Our 83% approval probability exceeds the roughly 75% implied by the price, supporting a modest positive expected return into the 22 November FDA decision. The commercial valuation is harder to support: our approved-case value of $826M sits roughly 39% below the $1.35B we expect the market to pay after approval. We would sell a rally toward that level; sustaining it requires stronger patient uptake, longer treatment persistence or better economics than our base case assumes.

Savara Inc. (SVRA) reaches its FDA decision on 22 Nov 2026 with two questions attached to one stock. Clock 1 is the next 28 days: will MOLBREEVI be approved for autoimmune pulmonary alveolar proteinosis (aPAP), and where does the stock trade the morning after? Clock 2 is the 12 to 36 months that follow: how many patients start paid treatment, how long they stay on it, and what that franchise is worth. The answers point in different directions.

On Clock 1 we sit above the market. At $4.22 ($1.14B), the price implies ~75% approval if the stock trades near $1.35B after approval and near $0.5B after a complete response letter (CRL). We put approval at 83%, a judgment above a ~74% first-cycle reference rate and an 81% eventual-approval count for extended reviews, resting on Breakthrough designation, Priority Review, no advisory committee and an extension letter that cited no safety, efficacy or manufacturing concern. The edge is modest and the tail is severe: +19% on approval, −56% on a CRL, about +6% expected. If the approval move is ~11% or less, the market already prices our view.

On Clock 2 we sit below it. Our base case of 1,700 US patients and $564M of mature US net sales, with operating costs that rise with sales, values the approved franchise at $826M ($3.06 a share): 27% below today’s price and ~39% below where we expect the stock to trade after approval. On our costs and price, ~$1.35B needs ~2,800 US patients on drug. That requires adoption beyond our central willingness assumption (~2,460 eligible and willing patients), or a different answer on price, margins or franchise duration.

So we are long the letter and cautious on the launch. After an approval we would sell a rally toward $1.35B and move to underweight, without a “short the launch” thesis: the gap can persist through the early launch, M&A interest can hold it open, and ~21% of shares are already sold short.

Two clocks Part I (Sections 1–3) is Clock 1: the clinical package, the approval case and what the price implies. Its dollar values are trading values, where we expect the stock to trade the day after the letter. Part II (Sections 4–7) is Clock 2: the launch, our base case and what the franchise is worth to us. Its dollar values are Clinaptis fundamental values, from a 12-year DCF with no terminal value. Every dollar figure below carries its clock.

Exhibit 1

One decision, two clocks: approval is worth ~+19% to the stock but leaves our value 27% below the price

Clinaptis probabilities for the 22 Nov PDUFA, with the trading value (Clock 1) and the Clinaptis value (Clock 2) on each branch. Equity values, $; per share at 269.81M diluted shares.

Probabilities are Clinaptis judgments, not FDA signals. On the same trading values the price implies ~75% approval; expected trading value at 83% is ~$1.21B, ~+6%. The Clinaptis CRL value assumes 85% eventual approval after ~18 months; derivation in Appendix A1. We carry no separate clinical-CRL branch. Market capitalization uses 269.81M diluted shares; on 205.46M basic shares it is ~$867M.

Source: Clinaptis estimates (modeled, not observed data); market data to 23 Oct 2026 close.Clinaptis Research


Part IClock 1: the letter

1. MOLBREEVI gives specialists a reason to start treatment

IMPALA-2 confirms a benefit in gas transfer and quality of life. The functional and lavage-avoidance claims remain unproven.

The alveolar macrophage is the lung’s garbage truck, and it only shows up for work when granulocyte-macrophage colony-stimulating factor (GM-CSF) tells it to. In aPAP, autoantibodies neutralize GM-CSF, the macrophages go idle and surfactant accumulates instead of being cleared: patients slowly fill their own lungs with detergent, and gas exchange fails. Patients present with dyspnea, hypoxemia and exercise limitation. The standard procedure, whole-lung lavage (WLL), washes surfactant out under general anesthesia, is repeated as the disease recurs and is concentrated in a few experienced centers. It does not correct the defect. The alternative is off-label inhaled sargramostim (Leukine, ~$115K a year at ~$310 for a 250 µg vial if dosed daily, about half that on the alternate-week regimen approved in Japan), and the 2024 ERS guideline already places inhaled GM-CSF alongside lavage for appropriate symptomatic patients.

MOLBREEVI is recombinant human GM-CSF, E. coli-derived and non-glycosylated, unlike yeast-derived sargramostim, delivered once daily through a PARI eFlow nebulizer. It replaces the missing signal locally; it does not remove the autoantibodies. It is not a clever mechanism, but it is the correct one. Treatment is therefore either chronic or episodic, and that distinction is the commercial question in Section 5.

Savara did not discover the molecule: Serendex assembled inhaled molgramostim with GEMA supply and a PARI device before Savara bought the program in 2016 (Appendix A3). Savara owns the dossier and the cell banks and depends on third parties for drug substance, drug product and device. Its first Ph2/3, IMPALA (n=138), missed its primary endpoint; the second, IMPALA-2 (n=164), met its primary in Jun 2024.

IMPALA-2 establishes the physiological effect. DLCO (diffusing capacity for carbon monoxide, a measure of gas transfer) improved by 6.0 percentage points (pp) against placebo at week 24 (95% CI 2.5–9.4) and by ~6.9 points at week 48 (P=0.0008, company topline). The total score on the St George’s Respiratory Questionnaire (SGRQ) improved by 6.6 points at week 24 (P=0.007), a confirmatory quality-of-life result; at week 48 the difference was ~4.9 points with a confidence interval crossing zero. Adverse-event and serious-adverse-event rates were similar between arms.

The functional claims need qualification. The testing hierarchy stopped at SGRQ-Activity, so the endpoints after it carry no confirmatory inference; the exploratory treadmill-distance result (+80.6 m at week 48, nominal P=0.03) is supportive only. Rescue WLL during the double-blind period was lower on drug (7.4% of patients against 13.3%), but this was an exploratory endpoint and does not establish lavage avoidance. The trial enrolled clinically stable patients with impaired DLCO, not patients needing urgent rescue, and the Oct 2026 analysis in mild diffusion impairment is small and exploratory. The original IMPALA missed its primary endpoint; a reanalysis excluding invalid blood-gas measurements was significant but not prespecified, so IMPALA supports the direction of effect without counting as a second pivotal win (Exhibit 2). IMPALA-2 then moved the primary to DLCO and met it (+6.0 pp). A one-for-two registrational record is still not a fortress.

Retention was high: 98% completed the double-blind period and 94% remained in the open-label extension at the reported data cutoff. That supports tolerability and acceptability. Multiyear paid persistence is a separate question, taken up in Section 5B.

Exhibit 2

IMPALA-2 confirms gas transfer and quality of life; functional and lavage claims remain exploratory

Placebo-adjusted results by endpoint, classified by what the testing hierarchy allows us to conclude.

EndpointResultStatusCommercial read
DLCO, week 24 (primary)+6.0 pp; 95% CI 2.5–9.4; P<0.001ConfirmedSupports initiation
DLCO, week 48+6.9 pp; P=0.0008SignificantEffect maintained to week 48
SGRQ total, week 24−6.6 points; P=0.007ConfirmedPatient-felt benefit
SGRQ total, week 48−4.9 points; CI crosses zeroInconclusiveDurability uncertain
SGRQ Activity, week 24Missed adjusted threshold; hierarchy stopsNot shownLimits function claims
Exercise distance (treadmill, exploratory)+80.6 m at week 48; nominal P=0.03No inferenceSupportive only
Rescue WLL, double-blind7.4% vs 13.3% of patientsExploratoryLavage avoidance unproven
SafetyAE and SAE rates similarSupportiveTrial tolerability reassuring; final label unknown
Retention98% completed DB; 94% in OLE at data cutoffTrial onlyNot paid persistence
IMPALA primary (A–aDO2)Missed; post hoc reanalysis significantMissedDirection, not replication

IMPALA-2 n=164, 48-week double-blind (primary endpoint at week 24), then a 96-week open-label extension. Status reflects the prespecified hierarchy, not clinical plausibility.

