B
2030$8.8B$7.1B$6.0B
2031$10.6B$8.5B$7.2B
2032$12.9B$10.5B$8.8B

Extra sales follow the modeled indication ramp, a 22% tax rate, 1% D&A, 2% capex, 10% working capital and erosion after 2040; the shared-cost burden is as in the SOTP. Illustrative, not a forecast.
Source: Clinaptis Analytical Model (modeled, not observed data); price as of 30 Oct 2026.

The programs that could change the valuation arrive as the franchise they would replace ends. Only three pivotal events fall before 2030, and we credit two: STELLAR-304 (Q1 2027) and the CRC action date (Mar 2027). The third, STELLAR-311 in NET, has a registry estimate near mid-2029 and competes with Cabo’s own label. The rest read out as generic cabozantinib arrives, on registry estimates: LITESPARK-034 in post-IO/TKI ccRCC (Nov 2030, at a 30% working probability) and LITESPARK-033 in post-adjuvant disease (Feb 2032). Both pair zanza with a partner drug in settings where standards are already established, closer to the combination work where Exelixis’s record is weakest (Section 1) than to its monotherapy successes, although belzutifan/lenvatinib’s win in LITESPARK-011 shows that the HIF-2α–TKI pairing can work. Spreading one molecule across tumors diversifies indication risk while leaving mechanism, tolerability and partner risk shared.

The comparators are also moving. By 2030 a belzutifan-plus-placebo control will look dated beside belzutifan/lenvatinib, making LITESPARK-034 easier to win and harder to translate into practice. After FDA’s Jun 12, 2026 approval of adjuvant belzutifan plus pembrolizumab (LITESPARK-022), a growing share of LITESPARK-033’s patients will already have received belzutifan. Any value for these trials has to be judged against the sequence patients will face at launch, not today’s treatment flow.

Even generous credit for the identifiable programs covers only ~15% of the residual (Exhibit 10). Published expectations are larger than ours, but not like-for-like: Visible Alpha consensus (2025 snapshot) put zanza at ~$1.6B of global risk-adjusted peak sales in 2035, against our $379M of US sales at maturity; later data may have raised it, and it includes ex-US sales and indications we do not value. The price therefore rests on one of two bets: that Exelixis’s revenue holds up after Cabo’s exclusivity far better than we model, or that zanza becomes several times the franchise any disclosed trial supports. The evidence does not yet support either.

Exhibit 10Generous credit for identifiable programs covers 15% of the residual

Five programs we do not value add $4.12 a share at generous odds and peaks. Bull cases on our three assets add $1.49. $32.16 a share is still unexplained.

Program and timing Generous credit (illustrative) Prob.-weighted peak Value, $ per share
Programs we do not value
STELLAR-311, NET vs everolimusPh2/3 enrolling; registry primary completion Jun 2029 55% × $500M US peak $275M
$1.74
STELLAR-316, MRD+ resected CRCPh3 started Aug 2026, DFS endpoint; readout after 2030 25% × $800M $200M
$0.86
Ex-US zanza (partners)Royalty and milestone economics not modeled Royalty stream $150M
$0.79
LITESPARK-033, post-adjuvant RCCMerck-sponsored Ph3; registry primary completion Feb 2032 40% × $300M $120M
$0.52
Bladder, mCRPC, NSCLC, meningiomaPh1b/2 cohorts, started Apr to Oct 2026; no registrational plan Aggregate 5% × $1B $50M
$0.22
Dated readouts, generous credit $795M
$4.12
Our three assets
Bull case on CRC, nccRCC and ccRCCCRC bull starts; nccRCC PoS 75%; ccRCC PoS 50%
$1.49
Total, as a share of the $37.77 residual
$5.61 (15%)

Generous credit is a Clinaptis illustrative upper bound, not a forecast or a model output: no peak sales are company-guided. Value uses the reverse-grid conversion at a 65% cash margin, 10.5% discount rate and first sales in 2030 (NET), 2031 (ex-US) or 2032 (others). HNSCC (STELLAR-305) is excluded: the company decided not to start its Ph3 in 2025. Bars share one scale; rows are as reported and may not sum because of rounding. The $5.61 is an illustrative stress test that uses a simplified reverse-grid conversion rather than indication-level cash-flow models; it is not an underwritten SOTP. NET partly replaces Cabo’s own label, so generous credit for it overstates incremental value.

