Celcuity Inc. (NASDAQ: CELC) — First-and-Only in an Empty Market, on a Patent That Expires in 2029

Report date: 2026-08-01 Framing: Fundamental, competitive-advantage-driven, evidence-based. No price target and no buy/sell recommendation appears anywhere in the main analysis (Sections 1–15). The single, deliberately-labeled exception is the Claude’s Take block immediately below.

⚡ Claude’s Take

This block is the author’s own subjective opinion. It is general information, not investment advice, and should not be relied upon as such. The analysis in Sections 1–15 below deliberately carries no recommendation and no price target; this block is the one place a view is expressed.

Verdict: HOLD / AVOID-HERE — accumulate on weakness, and explicitly not a short. Celcuity has done the hard part: gedatolisib is approved, the label came out broader than the trial, NCCN gave it preferred Category 1 status sixteen days later, and the balance sheet was refinanced into 0.25% paper at a 40% premium with every covenant extinguished. That is close to a perfect execution quarter. But here is the arithmetic that decides it: at $86.38 the enterprise value requires roughly $6.7 billion of probability-weighted peak sales — and one hundred percent of the entire 37,000-patient US second-line population, at the label-implied revenue per patient, is only $4.2 billion. The approved indication cannot justify this price at any market share. The price is therefore underwriting first-line success at high probability, from trials that do not read out until end-2028 and 2030 — while the composition-of-matter patent expires in December 2029. My accumulation zone is roughly $45–65, implying an enterprise value of about $2.5–3.7 billion, or 1.4–2.0× the risk-adjusted global peak. Conviction: medium.

The framing is event-driven special situation, not momentum and not deep-value contrarian — and the tape supports that rather than merely decorating it. Roughly 89% of CELC’s return variance is idiosyncratic, no style factor survives the model’s penalty, and the momentum loading is actually negative (−0.60) despite a +120.6% twelve-month return, because that return arrived in two overnight gaps rather than a trend anyone could accumulate. The biotech industry factor was positive at every horizon out to a year while this stock fell 31% in a quarter, so the 40% drawdown from the May high is entirely company-specific: the ASCO mutant data on 2 June (−25.7%) and the approval-plus-launch-delay on 15 July (−17.6%). Add that Baker Bros. — the 19.99% holder — put 3.1 million shares, about a tenth of the float, out the door at $102.50 on approval day with no 10b5-1 flag, and that a third of the remaining float is short at 6.85 days to cover, and you have a security whose next move will be decided as much by the share register as by the revenue line.

What the market is getting right: the science, the regulatory de-risking, the guideline listing, and the balance sheet. What I think it is getting wrong is narrower and more specific. First, duration of therapy. The price embeds the trial’s 9.3-month median PFS; the approved label reports a median duration of exposure of 6.2 months, with 64% of triplet patients suffering a dose interruption and 12% discontinuing outright — against the “2.3% and 3.1%” management has quoted for two years. Both numbers are honest; only one is what oncologists read. That gap is roughly a third of the revenue model and it has been sitting in a public document since July. Second, the calendar. Nobody is talking about the fact that the big opportunity — first line, where the multi-billion outcomes live — reads out in 2028 and 2030, after the composition-of-matter patent has expired, with IRA negotiation landing in the same window. Third, the royalty. Pfizer takes a low-to-mid-teens cut of every dollar forever, which permanently caps the margin 13–15 points below any peer you would compare this to.

I am not short it and would not be, for three reasons: the asset is real and de-risked, the downside is cushioned by a saleable franchise rather than by cash (an approved, Category 1, wholly-controlled drug), and a strategic bid plainly exists — Novartis paid $2.0 billion upfront in March for a Phase 1/2 asset in this exact indication. Being short a good drug into a launch with 34% of the float already short is a poor way to be right. But note what is not underneath this stock: net cash is negative $30 million, $746 million of cash against $776 million of converts. There is no cash floor here, and a VIKTORIA-2 failure in 2028 would land on a leveraged balance sheet after another half-billion has been spent. And ninety days ago, in this exact line of therapy, Arvinas and Pfizer got vepdegestrant approved and then declined to commercialise it — Arvinas trades at a negative enterprise value. Two sophisticated owners have just priced this niche.

Conviction: medium. What flips me bullish: the first two full quarters of REVTORPYK revenue implying revenue per treated patient at or above roughly $160,000 — an eight-month-plus average course — which would mean real-world persistence is tracking the trial rather than the label. What flips me bearish: a mature VIKTORIA-1 overall-survival readout at or above a hazard ratio of 1.0, or any NCCN downgrade from Category 1; either would break the access story and the first-line program at once.

Tag: a genuinely good drug on a genuinely short clock.