Source: IMPALA-2 publication and company presentations (ATS 2025) and releases (May and Jun 2026); IMPALA (NEJM 2020); Clinaptis classification.Clinaptis Research


For the regulator, this is an approvable package: a met primary endpoint, a confirmatory quality-of-life result and a clean safety table in a disease with no FDA-approved therapy. The May 2025 refusal to file turned on manufacturing, not efficacy; the subject of the information requests behind the current extension is not public. Manufacturing is where the approval case turns.

2. The approval case: why we sit above the base rate

We put first-cycle approval at 83%. The reference rates are ~74% first-cycle approval and 81% eventual approval for extended reviews. What lifts SVRA above it is procedural and public; what holds it back sits in manufacturing, where the deciding evidence is not.

Start from the base rate. FDA’s own analysis of first-cycle outcomes puts the median first-cycle approval rate for original NDAs and BLAs at ~64% over FY2008–24; for new molecular entities and new therapeutic BLAs, the FY2024 first-cycle CRL rate was 26%, or ~74% approval. Our own count of applications with extended review dates since 2015 finds 142 of 176 eventually approved (81%), but that counts approval in any cycle, not on the extended date. No published rate isolates the cohort SVRA belongs to, so our 83% is a judgment, not a statistic. Our edge is eight probability points on a public record that leaves the deciding manufacturing evidence out of view. We size it accordingly.

The case for clearing it. MOLBREEVI carries Breakthrough Therapy designation and received Priority Review in Feb 2026, and the Day 74 letter said no advisory committee was planned. The FDA then moved the PDUFA date from 22 Aug to 22 Nov after classifying Savara’s responses to information requests as a major amendment; the company says the extension letter raised no safety, efficacy or manufacturing concern. On the manufacturing side, Savara reports that Fujifilm has completed three process performance qualification (PPQ) batches, the validation batches the FDA reviews before approval (2025 10-K).

Where it could still go wrong. The first BLA drew a refusal to file in May 2025 that asked for more chemistry, manufacturing and controls (CMC) information; Savara said it raised no safety concern and required no new efficacy study. Savara resubmitted in Dec 2025 with Fujifilm as commercial drug-substance manufacturer, after comparability work. The supply chain has three outside dependencies, each of which the FDA must accept: drug substance at Fujifilm (replacing GEMA, which made the clinical material and assigned the cell banks to Savara), drug product at Patheon, and PARI’s eFlow nebulizer under license. A Dec 2023 amendment also obliges Savara to buy GEMA API for a redacted share of product for ten years after approval. One commercial drug-substance supplier, one nebulizer and $126M of cash leave little room for a manufacturing detour.

The Fujifilm 483 is a network signal, not a site finding. Savara’s 2025 10-K names Fujifilm’s Billingham, UK plant as the MOLBREEVI drug-substance site, where technology transfer and the three PPQ batches are complete. A Mar 2025 Form 483 for a different plant, Fujifilm’s College Station, Texas facility, lists eight observations across contamination control in multi-product areas, cleaning and environmental monitoring, analytical method validation, stability tests run 24 to 107 days late, electronic-record controls and quality-unit oversight. These are quality-system observations, not a final FDA compliance determination, and the public record does not tie them to Billingham or to molgramostim. They matter only as a read on quality systems across Fujifilm’s network. Completed PPQ batches are not a passed pre-approval inspection, and we have found no public record of an inspection outcome at Billingham.

The evidence that would settle the question, the outcome of any pre-approval inspection at Billingham and the content of the information requests behind the extension, is not public, and management has not described those requests publicly. Reticence before a letter is normal. Placing the 483 at a different plant removes one specific concern, but not the larger unknown of what the FDA asked for in the major amendment; that is why we stop at 83% and not 90%. Everything we say about CMC is inference from the public record.

If the letter is a CRL. We carry one CRL branch and assume the asset survives it: the clinical package stays intact, Savara remediates and resubmits, and a six-month review puts approval ~18 months out. The $75M RTW payment requires approval by 31 Mar 2027 and would be lost. Savara would face the delay with ~$126M of estimated cash against $32M of Hercules debt, so dilution follows. We drop a separate clinical-CRL branch because nothing in the public record points to one.

3. What the price implies, and what the stock’s year says

The price implies ~75% approval if the stock trades near $1.35B after approval. This year’s de-rating shows the market has already cut the value it will pay for an approved MOLBREEVI; it cannot show how much confidence in approval went with it.

One equation, two unknowns. Today’s price is the approval probability times the value after approval, plus the CRL probability times the value after a CRL. The price alone cannot separate the two. The size of the move on approval pins the value after approval, and the implied probability then falls out. So every implied probability in this section is stated with the trading values it assumes.

Pricing the approval branch. We use ~$1.35B, a 19% move from $4.22, and treat +10% to +30% as the plausible range. Two cross-checks bracket it, and our own break-even sits between them:

  • Street framing, discounted. Sell-side bull cases broadly pair US peak sales of $500M or more with a 4–5x multiple, or $2.0–2.5B. We treat that as a value at peak and discount it four to five years at 11%, to ~$1.2–1.6B, centered near $1.4B. Read as a present value, it would put approval nearer $2B and the implied probability at ~43%, so ours is the conservative reading.
  • Deal comps, the lower check. The 2.6x median of rare-disease acquisitions in Appendix A4, applied to our base year-4 US sales of $410M, gives ~$1.07B of enterprise value, or ~$1.16B of equity.
  • Our break-even. Approval must be worth more than today’s $1.14B, and our long needs ~$1.27B to break even at 83%. The deal-comp value and the low end of the discounted Street range ($1.16–1.20B) would make it a losing trade.

We anchor to a market-based value rather than our own model on purpose: after approval the stock trades on its holders’ view of the franchise, not ours. January’s $1.63B shows what the market paid before the de-rating, not that it will pay it again. Two things could shave the move. The +19% is not fully protected from dilution: $126M of cash and the $75M RTW payment carry our base through its first two years of launch losses without a raise, with a low point near $100M, but a raise into approval-day strength would be the usual move. And with ~44M shares sold short and an event-driven holder base, approval could become a liquidity event, with the stock weaker within days.

Pricing the CRL branch. We mark ~$0.5B, a ~56% fall. Applying the market’s own ~$1.35B approval value to a delayed approval (85% eventual approval, ~18 months, the RTW payment lost, burn and financing costs) gives ~$0.75B, with burn and financing costs already netted out; we take a third off that for forced selling and holder turnover. The two closest recent cases fell much less: Scholar Rock ~15% pre-market on a Sep 2025 CRL tied to a third-party fill-finish site, Ultragenyx ~5% on a Jul 2025 CRL citing facility observations. Savara has less room: eighteen months of burn, a lost $75M payment and a debt stack. The mark matters to the trade. At −40% the expected return rises to ~+9%; at −70% it falls to ~+4%.

The arithmetic. On those values, $1.14B implies ~75% approval against our 83%. The expected trading value is ~$1.21B, ~6% above the price (Exhibit 3). The result depends heavily on the approval value: each $50M moves the implied probability by ~4–5 points. Below an ~11% approval move, the price already holds our view. That is the trade: an 11% hurdle, a 56% cliff and one letter between them.

Exhibit 3

The price implies ~75% approval at a ~$1.35B approval value; our 83% is worth ~6% into the letter, and nothing below an ~11% approval move

Approval probability implied by the $4.22 price ($1.14B) for each approval move and CRL trading value, and the expected return at our 83%. Clock 1 trading values.