Source: Exelixis Q3 2026 release; ClinicalTrials.gov; Clinaptis Analytical Model (modeled, not observed data).

8. What would change our view

Upside. We would raise our valuation on three kinds of evidence: CRC launch data showing third-line use above the 9% in our base case and paid duration at or above our 4.5-month assumption (Section 4); STELLAR-304 efficacy and tolerability competitive with modern nccRCC combinations, such as a PFS hazard ratio at or below ~0.60 with durable responses (Section 5); or evidence in NET or post-adjuvant RCC that arrives early enough, and is large enough, to be valued before exclusivity ends. Each would change starts, duration or the indications we value, not just a probability. What would prove us wrong is evidence of scale, not of approval: even all three, at our bull-case levels, add only a few dollars a share (Exhibit 10).

Downside. We would lower it if CRC settles into a fourth-line niche with short paid exposure; if STELLAR-304 misses (~$0.70/sh) or beats sunitinib without a competitive modern profile; or if Cabo erodes earlier than assumed, through a loss on the ’039 patent (MSN entry from Jan 15, 2030; ~$3.6/sh before any cost response) or faster second-line share loss to belzutifan/lenvatinib (~$1.2/sh).

How we differ from the sell-side. We diverge from the first estimate year, and by 2032 our $1.05B of revenue is about a third of a representative sell-side long-dated forecast (~$2.96B). The difference is revenue, not costs: through 2030 our operating costs are within 13% of FactSet and ChartMill, and that forecast’s cost dollars on our revenue would lower our DCF, not raise it (Section 6). Near-term estimates and the multiples built on them describe Cabo before the cliff; our valuation rests on what remains after it (Exhibit 11).

Exhibit 11Revenue durability drives most of our gap to published forecasts
Revenue, $B0$1B$2B$3B20272029203120332035ClinaptisFactSetChartMillRepresentative sell-side
Total operating costs, $B0$0.5B$1.0B$1.5B$2.0B20272029203120332035ClinaptisFactSetChartMillRepresentative sell-side

Ours is the conditional operating case (all modeled zanza indications approved; restructuring in 2032; GAAP, including stock compensation). Total operating costs are revenue less EBIT. FactSet and ChartMill are as published; ChartMill 2032 implied costs (~$1.57B) exceed that forecast (~$1.12B), and the sources differ on EBIT definition (long-dated sell-side EBIT is non-GAAP, before stock compensation), so cost lines are not perfectly like-for-like. Holding our cost dollars constant, substituting that forecast’s revenue alone lifts our company DCF to $59.09 (Exhibit A11). Hollow circles are a representative sell-side long-dated forecast, unattributed.
Source: Clinaptis Analytical Model (modeled, not observed data); FactSet and ChartMill consensus as of 1 Nov 2026; a representative sell-side long-dated forecast (unattributed).

Timing. No scheduled near-term catalyst bridges the valuation gap on our current commercial assumptions. Investors have so far looked through exclusivity loss, citing buybacks and margin growth, and nothing in our work says that cannot continue. As 2031 and 2032 enter forward estimates, post-2030 earnings become harder to leave out. That says when the question becomes unavoidable, not where the stock goes.

Exhibit 12Three dated events can move the underwrite by mid-2027

The patent trial, STELLAR-304 and the CRC FDA decision change probabilities, timing or erosion date. The other four items are monitoring: they change paid exposure and earnings quality, not whether a gate is passed.