📈 Stock Price Action — Five-Year Event Map

Celcuity spent four years as a forgotten ~$10–16 microcap, then compounded ~10x in ten months on two Phase 3 readouts, printed an all-time high of $144.98 on 2026-05-04, and has since given back 40.4% to close at $86.38 on 2026-07-31 — even as the news flow turned unambiguously better (positive PIK3CA-mutant data, then an FDA approval). The 52-week closing range is $39.16 (2025-07-31) – $144.98 (2026-05-04), and the all-time intraday high of $151.02 was set the same day, leaving the stock 42.8% below it on an intraday basis. Trailing returns: +120.6% over 12 months, but −31.3% over 3 months and −17.3% over 1 month, with the price below its 21-day ($92.31), 50-day ($99.84) and 200-day ($95.34) EMAs. The single largest one-day move in the stock’s history is +167.2% on 2025-07-28. (Prices: AZI adjusted daily series; all moves are FACT.)

Cycle narrative. (1) The April 2021 Pfizer license is the origin of the entire equity: Celcuity paid $5.0M in cash and 349,406 shares for exclusive worldwide rights to a shelved Pfizer molecule, and the market roughly doubled the stock in three sessions. (2) Then nothing happened for four years. Between mid-2021 and July 2025 the stock traded between roughly $7.55 and $30 with no trend while VIKTORIA-1 enrolled 701 patients and the share count crept higher — the ordinary purgatory of a single-asset developer between readouts. (3) On 2025-07-28 the wild-type cohort hit its primary endpoint and the stock rose 167% in a single session — the largest one-day move in its history and the moment CELC stopped being a microcap. (4) The ESMO presentation on 2025-10-18 converted “positive” into “how positive”: a hazard ratio of 0.24 against fulvestrant, the most favorable ever reported in a Phase 3 in this population. The stock added another 36% on 2025-10-20. (5) From there the tape simply extrapolated — NDA submission in November 2025, Priority Review with a 2026-07-17 PDUFA date, and positive mutant-cohort topline on 2026-05-01 — to an all-time high of $144.98 on 2026-05-04, a ~10.5x advance in nine months. (6) The turn came on positive data. The detailed mutant results at ASCO on 2026-06-02 were, on their face, excellent — an 11.1-month median PFS against an active comparator (alpelisib) rather than a placebo-like control — yet the stock fell 25.7%. INTERPRETATION: by June the price already discounted a best-case mutant dataset, and the doublet’s descriptive-only p-value and immature overall survival left nothing to upgrade. (7) Management issued $500M of 0.25% converts due 2032 into that drawdown (underwriting agreement 2026-06-03) — well-timed for the company, and further supply on the tape. (8) The approval itself, announced after the close on 2026-07-14 and three days ahead of the PDUFA date, produced a 17.6% decline the next session on roughly six times normal volume. INTERPRETATION: the swing factor was not the label — which is broad — but the disclosure that commercial launch slips to late Q3 2026 on drug-supply readiness, against a company that had spent two quarters telling investors it was launch-ready. Since then the stock has drifted to $86.38, roughly where it traded the day the converts priced.

1. Executive Summary

Celcuity is a one-drug company that just became a real one. On 2026-07-14, three days ahead of its PDUFA date and without an advisory committee, the FDA granted regular approval to REVTORPYK (gedatolisib) — a pan-PI3K/mTORC1/2 inhibitor licensed from Pfizer in 2021 for $5.0 million in cash and 349,406 shares — for HR+/HER2−, PIK3CA wild-type locally advanced or metastatic breast cancer after at least one line of endocrine therapy. Sixteen days later, before a single vial shipped, NCCN listed the regimen as a preferred Category 1 second-line option. The company has never recorded a dollar of revenue; commercial launch is guided to late Q3 2026.

The commercial logic is genuinely good. Roughly 37,000 US patients a year progress after a CDK4/6 inhibitor; about 22,900 are PIK3CA wild-type, and roughly 18,800 of them had no approved targeted option at all — every competing PAM-pathway drug (Piqray, Itovebi, Truqap) is biomarker-restricted away from this population. The evidence that the segment was empty is the trial’s own control arm: fulvestrant monotherapy delivered a median PFS of 2.0 months, against 9.3 months for the gedatolisib triplet (HR 0.24). The approved label is also broader than the trial — it requires only prior endocrine therapy, not prior CDK4/6 exposure — and needs no companion diagnostic. In the separate PIK3CA-mutant cohort, gedatolisib beat an active comparator, alpelisib, at 11.1 versus 5.6 months (HR 0.50), the first Phase 3 ever to show one PAM inhibitor superior to another; an sNDA follows in Q3 2026.

The business underneath it is weaker than the drug. Applying Greenwald’s taxonomy honestly, Celcuity has no durable firm-level competitive advantage: no customer captivity (prescriber switching costs are zero and NCCN publishes the comparison), no network effects, and negative economies of scale — a 155-person company detailing an intravenous drug against AstraZeneca’s bag, which carries Truqap and Enhertu and Datroway on the same call. What it owns is a time-limited, revenue-shared legal monopoly on one molecule. Market shares in this niche have turned over three times in seven years (Piqray → Truqap → Itovebi → REVTORPYK), which is the diagnostic signature of an industry with no barriers to entry.