Approval probability the $4.22 price implies Expected return at our 83%, CRL at $500M 50% 60% 70% 80% 90% Price implies less than our 83% Our 83% Break-even ~+11% Below it, the price already embeds our 83% 87% 82% 78% 73% 69% 85% 79% 75% 69% 65% 79% 72% 67% 61% 56% CRL at $366M (Clock 2) CRL at $500M (trading mark) CRL at $700M Central case 0% +5% +10% +15% −1% +3% +6% +11% +15% +10% $1,252M +15% $1,309M +19% $1,355M +25% $1,423M +30% $1,480M Move on approval Approval value
View data
Move on approvalApproval value, $MImplied, CRL at $366MImplied, CRL at $500MImplied, CRL at $700MExpected return at 83%, CRL at $500M
+10%1,25287%85%79%−1%
+15%1,30982%79%72%+3%
+19% (our central case)1,35578%75%67%+6%
+25%1,42373%69%61%+11%
+30%1,48069%65%56%+15%

Implied probability = (price − CRL value) / (approval value − CRL value). Markers are our values at five approval moves; lines apply the same formula in between. $366M is our Clinaptis CRL value (Clock 2), shown for comparison; $500M is our trading mark. Break-even approval move at 83% and a $500M CRL value: ~+11%. The approval move can also be negative if approval becomes a liquidity event; the exhibit does not model that.

Source: Clinaptis estimates; price and diluted shares at the 23 Oct 2026 close.Clinaptis Research


A year of de-rating. SVRA is down 30% this year while the XBI is up 23%, and up only 2% over twelve months against the XBI’s 39% (Exhibit 4). Over the past month it fell ~20% against ~8% for the sector, so most of the recent weakness is specific to SVRA. The stock round-tripped: ~$1.11B in Oct 2025, ~$1.63B in Jan 2026 after the resubmission, $1.14B now. January’s $1.63B was above $1.35B even at certain approval, so at least part of the de-rating is a re-mark of the value the market will pay for an approved MOLBREEVI. The price cannot show whether approval confidence fell too.

The decline came in steps: into the mid-Apr review extension (announced 15 Apr), again in mid-Jul around the departure of the chief financial officer (announced 8 Jun, for health reasons, with the chief operating officer adding the role), and again over the past month. We take the stated reason at face value; the practical cost is continuity, since the executive who carried the extension message to investors left before the letter.

Short interest cuts both ways. Shares sold short doubled from 22.3M at the 15 Jan settlement to 43.6M at 30 Sep, ~21% of basic shares outstanding, with 17 to 36 days to cover. Rising short interest into a falling price is active selling, not just an absence of buyers: someone is paying to bet against the letter, or against the launch that follows it. The same position could prompt short covering if the letter is an approval, which could carry the first move past +19% before the launch question reasserts itself.

Exhibit 4

SVRA has fallen 30% this year while the XBI rose 23%, and short interest has doubled

Daily closes indexed to 100 at 31 Dec 2025. The 53-point gap opened in steps: after the mid-Apr review extension, in mid-Jul and over the past month. Shares sold short rose from 22.3M to 43.6M, ~21% of basic shares.

608010012014031 Dec 2025 = 100JanFebMarAprMayJunJulAugSepOctFeb: BLA accepted,Priority Review15 Apr: review extendedto 22 Nov8 Jun: CFO transitionannouncedXBI 122.6+22.6% this yearSVRA 70.0−30.0% this year53-point gapDaily closes, 31 Dec 2025 to 23 Oct 2026. Index points.
Settlement dateShares sold shortShare of basic sharesDays to cover
15 Jan 202622.3M~11%14
15 May 202627.9M~14%24
15 Jul 202634.1M~17%21
15 Sep 202643.3M~21%36
30 Sep 202643.6M~21%17

Shares sold short, millions; all bars on one scale. Latest settlement in blue.