Nov 2, 2026Trial began
MSN low-impurity patent caseSubsequent rulings and appeals
About $3.6/sh at stake, before any cost responseA loss permits MSN entry from Jan 15, 2030, about 10.5 months before our working case. A win protects the working case. A ruling would likely follow the trial by months, with appeal possible.
Feb 2027Expected month
FY27 guidanceCabo volume, NET contribution, gross-to-net, operating budgets
Funding capacity and recurring earnings quality
Q1 2027
STELLAR-304 readoutPFS, ORR, histologies, OS direction, safety; final statistical plan
ncc PoS and conditional share and duration
Mar 3, 2027FDA action date
CRC FDA actionExact label, safety language and obligations
Timing and usable approval probability
After CRC launchIf approved
CRC launch uptakeTreatment line, prior TAS-102, unique starts, refills and interruptions
CRC penetration and paid exposure
H1 2027After approval
NCCN and payer policyNCCN guideline placement and payer coverage policies
Treatment line, paid duration and CRC penetration
QuarterlyEvent-driven
NET and Cabo trackingNET oral share and duration; Handa availability and payer decisions
Cabo growth versus earlier competitive erosion

Rows are in expected order, not to scale. Which events count as dated is a Clinaptis judgment.

Source: Clinaptis diligence and company/regulatory calendar. Guidance-reporting month is expected, not a confirmed company event date. Criteria are observations that change model inputs; they are not stock-price triggers.

EXEL stays on our high-priority watchlist. Cabo can finance the search for a successor. The price already pays for one the evidence has not yet shown, and the proof arrives as Cabo’s cash flows taper.

Appendix: valuation conventions and supporting exhibits

Valuation conventions

The SOTP is our anchor: it values the finite assets we can underwrite. The company DCF is a cross-check that adds the post-exclusivity cost response and a continuing-company terminal value (40% GAAP EBIT margin, 1% growth). Multiples test near-term earnings, not the cliff. The two cash-flow methods are not interchangeable: the SOTP charges the modeled assets with 50% of scoped shared costs, while the DCF funds all resized operating budgets. Both discount at 10.5% and add back stock compensation without deducting a claim for future awards; risk sits in the probability of success, which is subjective and applied once, not in the discount rate.

The hybrid fair value (Exhibit 8) weights the SOTP 25%, the company DCF 15%, a DCF to 2029 with a 6–10x exit multiple on 2029 cash EBITDA 20%, oncology EV/sales 10% and FY27 EV/EBITDA 30%. The weights are an editorial choice (Exhibit A5). Ranges use 9% and 12% discount rates for the cash-flow methods; for multiples, the 25th to 75th percentile of eleven commercial oncology peers on EV/sales (EXEL excluded) and three GAAP EV/EBITDA anchors: INCY-adjusted, EXEL’s own 24-month median and NBIX-adjusted (Exhibit A6). We exclude precedent transactions, whose prices embed control premia, and peer P/E, which acquired in-process R&D charges distort.

The residual bridge (Exhibit 7) re-runs the company DCF on a representative sell-side long-dated P&L, unattributed. Its operating income is non-GAAP, before stock compensation, so we deduct $130M a year growing at 2% for tax, apply our 22% rate, add it back to cash flow, run the P&L explicitly to 2036 and apply the same terminal value. Alternative treatments give $50.65 (stock compensation of $200M a year, explicit to 2045, no terminal value) and $66.28 (terminal value at that forecast’s ~73% 2036 GAAP margin). The reverse grid (Exhibit 9) applies the model’s indication ramp, discount factors, 22% tax, 1% D&A, 2% capex, 10% working capital and post-2040 erosion to a hypothetical probability-weighted peak, and solves for the peak that equals the residual. FactSet and ChartMill figures in Exhibit 11 are as published.