Three structural facts govern the valuation and are under-discussed. (i) Gedatolisib is a legacy Wyeth/Pfizer compound: its US composition-of-matter patent expires in December 2029, 3.4 years after launch, reaching only ~December 2034 with a maximum Patent Term Extension — while the VIKTORIA-2 first-line trials that would justify a multi-billion-dollar franchise do not report until end-2028 and 2030, and IRA negotiation lands in the same 2033–2035 window. Effective commercial life is roughly eight to nine years, not sixteen. (ii) Pfizer takes tiered royalties in the low-to-mid-teens percent of net sales plus up to ~$330 million of remaining milestones, permanently capping the operating margin 13–15 points below an owned-asset peer. (iii) The approved label reports a median duration of exposure of 6.2 months against a 9.3-month median PFS, with 12% permanent discontinuation and 64% dose interruption — four to five times the “2.3% / 3.1%” figures management has quoted. Both sets of numbers are correct; the label is what prescribers read.

Financially, liquidity is strong and equity is nearly absent. FY2025 saw R&D of $145.0 million, SG&A of $27.2 million and a net loss of $177.0 million; Q1 2026 burned $55.1 million with SG&A tripling year over year on the commercial build. In June 2026 the company issued $575.0 million of 0.250% convertible notes due 2032 at a $124.53 conversion price — a 40% premium — and used part of the proceeds to pay off and terminate an 11–12.5% secured facility at ~$137.5 million, eliminating every financial covenant and cutting cash interest from ~$16 million to ~$7 million a year. Pro-forma liquidity is roughly $746 million, funding nine to thirteen quarters with nothing due before August 2031. Against that: the economic share count is up 433% since 2020, GAAP book equity is approximately zero, $776 million of convertible principal is unhedged (no capped call on either issue), an undrawn $400 million ATM — 8.4% of market cap — sits available at management’s discretion, and — the fact most often missed — net cash is negative $30 million, so there is no cash floor beneath the equity.

Capital allocation grades as mixed with a sharply improving recent record. The 2021 in-license was outstanding. The last twelve months of financing — equity at $38.00 three days after a +167% gap, converts at 40% premiums, an expensive covenanted loan retired — were top-decile. Against that, five years of pre-data financing (a December 2022 PIPE at $5.75 with 40% warrant coverage; October 2023 pre-funded warrants at $8.699) compounded the share base 38% a year, and management repriced options twice in four years, a practice the new 2026 plan bars only prospectively. Alignment is real where it counts — CEO Brian Sullivan owns 8.62% and has never sold a share — but the annual incentive plan contains no TSR, return or revenue metric, FY2025 CEO pay was $16.4 million at 94% option value, and Baker Bros. sold 3,100,000 shares at $102.50 on the day of approval for ~$318 million with no 10b5-1 flag.

On valuation, one calculation dominates. At a 13% discount and a royalty-adjusted 26.35% after-tax margin, the conditional value of the asset is roughly 0.80× peak sales; the ~$4.9 billion enterprise value plus ~$0.45 billion of remaining development spend therefore requires about $6.7 billion of probability-weighted peak sales against a bottom-up risk-adjusted build of ~$1.85 billion. But the decisive number is simpler and needs no probability weighting at all: one hundred percent of the entire 37,000-patient US second-line population, at the label-implied ~$113,500 of net revenue per treated patient, is $4.2 billion of peak sales — 37% below what the enterprise value requires, and below the requirement at every plausible discount assumption. The approved indication cannot support this price at any market share. The price necessarily embeds first-line success from trials reading out in end-2028 and 2030. Against that, the cross-sectional multiples the sector actually trades at put CELC at 2.0–2.65× street peak — low-to-middle of the 1.5–3.5× de-risked band, and in the upper half of the 2.0–4.0× EV/2031E band in whose peer set Celcuity is itself a named constituent. Expensive absolutely — the price is roughly 2.6× the probability-weighted NPV — and mid-to-rich relatively. Note also that no targeted agent has ever reached $1 billion in this setting: alpelisib peaked at $505M and is now declining, capivasertib is at $728M with its maker saying US share is already at peak, and elacestrant is rolling over at ~$650M.

What decides it is one observable number, and it arrives soon. The bull and bear cases pivot on the same hinge: real-world duration of therapy. Divide the first two full quarters of REVTORPYK revenue by patients on therapy. An eight-month-plus implied course validates the franchise thesis; a six-month course means no amount of label breadth or guideline positioning supports the current valuation. That number becomes visible in the Q3 2026 and Q4 2026 filings — roughly November 2026 and March 2027. Everything else in this memo is context for reading it.

2. Business Overview

2.1 What Celcuity is — and what it was

Celcuity Inc. is a Minneapolis-based, single-asset oncology company that, as of 2026-07-14, sells exactly one FDA-approved product: REVTORPYK™ (gedatolisib). Everything else in the enterprise — the trials, the salesforce, the balance sheet, the $4.2 billion of equity value — is a claim on that one molecule.