View data
DateSVRA close, $XBI close, $SVRA indexXBI index
31 Dec 20256.03121.93100.0100.0
2 Jan 20265.64121.5293.599.7
5 Jan 20265.8120.1596.298.5
6 Jan 20265.54122.0991.9100.1
7 Jan 20265.77126.4395.7103.7
8 Jan 20265.76123.7195.5101.5
9 Jan 20265.81124.596.4102.1
12 Jan 20265.77123.9395.7101.6
13 Jan 20265.92123.8598.2101.6
14 Jan 20266126.8299.5104.0
15 Jan 20265.88124.3597.5102.0
16 Jan 20265.82124.1896.5101.8
20 Jan 20265.97125.2999.0102.8
21 Jan 20266.19128.18102.7105.1
22 Jan 20266.46131.57107.1107.9
23 Jan 20266.02128.0499.8105.0
26 Jan 20265.78127.8695.9104.9
27 Jan 20265.85128.5397.0105.4
28 Jan 20265.61125.9493.0103.3
29 Jan 20265.58126.7492.5103.9
30 Jan 20265.4124.7589.6102.3
2 Feb 20265.52127.3591.5104.4
3 Feb 20265.66127.6593.9104.7
4 Feb 20265.36125.2988.9102.8
5 Feb 20265.16120.7285.699.0
6 Feb 20265.3125.5187.9102.9
9 Feb 20265.23126.1186.7103.4
10 Feb 20265.27125.3487.4102.8
11 Feb 20265.35124.6588.7102.2
12 Feb 20265.83123.1896.7101.0
13 Feb 20265.6122.8692.9100.8
17 Feb 20265.73124.6395.0102.2
18 Feb 20265.59125.692.7103.0
19 Feb 20265.72126.7494.9103.9
20 Feb 20265.53124.7891.7102.3
23 Feb 20265.7126.9794.5104.1
24 Feb 20265.64129.1693.5105.9
25 Feb 20265.84128.2896.8105.2
26 Feb 20265.97127.6799.0104.7
27 Feb 20266.02127.3799.8104.5
2 Mar 20266.01126.5599.7103.8
3 Mar 20265.58123.592.5101.3
4 Mar 20265.62126.4293.2103.7
5 Mar 20265.49123.991.0101.6
6 Mar 20265.2712487.4101.7
9 Mar 20265.43127.1290.0104.3
10 Mar 20265.61127.4393.0104.5
11 Mar 20265.53126.3891.7103.6
12 Mar 20265.32122.6288.2100.6
13 Mar 20265.25121.8387.199.9
16 Mar 20265.23123.9186.7101.6
17 Mar 20265.15124.1985.4101.9
18 Mar 20265.04121.1183.699.3
19 Mar 20265.16122.3385.6100.3
20 Mar 20264.96120.3182.398.7
23 Mar 20265.08121.784.299.8
24 Mar 20265.04119.5483.698.0
25 Mar 20265.28123.9287.6101.6
26 Mar 20265.25123.987.1101.6
27 Mar 20265.14119.5985.298.1
30 Mar 20265.05118.7883.797.4
31 Mar 20265.46127.7390.5104.8
1 Apr 20265.58128.5592.5105.4
2 Apr 20265.56128.9692.2105.8
6 Apr 20265.6128.892.9105.6
7 Apr 20265.78128.8595.9105.7
8 Apr 20265.94130.6598.5107.2
9 Apr 20265.87131.8297.3108.1
10 Apr 20265.66129.4493.9106.2
13 Apr 20265.92132.0598.2108.3
14 Apr 20266.07135.72100.7111.3
15 Apr 20265.87136.4597.3111.9
16 Apr 20265.64135.4693.5111.1
17 Apr 20265.78138.6795.9113.7
20 Apr 20265.61137.7993.0113.0
21 Apr 20265.28136.3987.6111.9
22 Apr 20265.14137.3185.2112.6
23 Apr 20264.94134.4981.9110.3
24 Apr 20265.12133.2884.9109.3
27 Apr 20265.16133.0485.6109.1
28 Apr 20265.02131.1683.3107.6
29 Apr 20264.85128.980.4105.7
30 Apr 20265.24131.3386.9107.7
1 May 20265.12130.4284.9107.0
4 May 20265.3133.587.9109.5
5 May 20265.25133.7987.1109.7
6 May 20265.27137.6887.4112.9
7 May 20265.16133.6685.6109.6
8 May 20265.23134.7186.7110.5
11 May 20265.12134.7584.9110.5
12 May 20265.08134.9184.2110.6
13 May 20265.27135.8587.4111.4
14 May 20265.27134.8587.4110.6
15 May 20265.03130.6983.4107.2
18 May 20264.91127.6781.4104.7
19 May 20264.72126.7978.3104.0
20 May 20265.08131.7184.2108.0
21 May 20265.06132.7883.9108.9
22 May 20265.02131.6683.3108.0
26 May 20265.14133.3585.2109.4
27 May 20265.21134.4586.4110.3
28 May 20265.09135.9984.4111.5
29 May 20265.2136.6986.2112.1
1 Jun 20265.15133.6285.4109.6
2 Jun 20265.25127.7687.1104.8
3 Jun 20265.25129.8387.1106.5
4 Jun 20265.3133.4287.9109.4
5 Jun 20265.09128.6784.4105.5
8 Jun 20265.06128.4383.9105.3
9 Jun 20264.93131.3881.8107.8
10 Jun 20265.09128.8584.4105.7
11 Jun 20265.27132.7487.4108.9
12 Jun 20265.26133.7987.2109.7
15 Jun 20265.33136.488.4111.9
16 Jun 20265.3135.3487.9111.0
17 Jun 20265.38139.3989.2114.3
18 Jun 20265.5140.7291.2115.4
22 Jun 20265.69145.8694.4119.6
23 Jun 20265.78147.0395.9120.6
24 Jun 20265.83149.7196.7122.8
25 Jun 20266.03151.59100.0124.3
26 Jun 20265.9155.3897.8127.4
29 Jun 20266.03158.31100.0129.8
30 Jun 20266.16158.25102.2129.8
1 Jul 20266.06156.55100.5128.4
2 Jul 20266.22160.46103.2131.6
6 Jul 20266.2160.81102.8131.9
7 Jul 20266.21163.87103.0134.4
8 Jul 20266.38162.97105.8133.7
9 Jul 20266.29164.28104.3134.7
10 Jul 20266.03159.03100.0130.4
13 Jul 20265.74155.3495.2127.4
14 Jul 20265.69155.4594.4127.5
15 Jul 20265.81156.2296.4128.1
16 Jul 20265.5215291.5124.7
17 Jul 20265.52154.2691.5126.5
20 Jul 20265.49150.9491.0123.8
21 Jul 20265.84154.596.8126.7
22 Jul 20265.65152.1193.7124.8
23 Jul 20265.75152.2395.4124.9
24 Jul 20265.51150.4891.4123.4
27 Jul 20265.43150.5990.0123.5
28 Jul 20265.33149.7888.4122.8
29 Jul 20265.28147.9187.6121.3
30 Jul 20265.54151.4691.9124.2
31 Jul 20265.46147.0190.5120.6
3 Aug 20265.33147.3188.4120.8
4 Aug 20265.41151.8889.7124.6
5 Aug 20265.51153.0191.4125.5
6 Aug 20265.79154.596.0126.7
7 Aug 20265.83157.3796.7129.1
10 Aug 20265.73158.0395.0129.6
11 Aug 20265.57158.0792.4129.6
12 Aug 20265.7159.3994.5130.7
13 Aug 20265.27156.8687.4128.6
14 Aug 20265.38157.4189.2129.1
17 Aug 20265.37159.5389.1130.8
18 Aug 20265.37160.1189.1131.3
19 Aug 20265.58169.5592.5139.1
20 Aug 20265.49163.3891.0134.0
21 Aug 20265.59165.7392.7135.9
24 Aug 20265.56164.1792.2134.6
25 Aug 20265.6169.192.9138.7
26 Aug 20265.61168.3993.0138.1
27 Aug 20265.49168.2391.0138.0
28 Aug 20265.26162.3887.2133.2
31 Aug 20265.39162.589.4133.3
1 Sep 20265.35163.488.7134.0
2 Sep 20265.46165.3790.5135.6
3 Sep 20265.37164.3889.1134.8
4 Sep 20265.37163.8189.1134.3
8 Sep 20265.53161.9391.7132.8
9 Sep 20265.42159.3889.9130.7
10 Sep 20265.32156.8288.2128.6
11 Sep 20265.3156.287.9128.1
14 Sep 20265.34157.688.6129.3
15 Sep 20265.26154.0387.2126.3
16 Sep 20265.13154.1885.1126.4
17 Sep 20265.42158.2589.9129.8
18 Sep 20265.37156.7289.1128.5
21 Sep 20265.45158.2390.4129.8
22 Sep 20265.3161.8787.9132.8
23 Sep 20265155.282.9127.3
24 Sep 20265.07156.0584.1128.0
25 Sep 20265155.0382.9127.1
28 Sep 20265.07156.6184.1128.4
29 Sep 20265.08156.8184.2128.6
30 Sep 20265.16157.6885.6129.3
1 Oct 20265.04154.5183.6126.7
2 Oct 20264.81154.4379.8126.7
5 Oct 20264.78156.1979.3128.1
6 Oct 20264.65150.977.1123.8
7 Oct 20264.88150.2380.9123.2
8 Oct 20264.83149.2980.1122.4
9 Oct 20264.92153.7781.6126.1
12 Oct 20264.84151.480.3124.2
13 Oct 20264.78146.879.3120.4
14 Oct 20264.76146.7778.9120.4
15 Oct 20264.74149.7878.6122.8
16 Oct 20264.45144.3273.8118.4
19 Oct 20264.63147.8676.8121.3
20 Oct 20264.76153.2478.9125.7
21 Oct 20264.67152.7877.4125.3
22 Oct 20264.54151.6475.3124.4
23 Oct 20264.22149.5470.0122.6

Twelve months to 23 Oct 2026: SVRA +2.4%, XBI +39.0%. Past month (22 Sep to 23 Oct): SVRA −20%, XBI −8%. The Feb line marks the month, not the day. Share of basic shares uses 205.46M (11 Aug 2026) for every date; the 15 Oct settlement is published 26 Oct.

Source: daily closing prices (FinanceCharts); FINRA short interest via Equibles; Savara releases for event dates. Observed data.Clinaptis Research


Options price more movement than our tree. At the reported 152% annualized implied volatility (23 Oct), a rough 30-day at-the-money straddle proxy is about 35% of spot, against an expected absolute move of ~25% in our tree. That is context for the event premium; it does not rank a directional option structure against the stock, and we have not priced the chain.

A ~6% expected return with a −56% branch is a trade to size for the tail, not a position: a 1% position makes ~19bp on approval and loses ~56bp on a CRL. What the stock is worth after the letter is a separate question, and on that we are more cautious than the market.

Part IIClock 2: the franchise

4. A concentrated prescriber base makes the launch reachable

Savara can reach most identified patients through a few hundred accounts. Reaching them does not determine how many convert to paid treatment or stay on it.

The company estimates ~5,500 US aPAP patients from claims, ~1,000 of them in line of sight, and 65% of identified patients in its top 500 accounts. The top 500 accounts are also not 500 centers of excellence; many will have one or two patients. The launch infrastructure is in place: ClearPath anti-GM-CSF antibody testing to confirm diagnosis, PANTHERx Rare as exclusive specialty pharmacy and hub, and patient-access support. ClearPath confirms patients who are already in the claims estimate rather than expanding it, and PANTHERx reduces friction without proving conversion.

We expect the launch to move from centers of excellence and trial sites in year one to regional interstitial lung disease (ILD) networks in years two to four, then to community referral. The first prescriptions should go to symptomatic patients with impaired gas transfer, a recurring need for lavage or inadequate control on off-label GM-CSF. Patients doing well on off-label sargramostim are the hardest to switch: the physician sees little reason to change a stable patient, and payers may ask why a ~$450K list-price product should replace a ~$115K one. We do not model formal payer step-therapy requirements; our 65% to 82% paid conversion absorbs some denials, but a payer that insists on sargramostim first would push conversion lower. Japan is a cautionary reference. Inhaled sargramostim (Sargmalin) launched for aPAP in Jul 2024 and recorded ¥167M of sales in its first five months and ¥798M in 2025, at Japanese list prices (Nobelpharma business reports for 2024 and 2025). At ¥42,359 a vial on the approved alternate-week regimen, 2025 sales equal roughly 100 patient-years of treatment, ~14% of Japan’s ~750 estimated patients, in a market where lavage is entrenched.