Exhibit A1The modeled indications support a useful franchise; the valuation depends on the rows we have not valued
Opportunity Evidence supporting the underwrite Conditional mature US sales Working PoS Asset rNPV
Later-line CRC Randomized Ph3 ITT OS benefit; modern sequencing unresolved $288M 80% $440M
First-line nccRCC Ph3 pending; class support, no verified direct early ncc efficacy $198M 50% $177M
Post-IO/TKI ccRCC, LITESPARK-034 Separate randomized combination trial; no verified efficacy yet $165M 30% $47M
Post-adjuvant ccRCC, LITESPARK-033 Ph3 vs Cabo; primary completion est. Feb 2032 Not valued — —
NET, STELLAR-311 Ph3; primary completion est. Jun 2029 Not valued — —
MRD CRC and other pipeline Earlier-stage; no quantified underwrite Not valued — —
Ex-US zanza Partner economics not yet modeled Not valued — —

Source: Clinaptis Analytical Model and clinical diligence. Market value above our SOTP at the 30 Oct close: $9,557M on a diluted basis. Sales are conditional commercial estimates, not trial-derived outcomes; individual mature course estimates are rounded. Asset rNPVs precede separate prelaunch/shared-cost deductions. Mature combined sales are $651M conditional and $379M probability weighted. PoS is subjective and applied once: the rNPV already includes it.

Exhibit A2Patient flow, paid exposure and price explain the revenue assumptions
Mature US model CRC nccRCC ccRCC / -034
Starting population proxy 23,000 3L+ 21,400 first-line RCC Specific post-IO/TKI population
Modeled screened candidates 11,685 3,411 Separate expansion funnel
Unique annual starts 3,199 across lines 1,100 918
Mean paid months 4.5 9.0 9.0
Net revenue / paid month $20,000 $20,000 $20,000
Conditional mature sales $288M $198M $165M
Probability-weighted sales $230M $99M $50M

Source: Clinaptis patient funnels/current model. All flow screens, uptake, price and paid means are analyst assumptions. CRC includes separate later-line routes; its screened pool is not the denominator for a simple all-line share calculation. RCC excludes unsupported histologies/clinical populations. Revenue belongs to zanza only, excluding partner-drug sales.

Exhibit A3Selected-asset SOTP, $M and $ per diluted share
Selected-asset SOTPValue, $M$ / diluted share
Cabo US6,29824.89
Net ex-US royalties7362.91
CRC / nccRCC / ccRCC asset rNPVs6662.63
Prelaunch development costs(303)(1.20)
Retained shared costs, 50%(1,600)(6.32)
Realizable cost tax shields4191.66
Incremental tail after 2045110.04
Stock compensation added back3741.48
Operating value6,59926.08
Net cash and securities1,4005.53
Modeled equity value7,99931.62
Market reference, diluted basis17,55669.39
Value outside current underwrite9,55737.77

Source: Clinaptis Analytical Model. 10.5% discount rate; 253.0M diluted valuation shares: modeled Q4 2026 weighted count after post-Q3 repurchases, including an existing-award proxy (Q3 reported diluted weighted average: 255.8M). Rounding affects displayed sums. Basic market EV is $15.53B; the diluted reference EV used for this bridge is $16.16B. Residual is not uniquely identifiable pipeline NPV or a market-implied success probability.

Exhibit A4Net cash pulls most established names well below their sales multiple
0×2×4×6×8×10×12×median EV/S 2.6×median P/S 4.4×URGN10.4׆11.0×ONC5.9×6.6×DCTH4.8×5.8×EXEL5.6×EV/S n/a, no cash capturedGMAB5.7×4.8×SNDX2.4׆4.7×IMCR2.3׆4.4×INCY3.6׆4.2×CHRS1.9׆4.0×GERN2.1׆3.7×RIGL2.6׆3.1×LEGN2.3×3.0×Price / sales (market cap ÷ TTM revenue)EV / sales, debt nettedEV / sales, debt not netted†

† Debt not captured in the source set; EV/S is a lower bound. Revenue is trailing twelve months to the latest reported quarter; market caps are as of mixed dates between Aug and Oct 2026, so treat gaps under about 1× as noise. GMAB market cap is flagged for verification and its EV/S includes about $5.2B of debt. EXEL has no cash data in the set.
Source: company filings via Clinaptis oncology comps v2 (internal, 2 Nov 2026).