That is not what the company was built to be. Celcuity was founded in 2011/2012 around CELsignia, a cellular-analysis companion-diagnostic platform intended to identify patients with hyperactivated signaling pathways who would respond to targeted therapy. It IPO’d on Nasdaq in September 2017 at $9.50 per share and spent four years as a diagnostics company running investigator-style signaling assays. The diagnostics business never generated meaningful revenue and is not a source of value today; the FY2025 10-K describes a company whose entire operating purpose is developing and commercializing gedatolisib. FACT: Celcuity has never reported product revenue. TTM sales per share as of 2026-07-31 were $0.0034 — a rounding artifact, not a business.

The pivot came on 2021-04-08, when Celcuity in-licensed worldwide rights to gedatolisib from Pfizer. The terms are important because they determine how much of the franchise’s economics Celcuity actually keeps:

INTERPRETATION. As an acquisition of optionality this was a spectacular trade: $5M in cash and roughly $5M in stock bought an asset the market now values in the billions. As a durable business, it is more qualified. A low-to-mid-teens royalty is a permanent, uncapped tax on the top line that never amortizes away — at a $2 billion revenue level it is roughly $260–300 million a year flowing to Pfizer before Celcuity pays for its own salesforce, its own R&D, or its own debt. Any peer comparison that treats CELC’s future gross margin as equivalent to a company that owns its molecule outright is wrong by that amount, and the remaining ~$330 million of milestone obligations sits ahead of the shareholder in the cash-flow waterfall as well.

2.2 The product

Gedatolisib is a kinase inhibitor of all four class I PI3K isoforms (α, β, δ, γ) and both mTOR complexes (mTORC1 and mTORC2). The PI3K/AKT/mTOR (“PAM”) pathway is one of the most-attacked targets in oncology, and every previously approved agent hits a single node — PI3Kα (alpelisib, inavolisib), AKT (capivasertib), or mTORC1 (everolimus). Gedatolisib is the first approved agent to blockade the pathway comprehensively. That is a genuine scientific distinction, and it is the entire differentiation argument.

The commercial form of that distinction is less flattering. REVTORPYK is dosed at 180 mg as a 30-minute intravenous infusion, weekly on Days 1, 8 and 15 of every 28-day cycle — roughly 39 infusions per year, indefinitely, until progression. It is supplied as a lyophilized powder in a single-dose vial requiring reconstitution and dilution. Its approved competitors are oral pills. A patient choosing REVTORPYK is choosing three infusion-center visits a month over a tablet, and the regimen is only approved in combination with fulvestrant (itself an intramuscular injection) with or without palbociclib. Celcuity has a subcutaneous formulation in development — the first patent application was filed in 2026 — precisely because management understands this.

2.3 The approved indication and the label

FACT. On 2026-07-14, three days ahead of its 2026-07-17 PDUFA date, the FDA approved REVTORPYK for:

“adult patients with hormone receptor (HR)-positive, human epidermal growth factor receptor 2 (HER2)-negative locally advanced or metastatic breast cancer without a PIK3CA mutation detected following progression on or after treatment with at least one line of endocrine therapy in the metastatic setting” — in combination with fulvestrant, with or without palbociclib.

Two features of that wording matter more than the approval itself.

First, the label is biomarker-gated negatively. REVTORPYK is approved for patients without a PIK3CA mutation — roughly 60% of HR+/HER2− breast cancer per the company’s cited epidemiology. This is the inverse of every approved PAM-pathway competitor: Novartis’s Piqray (alpelisib) and Roche’s Itovebi (inavolisib) are restricted to PIK3CA-mutant tumors, and AstraZeneca’s Truqap (capivasertib) to PIK3CA/AKT1/PTEN-altered tumors. Celcuity therefore launches into a segment where, for the first time, there is an approved targeted option — and it is theirs. It still requires a PIK3CA test to establish wild-type status, so the drug is dependent on NGS or liquid-biopsy testing infrastructure, but that infrastructure already exists because the competitors created it.

Second, the label is broader than the trial. VIKTORIA-1 enrolled patients who had progressed on or after a CDK4/6 inhibitor and an aromatase inhibitor. The approved indication requires only progression after “at least one line of endocrine therapy in the metastatic setting.” INTERPRETATION: that is a meaningful regulatory win. It permits use in endocrine-pretreated patients who did not receive a CDK4/6 inhibitor and does not confine the drug to a strict post-CDK4/6i slot, widening the eligible pool relative to the studied population.

Third — and offsetting both — the safety section is not a footnote. There is no boxed warning, but the Warnings and Precautions are substantial and, in a real-world community-oncology setting, are the swing factor in whether patients stay on drug:

Prophylactic steroid-containing, alcohol-free mouthwash is required four times daily for a minimum of eight weeks. Administration is fussy in its own right: the drug must be diluted in 5% dextrose only — chloride-containing solutions including normal saline are prohibited — and run through a dedicated line with an in-line 0.2/0.22-micron filter.