Launch analogs inform speed more than a ceiling. Genetic orphans converted ~50% of a known pool within months. Orphan drugs that had to find patients after approval reached ~2% to 15% after one to three years. MOLBREEVI starts like a pre-identified launch, with ~1,000 known patients, and then has to behave like a conventional one. The first few hundred patients are the easy part; the base case depends on what happens after them. Our base case reaches 12.5% of the 5,500 pool at 24 months and 31% by year 7, at the strong end of conventional launches. The Street’s CY2033 revenue needs 35% to 59% of the pool, depending on the ex-US share. At the top of that range the Street needs three in five claims-identified US patients on a daily nebulizer; we need fewer than one in three.

Exhibit 5

Genetic orphans reached ~50% within months; non-genetic orphans sit at ~2% to 15% after one to three years

Share of the pool on drug against months since launch, log scale. Analogs are colored by how patients were found; open diamonds are context (expanded-access conversion, identified patients), not on-drug share. The MOLBREEVI base case reaches 1,700 patients, 31% of the 5,500 claims-identified pool, by year 7, above conventional orphans and below genetic ones; the bar at year 7 is what Street needs. Hover a marker for counts and denominators.

Metrics are not harmonized (on drug, dispensed, prescribed, enrolled, start forms) and denominators are company or analyst estimates, so read shares as orders of magnitude. Epidiolex (conversion of expanded-access patients) and the MOLBREEVI identified cohort are context, not on-drug share; Epidiolex is plotted at ~5 months, inferred; the MOLBREEVI identified cohort is pre-launch context placed at the left edge of the axis, not a two-month observation; Fintepla and Lupkynis are not plotted (denominator unverified). MOLBREEVI is plotted against the 5,500 claims-identified pool; the ~1,700 patients at maturity are 31% of it and 48% of the 3,520 moderate or severe treatment-eligible patients (Section 5A); Street requirements assume a $360K net price and 92% paid coverage.

Source: company disclosures and published launch analogs (secondary); Clinaptis estimates (modeled, not observed data).Clinaptis Research

Analog values as a table
AnalogTime since launchPatientsShare, as reported or derived
Genetic orphan: genotype known before launch
Kalydeco (CF, G551D)~4 mo~600 on drug~50% of ~1,200 pool
Orkambi (CF, F508del homozygous)~6 mo>4,50053% of ~8,500 pool
Context, not on-drug share: patients known before launch
Epidiolex (Dravet, LGS)At approval~1,200 in expanded access~75% of ~900 converted by Q1'19
Epidiolex5 mo>7,600 dispensedNot computed
Fintepla (Dravet)~18 mo1,129 prescribedPrevalence share not verified
MOLBREEVI identified cohort (pre-launch)Launch~1,000 identified (company claim)~18% of 5,500 claims count
Conventional orphan: patients found after launch
Arikayce (MAC lung disease)3 mo / ~6.2 yr500 / ~2,100~14% of 15,000 at ~6.2 yr
Winrevair (PAH)6 mo / ~21 mo2,600 new / ~6,060~6.5% / ~15% of 40,000
Filspari (IgAN)19 mo2,989 start forms~6% to 10% of 30,000 to 50,000
Tarpeyo (IgAN)24 mo2,791 enrollments~1.9% of 150,000, or 6% to 9% of 30,000 to 50,000
Lupkynis (lupus nephritis)35 mo2,066 on therapyDenominator unverified
MOLBREEVI base case, year-end patients on therapy
Base, month 12 / 24 / 3612 / 24 / 36 mo291 / 685 / 1,0795.3% / 12.5% / 19.6% of 5,500
Base, month 8484 mo1,70230.9%
Street CY2033, all US84 mo~3,23059% of 5,500
Street CY2033, 20% / 30% / 40% ex-US84 mo~2,590 / 2,260 / 1,94047% / 41% / 35% of 5,500

The analog definitions mix enrolled, prescribed, dispensed and on-therapy patients, and the denominators are estimates of different vintages, so the exhibit fixes orders of magnitude rather than a precise MOLBREEVI ceiling.

5. The Clinaptis US base case: 1,700 patients and $564M, if they stay

Twelve-month persistence is the assumption that moves the conclusion, and it is the least validated.

Clinaptis base case We use a 1,600–1,800 ongoing-patient base range, centered on 1,700, conditional on approval. After allowing for short unpaid refill gaps, the center produces 1,565 paid patient-equivalents a year. At $360K net per paid year, that supports $563M of annual US net sales by 2033.

5A. Who starts

The starting population is the company’s ~5,500 claims-derived patients. Claims support patient-finding, not confirmed antibody status or current symptoms, and the ~1,000 line-of-sight cohort sits inside that pool rather than adding to it. Applying its physician-survey severity mix gives ~1,430 severe, ~2,090 moderate and ~1,980 mild patients; this is a scenario proxy, not a confirmed eligible registry. As a cross-check, treating 55% of severe, 40% of moderate and 4% of mild patients gives ~1,700 ongoing patients, of whom ~1,620 are moderate or severe (~46% of that group). The cross-check tests plausibility; the model does not apply it as a further haircut.

The cohort model runs on initiation episodes: 500 in 2027, peaking at 780 in 2029, then ~463 a year at maturity, including restarts. Effective paid-access conversion, the share generating a first paid fill within the year net of denials, abandonment and timing, rises from 65% in 2027 to 82% from 2032. Both are Clinaptis assumptions. Year one produces 325 paid starts and 291 ongoing patients by Dec 2027.

Reach, willingness and persistence. Three filters sit between the claims pool and a treated patient. Reach: of ~5,500 claims-identified patients, ~3,520 are moderate or severe treatment-eligible, the group a pulmonologist would plausibly treat. Willingness: a company-commissioned physician survey, cited in the company’s current corporate presentation, puts 70% willing to start, which leaves ~2,460 patients and treats the other 30% as permanent non-starters. Persistence: 80% at 12 months, an ~4.5-year average course.

The annual flow is the harder test. Our 1,700 ongoing patients at maturity are 48% of the eligible group and 69% of the willing group, so the base needs most willing patients to start and stay on drug. It also needs ~380 starts or restarts a year to hold that level, against only ~110 new eligible diagnoses a year. That figure uses an aPAP incidence of 0.49 per million, the minimum found in Japan’s national registry (Inoue et al., AJRCCM 2008); at the 0.2 per million more often cited for all PAP (StatPearls), it falls to ~45. Restarts and mild patients must fill a ~270-patient-a-year gap between the starts we need and the diagnoses that arrive, and we cannot verify either source. That makes our base demanding, not conservative, and is one reason we flag persistence as the largest sensitivity.

5B. Who stays

We assume 80% of patients remain on their treatment course 12 months after a paid start, modeled as a constant exit hazard of ~22% a year. That implies an average course of ~4.5 years, with ~33% of patients still on the same course after five years. This duration is unvalidated and it materially moves value. Trial retention does not prove it: IMPALA-2 patients received free drug under scheduled visits after selection for stability. Commercial patients face a daily nebulizer, a $360K net annual cost to payers and, if they improve, a reason for their physician to try a treatment holiday. Inhaled therapy has a commercial persistence problem. In US claims, 46.7% of patients with pulmonary arterial hypertension (PAH) on nebulized inhaled treprostinil and 16.1% on inhaled iloprost were still on drug at 12 months, counting a gap of 60 days or more as discontinuation (Optum claims, 2010–20). PAH is a warning about nebulizer burden, not a forecast for MOLBREEVI: those therapies need four, or six to nine, sessions a day, so our 80% asks once-daily dosing to make a substantial difference. It is our base, and it is not a cautious one.

The model separates three behaviors:

  1. Refill gaps: short gaps sit inside the paid-coverage assumption (92% at maturity).
  2. Discontinuation (D/C): longer cessation leaves the census.
  3. Restart: a patient who later returns is counted as a new paid start but not as a new unique patient.