Exhibit A5No weighting of the five methods reaches the price
WeightingCentral valuevs price
Proposed weights$41.57-40%
Equal weight, five methods$38.76-44%
Intrinsic only: rNPV 60 / DCF 40$28.91-58%
Intrinsic 50 / multiples 50$37.84-45%
Cash-flow methods only$36.24-48%
Multiples only: EV/S 25 / EV/EBITDA 75$51.15-26%

Weights apply to the central case of each method. Cash-flow methods are rNPV, company DCF and the exit-multiple row. Modeled; not observed data.
Source: Clinaptis Analytical Model (Valuation Summary).

Exhibit A6. Operating-profit comparisons

Comparable basis EV / TTM EBIT EV / TTM EBITDA
EXEL, current 15.48x 15.06x
EXEL, 24-month median 12.99x 12.54x
EXEL, 36-month median 14.51x 13.81x
NBIX, specified transaction adjustments 17.11x 16.20x
INCY, specified transaction/rebate adjustments 12.85x 12.11x

SBC and recurring R&D remain expenses. NBIX uses TTM Jun 2026; EXEL and INCY use Q3. INCY removes $1,231.5M acquired IPR&D and the $246M CMS rebate benefit; NBIX removes specified transaction costs. These are limited analytical normalizations, not issuer-adjusted EBITDA. Trading EV uses observed basic equity capitalization; implied per-share values use the diluted model denominator. The early portion of the 36-month EXEL series has an unusually small EBIT denominator; medians are more informative than a simple mean.

Sources: company SEC filings; Clinaptis Analytical Model.

Exhibit A7Cutting the discount rate to 8% lifts the hybrid by $1.60, not the gap

A single lower rate on every asset moves the hybrid from $41.57 to $43.17, still 38% below the $69.39 price. Splitting the rate, with a lower one on Cabo and ex-US royalties only, adds about $0.26.

Discount raterNPV / SOTPCompany DCFExit multiple, 8xHybridvs price
12.0%$30.41$23.53$45.88$40.77−41%
10.5% (base)$31.62$24.84$47.40$41.57−40%
9.0%$32.94$26.55$49.01$42.48−39%
8.0%$33.90$28.03$50.14$43.17−38%
Split: Cabo and ex-US 9%, zanza 10.5%$32.63$24.84not re-run$47.40not re-run$41.83−40%

Hybrid reweights the central case of each method at the stated rate (weights 25/15/20/10/30); the multiple-based legs do not depend on the rate. The split case changes the rNPV / SOTP leg only; the company DCF and exit-multiple rows are not re-run on a split rate. Headline valuation stays at the 10.5% lifecycle rate. CAPM cross-check: 4.93% risk-free, 4.20% equity risk premium, beta 0.65 to 1.0 gives a 7.7% to 9.1% cost of equity. Modeled; not observed data.
Source: Clinaptis Analytical Model (Valuation Summary, Firm DCF, Asset DCF); discount-rate reruns 2 Nov 2026; price as of 30 Oct 2026.

Exhibit A8Cabo sensitivities and cash-flow reconciliation behind Section 2
ItemCaseResult
Second-line share, 2028 (from 37%)27% base; 22% bear; 32% bull−$1.16; −$1.74; −$0.58 per share2028 Cabo revenue $2.35B; $2.28B; $2.42B, against $2.49B
Second-line-and-later share of Cabo RCC revenue20%; 25% base; 30%; 35%−$0.93; −$1.16; −$1.39; −$1.63 per shareAnalyst assumption; not disclosed by the company. Applied to RCC revenue of about $2.1B in 2028; 2028 Cabo revenue $2.38B; $2.35B; $2.32B; $2.30B, against $2.49B
Further second-line pool contraction2% a year on top of base share−$1.37 per share$0.21 beyond the base case
Cabo growth above our case, 2027–30+3 points; +6 points a year+$1.75; +$3.61 per share2030 Cabo revenue $2.72B; $2.99B; about $1.2 per $100M of average annual Cabo revenue
’039 patent loss, MSN entry from Jan 15, 2030About 10.5 months earlier than Dec 2030About −$3.6 per share before any cost response$3.0 to $3.9 across Handa scenarios; 10.5% discount rate
Nine-month operating cash flow less capexReported; excluding $94M prepaid release; also treating $90M SBC as a cost~$829M; ~$735M; ~$645M