FACT vs. FRAMING — the most important disclosure gap in this story. Celcuity has consistently described tolerability using treatment-related AE discontinuation rates of “2.3% and 3.1%” (10-K FY2025; CEO Sullivan on the Q1 2026 call: “2% and 3% adverse event-related discontinuation rates”). The approved label reports permanent discontinuation of REVTORPYK due to an adverse reaction of 12% (triplet) and 9% (doublet) — four to five times higher — alongside 64%/37% dose interruptions and 41%/22% dose reductions. Both figures are technically correct: the company quotes the investigator-attributed “treatment-related” subset; the FDA label uses all-cause adverse reactions per its own convention. But the label is the document oncologists read, and a median duration of exposure of just 6.2 months against a 9.3-month median PFS suggests real-world persistence will be materially shorter than the efficacy headline implies. Any revenue model built on the press-release framing overstates duration of therapy.

2.4 Segments, customers and how money will be made

There are no segments. There is one product, one geography (the United States), and — until late Q3 2026 — no sales at all.

• Revenue model: REVTORPYK is a physician-administered infusion and will therefore be sold through a buy-and-bill channel under the medical benefit (Medicare Part B, not Part D), not through specialty pharmacy. No permanent J-code existed at approval; interim billing runs through J9999 (physician office) and C9399 (hospital outpatient) with manual claims review until CMS assigns a permanent code, plausibly 2026-10-01 or 2027-01-01.
• Price: the WAC has not been disclosed. Management said only that REVTORPYK would be priced “at a premium to currently available therapies.” Sell-side estimates cluster at $24,000–28,000 per month (roughly $288,000–336,000 per year) — these are third-party estimates, not company guidance, and should be treated as such.
• Customers: community and academic oncologists treating second-line HR+/HER2− advanced breast cancer. Management sizes this at approximately 37,000 US patients progressing after a CDK4/6 inhibitor.
• Recurring vs non-recurring: revenue will be recurring in the sense that patients dose until progression, but the duration is bounded by median PFS — roughly 9.3 months in the approved wild-type population on the triplet. This is not a subscription business; it is a business that must continually replace its patients.
• Commercial infrastructure: Celcuity is going it alone in the US. As of the Q1 2026 call, the commercial organization was built and all oncology sales specialists were hired and onboarded, averaging 24 years in pharmaceutical sales and 16 in oncology. Ex-US, management intends to file with other regulators after the mutant sNDA but has not disclosed a partnering structure.
• Manufacturing: entirely outsourced. Celcuity relies on third-party CMOs for both drug substance and finished drug product — and the late-Q3 2026 launch delay was attributed by the CEO to ensuring “sufficient drug supply at launch.” For a single-product company, third-party supply is not a detail; it is the whole cost of goods and the whole delivery risk.

2.5 Verdict

Celcuity is not yet a business; it is a very well-validated product in search of one. The scientific asset is real, the label is broader than the trial that produced it, and NCCN listed the regimen as a preferred Category 1 option within sixteen days of approval — an unusually fast and unusually strong endorsement that materially de-risks payer coverage. But as of the report date the company has zero product revenue, burns roughly $55 million a quarter, pays a low-to-mid-teens royalty on every future dollar, sells an infusion against pills, and must manage a 72% stomatitis rate in the community. The next four quarters convert this from a clinical story into either a commercial franchise or a cautionary tale, and nothing in the historical financials tells you which.

3. Industry Dynamics

3.1 Sizing the opportunity from the bottom up

Celcuity’s own headline — “more than $5.0 billion” of second-line TAM and “peak revenue of up to $2.5 billion annually” — is a fully-successful-development number. It requires both the PIK3CA-mutant expansion and material first-line penetration. It is worth rebuilding from the epidemiology.

FACT. The single most persuasive piece of evidence that this segment was a genuine void is the VIKTORIA-1 control arm itself: fulvestrant monotherapy delivered a median PFS of 2.0 months (95% CI 1.8–2.3) in these patients. Whatever one thinks of fulvestrant as a comparator, a 2.0-month result is the signature of an untreated population.

Pricing and the peak-sales build. WAC is undisclosed. Approved comparators price at roughly $22,900–23,800 per 28-day cycle (Itovebi $22,867; Truqap ~$23,762; Piqray launched at $15,500 and now runs ~$19,600), and management has said REVTORPYK will price “at a premium.” Sell-side estimates cluster at $24,000–28,000 per cycle. Assuming a ~75% net-to-gross for a Part B infused oncology drug (340B erosion offset by the absence of Part D manufacturer discounts and PBM rebate ladders), and — critically — anchoring duration of therapy on the 6.2-month median exposure in the label rather than the 9.3-month median PFS:

Layering the mutant sNDA (approval ~2H 2027, less erosion from inavolisib moving into first line) adds $0.10B / $0.38B / $0.96B. The first-line VIKTORIA-2 opportunity, if both studies read out positive, adds $0 / $0.99B / $5.30B — but not before 2029 and 2031. Unadjusted US totals: $0.51B bear / $2.49B base / $8.36B bull; risk-adjusting the mutant expansion at ~85% probability of success and the first-line program at ~35%, the risk-adjusted US figure is roughly $1.79B.