At 80% persistence, holding 1,700 ongoing patients requires 379 paid starts or restarts a year, or 463 initiation episodes at 82% conversion. The forecast is a flow of treatment courses, not a closed pool of lifelong users; it does not model depletion of treatment-naive patients. Exhibit 9 is our cross-check on that flow: a separate model that tracks reached, interrupted and restarted patients as distinct pools.

The sensitivity is steep. Holding the mature paid-start flow fixed, 70% persistence supports 1,064 ongoing patients and $352M of sales; 85% supports 2,334 and $773M. At 70%, the Clinaptis value of the approved franchise falls from $826M to $359M; at 85% it rises to $1,221M. It is the largest single-input sensitivity we ran.

5C. What gets paid

Ongoing patients are retained treatment courses, including temporary refill gaps. Paid patient-equivalents are annual paid supply divided by a full year’s supply. Price is the midpoint of the company’s $400–500K expected list range, $450K, less 20% gross-to-net, held flat at $360K net. The bridge in the box above multiplies the 1,700 census only by 92% paid coverage: access and discontinuation have already determined the census, so they are not applied again. Our valuation maps the same monthly ramp into approval years; the unrounded census is ~1,702, which gives the $564M shown at maturity, so the difference is rounding.

Exhibit 6

US net sales reach ~$295M in 2029 and ~$564M by 2033 in the Clinaptis base case

Annual US net sales, calendar years, approval-conditional, with ongoing patients and paid patient-years for selected years.

US net sales, $M015030045060049202716220282952029410203049620315472032564203356420345642035Plateau held flat
Calendar year20272028202920312033
Paid starts and restarts325504593544377
Year-end ongoing patients2916851,0791,5981,702
Paid patient-years1364518181,3771,567
Net sales, $M49162295496564

Assumes a Jan 2027 launch after approval. Calendar years; the valuation uses approval years (year 1 = Jan–Nov 2027, then Dec–Nov), so its early-year sales are slightly lower ($41M in approval year 1). Rows rounded; calculations use unrounded monthly values.

Source: Clinaptis US base case model (modeled, not observed data).Clinaptis Research


Where we differ from consensus in 2027. Sell-side consensus, as of 23 Oct 2026, sits at $70M or more for 2027 on an assumed ~$400K net price, which needs ~175 paid patient-years in the first year. Onboarding 175 patients in Q1 2027 would not deliver that: time to first fill, benefits verification and payer review mean the year’s paid exposure trails the patients onboarded. Our base has 325 paid starts in 2027 but only 136 paid patient-years, for $49M at $360K net. The price and the pace of conversion are both checkable within two quarters of launch.

6. What the franchise is worth to us

Approved, the franchise is worth $826M ($3.06 a share) on our base case: 27% below today’s price and ~39% below where we expect the stock to trade after approval. Weighted at 83% approval, it is worth $748M ($2.77).

6A. US cash flows when costs grow with the business

Costs rise with sales. A company that sells $564M of a rare-disease biologic does not run on a pre-launch budget, so we treat today’s run-rate as a floor (commercial $45M, R&D $35M, G&A $20M) and let each line rise to a share of sales once that share is larger: commercial and medical 15%, R&D 8% (lifecycle and post-marketing work), G&A 7%. Mature cash opex reaches ~30% of sales and the plateau EBITDA margin ~48%. As a check, Catalyst, a profitable rare-disease company, reported FY2025 SG&A of $194M and R&D of $13M on $589M of revenue (company release), ~35% of sales in total; it spends more on selling and less on R&D than we assume, at a similar total.

On that basis, US enterprise value is $645M and US equity if approved is $698M, after $126M of estimated cash, $32M of Hercules debt, the $75M RTW payment and the RTW royalty. Europe adds $128M after a 90% approval weight, for $826M. Margins remain the risk inside the value: moving commercial spend from 15% to 20% of sales (~$28M a year at maturity) costs ~$119M, and cutting stock compensation from 30% to 20% of cash opex adds ~$72M.

Assumptions and method choices:

  • DCF: 12 approval years to Nov 2038, 11% discount rate, mid-year discounting from 22 Nov 2026, no terminal value.
  • Costs: PARI royalty 3.5%, COGS 12.5%, variable distribution and hub 6% of sales; opex as above; launch costs $20M and $10M in years 1–2; stock compensation at 30% of cash opex, treated as a cost.
  • Tax: 25%, with a $195.3M net operating loss (NOL) usable against 80% of income.
  • Not modeled: working capital (~$50M of value at 12% of sales) and the GEMA royalty (~$20M of value per 1% of sales, so ~$20–60M at 1–3%). RTW terms, debt, cash and share count are in Appendix A2.

6B. Europe as a separately built block

We build Europe country by country across Germany, the UK, France, Italy and Spain, with launches lagging US approval by 0.6 to 2.5 years; inputs are in Appendix A5 and are model assumptions, not a sourced table. The base reaches ~760 ongoing patients and ~$138M of mature sales, worth $142M under a direct launch ($99M if partnered), or $128M after the 90% EU approval weight we apply given US approval. The bear case is partnered and worth $28M; the bull is worth $565M. The European block carries its own country-level costs and has not been re-run with the US cost scaling. After a CRL we delay Europe with the US and apply the same 85% eventual-approval weight, since Fujifilm and Patheon supply both regions.

6C. What the DCF leaves out

No terminal value. The DCF ends in Nov 2038. US orphan exclusivity runs seven years from approval, to ~Nov 2033; a new BLA may carry 12 years of reference-product exclusivity, to ~Nov 2038, but eligibility is unverified, so this is an inference. That cuts against us: if another GM-CSF product can enter after 2033, the DCF’s last five years are generous, not conservative. In Europe, a formulation patent is reported to Mar 2041 and a joint Savara–PARI device patent to Mar 2043, with US equivalents and claim scope unverified. None of it blocks a different GM-CSF product. Any value after 2038 is upside to our number, and a buyer or a market willing to capitalize it would close part of the gap; we choose not to underwrite it.

The approved value by patient count. On our costs and price, the approved franchise matches today’s $1.14B only at 2,340 US patients, and the ~$1.35B we expect the market to pay after approval needs 2,780 (Exhibit 7). At the bull case of 2,400 patients, with a direct European launch near its bull, it is worth $1,549M ($5.74 a share).

Exhibit 7

At 1,700 patients the approved franchise is worth $826M to us; today’s price needs ~2,340 and the expected post-approval price ~2,780

Clinaptis value if approved (Clock 2), US plus Europe held at base, against peak US patients on an ongoing treatment course. Pick a case or drag the marker.

Peak US patients
Peak US patients
Mature US net sales
Clinaptis value if approved
Versus $1.14B price

Approval-conditional DCF, 12 years to Nov 2038 at 11%, no terminal value, with opex rising with sales; includes cash, debt, the RTW payment and royalty, and Europe at its base value ($142M at a 90% approval weight). Europe is held at base along the line, so the bull point here ($1,169M) is below the full bull case in Exhibit 8 ($1,549M), which adds the bull European launch. Other patient counts scale the base ramp.

Source: Clinaptis estimates (modeled, not observed data). Price is market capitalization at the 23 Oct 2026 close.Clinaptis Research

View data
Peak patientsMature US sales, $MValue if approved, $MVersus price, $M
Bear1,200398547−592
Base1,700564826−313
Bull2,4007961,169+30
Matches price (approx.)~2,340~7761,140+1
Matches ~$1.35B (approx.)~2,780~9221,352+213

Street sits at our bull case. Fintel consensus (23 Oct 2026) is $1,071M for CY2033, which we treat as worldwide; by then it rests on six analysts. On approval year 7 (Dec 2032–Nov 2033, ~CY2033), our base is ~$689M worldwide ($564M US plus ~$125M from a Europe still ramping) and our bull ~$1,117M.

Exhibit 8

On our numbers the approved franchise is worth $3.06 a share against a $4.22 price; only the bull case clears it

Clinaptis values (Clock 2) by case. Base: 1,700 US patients, $564M mature US net sales, Europe adding $138M of sales. DCF only, no terminal value, opex rising with sales.