Per-share impacts are changes in SOTP value from the base $31.62, holding all other inputs. Second-line scenarios apply the share change to the stated slice of Cabo RCC revenue from 2027, with a one-tenth step in 2027. Modeled; not observed data.
Source: Clinaptis Analytical Model (scenario runs, 1–2 Nov 2026); Exelixis Q3 2026 10-Q.

Exhibit A9CRC funnel, duration and denominator checks behind Section 4
ItemDetail
Third-line screens, company reference23,000 company-cited US third-line-plus patients → 22,080 non-MSI-H (96%) → 17,664 with required prior therapy (80%) → 17,311 without pivotal-excluded prior treatment (98%) → 12,983 fit for a TKI-IO combination (75%) → 11,685 not deferred to a biomarker-directed option (90%)Each screen is an analyst assumption and may overlap exclusions already in the company figure; the result is an illustrative candidate pool, not a validated estimate. The company’s materials do not reconcile 23,000 to unique annual third-line entrants.
DenominatorsThird-line starts use the brief’s ~18,000 non-MSI-H pool (9% = 1,620). The later-line pool is built on 22,080, so fourth- and fifth-line starts are 1,579. Aligning both to 18,000: 2,884 starts, ~$260M conditional (~$208M after 80%). No patient is counted as a start more than once: each later line draws only on patients who have not started zanza.Later-line capture is 30% of trial-naive and 15% of exposed fourth-line candidates, 22.5% and 11.25% in fifth line. At 20% and 10% in fourth line, conditional sales are ~$250M; with both changes, ~$229M.
Paid duration4.5 months assumed. An exponential curve with the trial’s 3.4-month median would imply a ~4.9-month mean; paid months can run below treatment duration because of dose holds and dispensing gaps.Setting paid months equal to the 3.4-month median gives $218M conditional.
STELLAR-303 safety detailSerious treatment-related events 26% versus 10%; discontinuation for AEs 18% versus 15%; treatment-related deaths five versus one, too few events to compare.
STELLAR-001, RAS wild-type CRC54 versus 53 patients; median OS 11.7 versus 11.1 months, combination versus zanza alone (HR 0.89, 95% CI 0.56–1.42); PFS HR 0.65.
Immunotherapy competitorAgenus’s Ph3 BATTMAN, botensilimab plus balstilimab against best supportive care in refractory MSS CRC, began enrolling in April 2026 (~830 patients). It has no data, and a positive result would land after our launch window, but it targets the same patients.

Modeled except where trial data are cited; not observed data.
Source: Clinaptis Analytical Model and CRC patient funnel; Clinaptis CRC landscape brief; Exelixis CRC materials; STELLAR-303 and STELLAR-001 publications.

Exhibit A10STELLAR-304 outcome cases, risk weight and price check behind Section 5
ItemDetail
Commercial cases, conditional on approvalStrong result: mature share 55% papillary, 45% unclassified, 35% translocation (52% of suitable patients), ~1,774 annual starts, 12 paid months, ~$426M. Working: 35% / 25% / 20% (32.25%), ~1,100 starts, 9 months, ~$198M. Conservative: 20% weighted share, ~682 starts, 6 months, ~$82M. All at $20,000 net per paid month on ~3,400 combination-suitable patients.Shares are analyst hypotheses for mature use, not a prescriber survey. A statistically positive result against sunitinib alone does not justify the strong case.
Risk weight on the working case40%: ~$79M. 50%: ~$99M. 65%: ~$129M risk-adjusted. The weight applies once to conditional value; the cases above are not a second probability weighting.
Price per paid month (working case)$17,000: ~$168M. $20,000: ~$198M. $23,000: ~$228M conditional. Revenue is linear in price.Inference, not re-run: the three zanza assets together are worth $2.63/sh in our SOTP, so a price change of this size moves company value by well under $1/sh even allowing for operating leverage. It does not change the conclusion. The $15,000 Cabo counterfactual is the benchmark; we have not shown that payers will pay a premium in overlap settings.
Trial designPublished design: response rate in the first 222 patients at two-sided alpha 0.005, PFS at 0.045, each with ~90% power; the event count is not published. On that arithmetic, PFS significance needs an observed HR of about 0.74 (192 events) to 0.76 (230 events). Response-rate power (about 148 versus 74 patients): ~83% at 35% versus 12%, ~57% at 30% versus 12%, ~90% at 45% versus 18%. A two- to three-month delay before the curves separate would cut PFS power materially (at a true HR of 0.65, from ~0.85 to ~0.4–0.7).Clinaptis calculations, not a probability of success. The amended statistical plan may differ (317 enrolled against 291 planned), so thresholds are approximate.