INTERPRETATION. Management’s $2.5B is not fantasy — it lands close to our unadjusted base — but it is a number that assumes everything works. On the label as approved today, the honest base case is roughly $1.1 billion of US peak revenue, arriving several years out. That distinction is the crux of the valuation debate.

The empirical anchor that matters most is Truqap. AstraZeneca’s capivasertib — an oral drug, detailed by a global sales force, addressing roughly half of HR+/HER2− metastatic breast cancer — did $728M globally in FY2025 (+69%), with AstraZeneca stating US share was already “at peak” in the biomarker-altered second-line setting. Itovebi did CHF 36M in H1 2025 against Roche’s ~CHF 2B peak guidance. Piqray peaked around $373M. Evaluate forecasts gedatolisib at $2.1B by 2032. A reasonable reading: the entire second-line PAM-inhibitor pool is currently $1.2–1.5B globally, and a billion-dollar US franchise for REVTORPYK would mean taking most of it plus expanding it.

3.2 Is this a structurally good industry? No — with one qualification

Oncology drug development is structurally bad. The fully-capitalized cost per approved compound is roughly $2.6B in 2014 dollars (~$3.9B in 2024 dollars), and oncology has the lowest Phase-1-to-approval success rate of any therapeutic area at 3.4–6.7%. Aggregate industry returns barely clear the cost of capital and the distribution of outcomes is brutally skewed.

More damning in Greenwald’s terms: market shares in this niche are violently unstable. Piqray was the category leader in 2019; it was displaced by Truqap in 2023, then by Itovebi in 2024, and in 2026 it is the losing comparator arm in VIKTORIA-1. Share churn on that scale is the diagnostic signature of an industry with no barriers to entry. There is no customer captivity — oncologists switch on the next randomized trial, search costs are near zero because NCCN publishes the answer, and prescriber switching costs are literally zero. There are no network effects.

Where the profit pool actually sits. The economies of scale that make oncology commercialization a good business — sales reach across ~10,000 US medical oncologists, payer and GPO contracting leverage, ex-US infrastructure, portfolio-level risk pooling — belong to Big Pharma, not to single-asset developers. The scale of the asymmetry is worth stating: the first-line CDK4/6 inhibitor pool alone was $14.6B in 2025 (Verzenio $5.7B, Kisqali ~$4.8B, Ibrance $4.1B). The second-line PAM-inhibitor pool Celcuity just entered is roughly $1.2–1.5B — about 10% the size of the pool sitting one line ahead of it.

The qualification. Within a bad industry, the specific node Celcuity occupies is the best available position: a first-and-only approved therapy for ~18,800 US patients a year whose standard alternative delivered 2.0 months of PFS, with NCCN preferred Category 1 status, no gating companion diagnostic, and a tolerability profile that solves the class’s defining commercial problem.

3.3 Reimbursement: the IV question, answered properly

REVTORPYK is physician-administered and therefore falls under the medical benefit — Medicare Part B buy-and-bill — while every one of its targeted competitors (Truqap, Itovebi, Piqray, Orserdu) sits under Part D. This cuts both ways and the popular framing gets it wrong in both directions.

The tailwind, which is real and quantifiable. CMS retained ASP+6% for Part B drugs in the CY2026 Physician Fee Schedule (~ASP+4.3% after sequestration). At roughly $28,000 a cycle, that is about $1,680 of gross margin per patient per cycle flowing to the administering practice. A community oncology practice earns essentially nothing on an oral competitor dispensed through a specialty pharmacy. This is precisely Marathon’s agency relationship — the economic decision-maker is not the end consumer — and it is a genuine, quantified incentive at the point of prescribing.

The headwind, which is bigger. Part B carries 20% coinsurance with no statutory out-of-pocket cap. At a ~$336K annual list, a fee-for-service Medicare patient without supplemental coverage faces roughly $67,000 a year of exposure. Meanwhile the IRA’s Part D redesign capped patient out-of-pocket at $2,000 (2025), rising to $2,100 (2026) — so since 2025 the oral competitors have had a structurally better Medicare patient-affordability profile. Manufacturer copay assistance is prohibited for Medicare beneficiaries under anti-kickback rules, so REVTORPYK must route affordability through independent charitable foundations. Mitigants: roughly 90% of Medicare beneficiaries hold supplemental, Medicare Advantage or Medicaid coverage, and MA plans carry out-of-pocket maxima.

Net: roughly neutral-to-modestly-favorable, not the disaster the IV route first appears — but the patient-affordability edge flipped to the orals in 2025.

The J-code gap is a real, dateable launch drag. No permanent HCPCS J-code exists. Interim billing runs through J9999 (physician office) and C9399 (hospital outpatient), with MACs pricing claims manually from submitted invoices. A product-specific J-code is realistically 2026-10-01 at the earliest and more likely 2027-01-01. Smaller practices are reluctant to buy-and-bill a $28,000-per-cycle unclassified product on manual pricing. The late-Q3 launch delay partially masks this, and a January 2027 code would align with a Q4-26/Q1-27 revenue inflection.