BearBaseBull
Patients, exposure and sales
US patients on therapy, mature1,2001,7002,400
US paid patient-years, year 3 / mature556 / 1,105787 / 1,5651,112 / 2,210
US net sales, $M, year 3 / year 5 / mature200 / 346 / 398283 / 490 / 564400 / 691 / 796
EU net sales, $M, mature (structure)48 (partnered)138 (direct)342 (direct)
US EBITDA margin, mature47%48%48%
Clinaptis value per share
If approved (US, EU at 90%)$1.65$3.06$5.74
After a CRL, asset intact$0.33$1.36$3.31
Probability-weighted, 83% approval$1.42$2.77$5.33
If approved, versus $4.22−61%−27%+36%
Probability-weighted, versus $4.22−66%−34%+26%
What has to happen commercially
Persistence nearer 73% than 80%, or access stalls well short of 82% conversion. Functional benefit does not persuade payers; Europe only through a partner.~1,700 patients on drug at maturity, 2032–33 (31% of the 5,500 claims-identified pool): 80% 12-month persistence, access converting 65% to 82%, flat $360K net price. Direct EU launch from Germany and the UK in 2027–28.2,400 patients (44% of the pool): ~85% persistence at base-case starts, or ~40% more starts, plus a direct EU launch reaching ~1,500 patients at 70% of US net price. Opex rises with sales at the same ratios.
View data
BearBaseBull
US + EU equity if approved, $M4448261,549
US equity if approved, $M4196981,041
EU value if approved (before 90% weight), $M28142565
Value after a CRL, asset intact, $M89366892
Probability-weighted at 83%, $M3847481,437
Worldwide net sales, approval year 7, $M4366891,117

DCF runs 12 approval years to Nov 2038 at 11% with no terminal value. Opex floors at today’s run-rate and rises to 15% (commercial), 8% (R&D) and 7% (G&A) of sales. CRL values assume 85% eventual approval after ~18 months, the RTW payment lost, $120M a year of burn and $25M of financing costs; Europe is delayed with the US and carries the same 85% weight after a CRL. Years are approval years (year 1 = Jan–Nov 2027), so year 3 and year 5 differ from the calendar years in Exhibit 6. Bear and bull scale the base ramp by patient count. Per share at 269.81M diluted shares; bars scaled to $6.50.

Source: Clinaptis estimates (modeled, not observed data).Clinaptis Research


7. The gap between the market and our value, and what would close it

If approval comes, we expect the stock to trade on a launch curve steeper than ours. We will sell into that, but we will not bet in a hurry on the gap closing.

The gap, stated plainly. We expect the market to pay ~$1.35B for an approved MOLBREEVI that we value at $826M. On our costs and price, that value needs ~2,800 US patients on drug, about half of the 5,500 claims pool and above the ~2,460 eligible and willing patients in our central case, so it needs adoption beyond that assumption. The market can get there by other routes: a higher net price (consensus uses ~$400K, against our $360K), better margins, a terminal value, or an acquirer paying for the franchise it sees. Each route is plausible on its own. Together they are our bull case, not our base.

Mapped across launch curves. Exhibit 9 runs a patient-flow analysis across 45 combinations of reach speed, willingness to start and persistence, and values each on our current costs. At our fixed price, costs and European value, ~$1.35B sits at the strong end of this grid. One tested cell clears it (high persistence, 90% willing, the whole eligible pool reached in year one); with 40% reach the same profile reaches $1.27B, so the crossing lies somewhere between the two. Two cells clear today’s price and two more miss it by less than 1%. The count describes the scenarios we tested; it does not measure their odds. Mid-persistence cells, where our base sits, top out at ~$1.09B.

Exhibit 9

~$1.35B sits at the strong end of 45 tested launch curves: one cell reaches it, and two clear today’s $1.14B

Clinaptis value if approved (Clock 2), $M, by annual reach hazard (rows) and ultimate willing share (columns), for three persistence profiles. The outlined cell is closest to our base case census. Cells are tested scenarios, not weighted by likelihood.

Low persistence: 25% D/C a year, no restarts

Reach hazard50% willing70% willing90% willing
8%287287390
15%287345538
25%287456652
40%297531731
100%383602813

Mid persistence: 10% D/C, 10% interruptions, 50% restart

Reach hazard50% willing70% willing90% willing
8%287386574
15%282527724
25%398649854
40%485746961
100%5828541,090

High persistence: 5% D/C, 5% interruptions, 60% restart

Reach hazard50% willing70% willing90% willing
8%338586790
15%478737967
25%5988761,130
40%6969921,273
100%8141,1341,441
Below $500M$500M to $826M$826M (our base) to $1.14B$1.14B (price) to $1.35B$1.35B or more
View data
PersistenceReach hazardWillingPatients on or between treatment, year 7US net sales, year 7, $MValue if approved, $M
Low8%50%429157287
Low8%70%600220287
Low8%90%771283390
Low15%50%538198287
Low15%70%753277345
Low15%90%968356538
Low25%50%623232287
Low25%70%872325456
Low25%90%1,121417652
Low40%50%662252297
Low40%70%927352531
Low40%90%1,192453731
Low100%50%611238383
Low100%70%856334602
Low100%90%1,100429813
Mid8%50%742214287
Mid8%70%1,038299386
Mid8%90%1,335384574
Mid15%50%924266282
Mid15%70%1,294373527
Mid15%90%1,663480724
Mid25%50%1,085315398
Mid25%70%1,519440649
Mid25%90%1,952566854
Mid40%50%1,195350485
Mid40%70%1,672489746
Mid40%90%2,150629961
Mid100%50%1,209357582
Mid100%70%1,692500854
Mid100%90%2,1766431,090
High8%50%894289338
High8%70%1,252405586
High8%90%1,610521790
High15%50%1,111360478
High15%70%1,555503737
High15%90%2,000647967
High25%50%1,308426598
High25%70%1,832597876
High25%90%2,3557671,130
High40%50%1,455478696
High40%70%2,037670992
High40%90%2,6188611,273
High100%50%1,512504814
High100%70%2,1177051,134
High100%90%2,7229061,441

Reach hazard is the share of still-unreached eligible patients reached in a year; 100% reaches the whole pool in year 1 (half in the launch year). All cells use a 3,520 eligible pool, a 1,000-patient line-of-sight cohort, 110 new diagnoses a year, ClearPath at 150 a year (half overlapping), a 40% start hazard and $360K net per treated year; price, costs, European value and franchise duration are held fixed. Two high-persistence cells ($1,130M and $1,134M) miss the price by less than 1%. Weak cells floor near $287M, where stopping operations beats continuing; the RTW royalty stops with them. The outlined cell (mid persistence, 40% reach, 70% willing) has ~1,670 patients at year 7 against our base of 1,700, but lower sales ($489M against $564M) because this analysis deducts interruptions and 95% adherence; its value ($746M) sits below our base ($826M) for that reason. Values include Europe held at base ($128M after a 90% approval weight).

Source: Clinaptis estimates; patient-flow analysis valued in our DCF (opex rising with sales, DCF to Nov 2038 at 11%, no terminal value). Modeled, not observed data.Clinaptis Research


What would close it upward. Twelve-month persistence above 80%; paid conversion faster than 65% in year one; prescribing that spreads into regional ILD practices by the second year; a direct European launch nearer the bull; opex discipline that holds margins above ~48%. What would widen it. Persistence nearer 70% (value $359M if approved), payer step edits behind off-label sargramostim, a narrow label with a severity floor, or slow time to first fill.

The tests and their dates.

  1. ~11 Nov (expected, not yet confirmed): Q3 results and cash; every equity value here moves with the $126M estimate.
  2. 22 Nov: the letter, the label’s breadth (symptomatic aPAP without a narrow severity floor) and any supply conditions on Fujifilm, Patheon or PARI.
  3. Q1–Q2 2027: first paid starts and time to first fill, against 325 paid starts and $49M of sales in 2027 in our base.
  4. From Jan 2028: the first 12-month persistence data, near 80% in our base, with 24-month data (near 64%) from 2029. A cohort tracking toward 70% at 12 months would put the base near ~$350M of sales.
  5. Europe: the CHMP opinion, the first reimbursement decisions in Germany and the UK, and launch timing against our lags.