Modeled except where trial design is cited; not observed data.
Source: Clinaptis Analytical Model and nccRCC patient funnel; STELLAR-304 registry and design publication; PAPMET.

Exhibit A11SOTP, DCF and revenue-substitution checks behind Section 6
ItemDetail
SOTP by shared-cost retention0%: $36.00. 30%: $33.37. 50% (base): $31.62. 100%: $27.23. 100% plus funding all future growth R&D: $21.93.
Why the company DCF is $24.84, not $31.62The DCF funds every dollar of the resized operating budget, counts explicit cash flows to 2036 and adds a continuing-company value after (1% growth, 40% EBIT margin), which is ~15% of enterprise value. The SOTP charges the assets half of shared costs and runs finite drug cash flows to 2045 with a declining tail. Zanza revenue enters each method once.Both methods add stock compensation back to free cash flow with no future claim. Charging it as a claim would lower the company DCF to $22.01.
Operating-forecast substitutionCompany DCF $24.84 to $55.09 when a representative sell-side long-dated forecast (revenue and costs) replaces ours at 10.5%: +$30.25 of the $44.55 gap; $14.30 remains.Representative, not consensus: one long-dated forecast, unattributed, not verified against multiple brokers. Operating income there is non-GAAP and before stock compensation; we deduct a modeled $130M of stock compensation growing 2% a year to reach GAAP-equivalent EBIT.
Revenue-only and cost-onlyRevenue-only (that forecast’s revenue, our cost dollars, D&A, capex and working capital held): $59.09, +$34.25. Cost-only (its cost dollars on our revenue): $19.17, −$5.67. Interaction +$1.67. Together: +$30.25.The revenue-only case implies EBIT margins of ~70% in 2032 and ~88% in 2036, so it is an upper bound. If 30% of the added revenue is absorbed by cost: $52.55 (+$27.71); 10%: $56.91; 50%: $48.20. Terminal value is unchanged in these runs because it is a 40% margin on 2036 revenue.
Remaining gapOn that forecast the price implies a discount rate of ~8.1%. Alternative terminal treatments: $50.65 to $66.28, which does not reach $69.39.
Exit multiplePost-2029 flows imply ~2.2x 2029 cash EBITDA; the price needs ~13.6x. Both use cash EBITDA with stock compensation added back, the same explicit 2026–2029 forecast and 10.5% rate. On GAAP EBITDA: 2.45x and 15.1x. Applying 6x to 10x gives $39.60 to $55.21.
Hybrid arithmetic0.25 × 31.62 + 0.15 × 24.84 + 0.20 × 47.40 + 0.10 × 32.60 + 0.30 × 57.34 = $41.57. Intrinsic-only (rNPV 60 / DCF 40) $28.91; multiples-only (EV/S 25 / EV/EBITDA 75) $51.15, as in Exhibit A5.

Modeled; not observed data. Runs reproduce the Firm DCF with a replica that ties to the workbook ($24.84).
Source: Clinaptis Analytical Model (Firm DCF, SOTP, scenario runs 2 Nov 2026); a representative sell-side long-dated forecast (unattributed).