IRA exposure is real but largely moot. As a small molecule, REVTORPYK runs the 9-year “pill penalty” clock rather than the 13-year biologic clock — selection around 2033 with a negotiated price applying around 2035 (the statutory reading admits a 2035/2037 alternative; either way it lands in the same window). The Small Biotech Exception is irrelevant — it lapsed after initial price applicability year 2027 and would void on acquisition anyway. The pill-penalty drag on NPV is roughly 3–6%. The reason it barely matters is uncomfortable rather than comforting: loss of exclusivity arrives in the same early-to-mid-2030s window, so the IRA clock largely overlaps generic entry rather than preceding it.

3.4 Competitive intensity, and the threat that isn’t a PI3K inhibitor

At the line level, second-line-and-beyond HR+/HER2− is crowded: roughly twelve approved options (generic fulvestrant, generic everolimus+exemestane, Piqray, Truqap, Itovebi, Orserdu, Enhertu, Datroway, Trodelvy, PARP inhibitors for gBRCA, chemotherapy, and now REVTORPYK), with at least six to eight late-stage programs behind them (imlunestrant, camizestrant, vepdegestrant, giredestrant, atirmociclib, culmerciclib, SNV-4818). At the node level — a PIK3CA-wild-type patient needing a targeted endocrine-backbone option after a CDK4/6 inhibitor — it was empty. Both statements are true simultaneously; the differentiation is real but narrow.

The principal structural threat is the ADC wave, not the PI3K competitors. Enhertu’s DESTINY-Breast06 label (US approval January 2025) covers HR+/HER2-low or HER2-ultralow after ≥1 endocrine therapy, chemotherapy-naive — the same line, no PIK3CA restriction — and HER2-low plus ultralow captures roughly 85% of HR+/HER2− tumors by IHC. Enhertu did ¥698.4B (~$4.5B) globally in FY2025, +26.3%. Datroway (mPFS 6.9 vs 4.9, HR 0.63) sits one line later. Gedatolisib does not compete with the ADCs on mechanism; it competes with them for sequence position. Realistically REVTORPYK is fighting for the “one more endocrine-based line before ADC or chemotherapy” slot — which supports adoption but caps duration of therapy, and duration of therapy is the single most sensitive input in the revenue build.

3.5 The capital cycle

FACT. XBI is +70.2% over the trailing twelve months and +10.3% YTD 2026. Q1 2026 saw the highest quarterly biotech IPO proceeds since 2021. Biotech M&A reached $106B across 201 transactions by early June 2026, with H1 totals of ~$149B across 49 deals and ~$46.1B of that in oncology; the median premium across 18 public deals was ~36%. Cancer and immune-disease developers took over 40% of all 2026 YTD funding.

Capital is flowing into this exact node. Novartis agreed on 2026-03-19 to acquire Synnovation’s pan-mutant-selective PI3Kα inhibitor SNV-4818 — a Phase 1/2 asset in HR+/HER2− metastatic breast cancer — for $2.0B upfront and up to $3.0B total. Lilly acquired Scorpion’s mutant-selective STX-478 for ~$2.5B. Relay’s RLY-2608 is approaching Phase 3.

INTERPRETATION (Marathon). The 2021–2024 bear market performed a genuine supply-side cull — small and mid-cap biotechs cut Phase 1/2 programs by roughly 35% between 2021 and 2023 — and REVTORPYK arrives into the scarcity that cull created, against a Big Pharma cohort facing a 2026–2032 patent cliff. That is the favorable half of the cycle and it is why the M&A bid is where it is. But the cycle has turned. A Phase 1/2 PI3Kα asset clearing $2B of cash upfront is a textbook boom-phase marker, and by Marathon’s logic the excess returns available to the marginal PAM-pathway entrant are already being competed away. Note also where the capital is going: every dollar of it is aimed at the PIK3CA-mutant node — precisely where Celcuity’s Q3 2026 sNDA is headed. The wild-type segment shows the opposite signature; capital avoided it for a decade because there was no biomarker to select on and generic everolimus was good enough.

Cycle rating: mid-to-late-stage capital inflow. Favorable for a seller today; unfavorable for a holder of the asset’s terminal 2030s economics.

3.6 Regulatory landscape

REVTORPYK received regular approval — not accelerated — on a PFS-only package with overall survival immature at 25% events, and no ODAC was convened. That is a strong outcome, and it is worth understanding why it happened, because the counter-example is only three months old.

On 2026-04-30, ODAC voted 6–3 against clinically meaningful benefit for AstraZeneca’s camizestrant in SERENA-6, despite a PFS hazard ratio of ~0.44 and mPFS of 16.0 vs 9.2 months; FDA subsequently extended the decision date. The objections were the artificiality of the ctDNA-triggered switch endpoint, PFS2 confounding, immature OS, absence of crossover, and thin patient-reported outcomes. INTERPRETATION: ODAC risk on a PFS-only breast-cancer package scales inversely with effect size and directly with endpoint artificiality. REVTORPYK’s combination of a 2.0-month control arm, a 7.3-month absolute PFS delta, and a hazard ratio of 0.24 on a conventional RECIST endpoint sat far outside that danger zone.