Stance Rent the letter; do not marry the launch. Into 22 Nov we are long, sized so the CRL branch is a loss the book can take: 83% approval against ~75% implied, +19% on approval against −56% on a CRL, about +6% expected, and no edge if the approval move is ~11% or less. A 1% position makes ~19bp on approval and loses ~56bp on a CRL. We express it in the stock; we have not priced specific option structures. We sell a rally toward $1.35B and move to underweight, since the stock should then trade well above our value ($826M if approved, $3.06 a share). If approval arrives without the rally, we reassess the price rather than sell the letter mechanically. We do not short the launch. A gap like this can persist for several quarters, M&A interest can hold it open, and with ~21% of shares already short the trade is crowded. We re-underwrite on the first paid starts and the first 12-month persistence cohort.

Appendix

A1. PDUFA tree with branch values, by clock
Branch, 22 Nov 2026Clinaptis probabilityClock 1: trading value, $MContributesClock 2: Clinaptis value, $MContributes
Approval, first cycle83%~1,350~1,121826685
CRL, asset intact17%~500~8536662
Probability-weighted100%~1,206 (+6%)748 (−34%)
Implied by $1,139M price~75%On the Clock 1 values; Exhibit 3 shows the range

Clinaptis CRL value ($366M): 85% eventual approval × ($645M US EV − $116M PV of RTW royalties + $128M risked Europe, delayed with the US) discounted 1.5 years at 11% = $477M; plus $94M net cash ($126.0M estimated cash less $32.1M Hercules debt), less $180M of burn over 18 months and $25M of financing costs. The $75M RTW payment is lost; the RTW royalty is assumed to survive, which is conservative and unverified.

Trading CRL value (~$500M): the same arithmetic on the market’s ~$1.35B approval value gives ~$747M; we take about a third off for forced selling and financing at a depressed price. Recent third-party-site CRLs fell less (Scholar Rock ~−15%, Sep 2025; Ultragenyx ~−5%, Jul 2025), but both companies were better funded. Unrounded probabilities are as shown; no separate clinical-CRL branch is carried.

A2. Royalties, financing and other obligations
ClaimTermsModel treatmentStatus
PARI3.5% of net sales; term is the later of last valid device patent claim or 15 years from first sale3.5% for the full 12 yearsConfirmed
GEMA royaltyLow-single-digit % of net sales; ends per country at earlier of 10 years after approval or first biosimilar saleNot modeled; ~$20M of approved value per 1%Unresolved
GEMA supplyDec 2023 amendment: 10-year purchase of a redacted share of API, with price and supply exceptionsNot modeled separatelyUnresolved
RTW$75M on FDA approval by 31 Mar 2027; US net-sales tiers 1–7%, possible 9.5% tier; $187.5M cap$75M on approval; flat 7% to cap (approval year 8)Simplified
Hercules$32.085M drawn; further tranche undrawnFace value; undrawn tranche has no valueConfirmed
Cash at decision$173.0M at 30 Jun 2026$126.0M at 22 NovEstimated; Q3 on 11 Nov
Shares205.46M basic (11 Aug 2026), 48.60M pre-funded warrants, options and PSUs269.81M treasury methodEstimated
Operating costsRun-rate commercial $45M, R&D $35M, G&A $20MFloors; each rises to 15% / 8% / 7% of sales when largerClinaptis assumption
Working capitalReceivables and inventory for a specialty biologicNot modeled; ~$50M of value at 12% of salesOmitted
A3. How Savara assembled the program

Serendex developed inhaled molgramostim (then Molgradex) from academic GM-CSF work, signing a GEMA supply and license agreement in Dec 2012 and a PARI device collaboration in Nov 2014. Savara acquired the program in Jul 2016 for 1,965,400 shares and up to $21.5M of regulatory milestones, ~$12.4M of acquisition-date accounting consideration, as summarized in secondary reviews of company filings. Savara merged with Mast Therapeutics in Apr 2017, amended the PARI agreement in 2018 and took assignment of the GEMA cell banks in Apr 2019. IMPALA missed its primary endpoint (NEJM 2020); IMPALA-2 began in Jun 2021 and met its primary in Jun 2024. Fujifilm was contracted for drug substance in Feb 2024. The BLA followed in Mar 2025, the refusal to file in May 2025, resubmission in Dec 2025 and Priority Review in Feb 2026.

A4. Deal comps
Target / acquirerYearBasisMultipleStatus
Rare-disease and specialty deals: adviser precedent analysis
Day One / Servier2026EV / year-4 revenue3.3xSecondary
Avadel / Alkermes2025EV / year-4 revenue4.1xSecondary
SpringWorks / Merck KGaA2025EV / year-4 revenue2.6xSecondary
Calliditas / Asahi Kasei2024EV / year-4 revenue (consensus)1.7xSecondary
CTI BioPharma / Sobi2023EV / year-4 revenue3.5xSecondary
Albireo / Ipsen2023EV / year-4 revenue2.5xSecondary
Zogenix / UCB2022EV / year-4 revenue2.3xSecondary
Median, seven deals2.6x
Closest to SVRA: recently approved or launching single-asset rare-disease companies
Reata / Biogen2023EV / year-4 revenue6.2xSecondary
Soleno / Neurocrine2026EV / revenue, 2027 / 2028 / 20294.1x / 3.2x / 2.9xSecondary
Soleno / Neurocrine2026Equity value ($2.9B) / management peak revenue ($1,100M, 2034)~2.6xCalculated from the company release and 14D-9; EV would be lower
Development-stage deals: context only
Alpine / Vertex; DICE / Lilly; Prometheus / Merck; Momenta / J&J2020–24Risk-adjusted peak revenue0.9x; 1.0x; 1.2x; 3.0xSecondary; not comparable to an approved asset

Multiples are adviser analyses in merger filings as summarized in secondary research. Soleno’s deal value and revenue projections are checked to the company release and Schedule 14D-9. Bases differ (EV versus equity value; year-4 versus peak revenue; management projections versus consensus for Calliditas) and geography is not harmonized, so we use them only as a cross-check on where the market may price an approved MOLBREEVI (Section 3), not in our own value.

A5. Europe assumptions
InputBearBaseBull
Prevalence, patients per million71116
Eligibility70%
Peak ongoing share of eligible patients20%30%40%
Net price, % of US45%55%70%
Paid coverage92%
Launch lag after US approvalGermany 0.6 years, UK 1.0, France 2.0, Italy 2.0, Spain 2.5; a year later in the bear case
Regulatory timingUK decision guided for Q4 2026; CHMP opinion for Q1 2027

Prevalence and populations are model inputs, not a dated sourced table. Launch lags draw on EFPIA reimbursement data and are indicative.

A6. Sources

IMPALA-2 (publication); IMPALA (NEJM 2020); ERS aPAP guideline; Savara releases (May 2025 refusal to file, Dec 2025 resubmission, Q2 2026 update, Jun 2026 CFO transition); Savara 2025 10-K (manufacturing and supply) and Q2 2026 10-Q; Savara releases on IMPALA-2 exercise capacity (May 2026) and open-label extension data (Jun 2026); aPAP incidence (Inoue et al., AJRCCM 2008; StatPearls, Pulmonary Alveolar Proteinosis); Nobelpharma business reports (FY2024, FY2025) for Sargmalin sales; Leukine price from public pharmacy price listings (Drugs.com); FDA Form 483, Fujifilm Diosynth Biotechnologies Texas (Mar 2025); FDA first-cycle CR rates; Catalyst FY2025 release; company presentations (Jan and Aug 2026); Soleno 14D-9 (Apr 2026) and Neurocrine release; short interest.

Clinaptis Research. Modeled values, not observed data. Not investment advice.