The mutant sNDA is a materially harder ask — HR 0.50 against an active comparator, into a segment with three approved options. Whether FDA convenes ODAC for it is an open question, with the head-to-head superiority design as the strongest mitigant.

REVTORPYK is also a Project Optimus success case, and this deserves credit. The 180 mg weekly three-of-four schedule plus mandatory prophylactic mouthwash was explicitly dose- and schedule-optimized, and the resulting hyperglycemia rate of ~9–11% against alpelisib’s ~64% is the direct product of that work. Alpelisib’s pre-Optimus maximum-tolerated dose is precisely why Piqray never scaled past ~$373M. The regulatory regime that raised development cost created gedatolisib’s principal commercial differentiator.

3.7 Verdict

Structurally bad industry; unusually good position within it; and the position is a window, not a franchise. Drug development confers only a transient, patent-based supply advantage, market shares in this niche have churned three times in seven years, and the profit pool belongs to the companies with scale — which is not Celcuity. Against that, the specific node is genuinely uncontested, the reimbursement mechanics are neutral-to-slightly-favorable rather than the disaster the IV route suggests, and the NCCN Category 1 listing removes most of the coverage risk. But the barriers are patent-based and expiring, the ADCs are compressing the line from above, competing capital is arriving at $2–3B a clip, and Pfizer takes a low-to-mid-teens royalty off the top. Value accrues to whoever converts this window into cash fastest — which structurally means either a rapid, well-executed US launch or a sale into the strongest oncology M&A market since before Covid.

4. Competitive Position

4.1 The mechanism, and why it is genuinely different

Gedatolisib inhibits all four class I PI3K isoforms (α, β, γ, δ) and both mTOR complexes (mTORC1 and mTORC2) at low-nanomolar potency. Every previously approved agent in the pathway hits one node: alpelisib and inavolisib hit PI3Kα, capivasertib hits AKT, everolimus hits mTORC1. The mechanistic argument is that single-node inhibition triggers compensatory feedback activation of the un-inhibited subunits — and the company’s own preclinical work (Rossetti et al., npj Breast Cancer, 2024) found gedatolisib ≥300-fold more potent on average across a 28-cell-line panel, uniquely cytotoxic rather than cytostatic, and equipotent in PIK3CA-mutant and wild-type lines.

That last point is the whole commercial thesis. If potency does not depend on the mutation, the drug can serve the 60% of patients that a mutation-selective inhibitor cannot.

But note carefully what this differentiation is and is not. It is a molecule-level distinction — and Pfizer invented the molecule. Celcuity’s contribution was recognizing its value, licensing it cheaply, and running the trial that proved it. That is a genuine act of skill. It is not a competitive advantage of the firm.

4.2 The head-to-head competitive set

Two honest readings of this table.

The favorable one. Every approved PAM-targeted agent other than generic everolimus is biomarker-restricted to alterations REVTORPYK’s label excludes. In PIK3CA-wild-type post-endocrine disease the only prior targeted option was a 2012-vintage generic. REVTORPYK is the first agent there with a positive randomized Phase 3 and NCCN preferred Category 1 status. And the tolerability trade is genuinely favorable where it counts: hyperglycemia at ~9–11% any-grade versus alpelisib’s 64–79% with 26% discontinuation — hyperglycemia being the toxicity that has commercially hobbled the entire PI3Kα class. Trading a metabolic toxicity requiring endocrinology co-management for a mucosal toxicity managed with a mouthwash is a real, physician-visible advantage.

The skeptical one. The wild-type control arm was a straw man. Fulvestrant monotherapy at 2.0 months is not what a US oncologist actually does post-CDK4/6i — everolimus combinations and ADCs are. There is no head-to-head against everolimus. The hazard ratio of 0.24 is measured against the weakest available comparator. Contrast the mutant cohort, where the comparator was an active approved drug and the hazard ratio moved to 0.50 — a far more honest number, and one that sits right on top of capivasertib’s 0.50 and alpelisib’s 0.65. The company’s own CEO made this comparison on the Q1 call and framed it as vindication; read the other way, it says that against a real comparator gedatolisib performs like a good drug in its class rather than a categorically different one.

4.3 The patent estate — the hard number that governs everything

This is the most under-discussed fact in the equity. Gedatolisib is a legacy Wyeth/Pfizer compound (PF-05212384 / PKI-587) whose priority dates long predate Celcuity’s involvement. The composition-of-matter claim expires before the NCE exclusivity does. Even with a maximum five-year Patent Term Extension, the molecule patent yields roughly 8.4 years of post-launch protection. Beyond ~2034 the franchise rests entirely on the formulation and method-of-use patents — meaningfully weaker and more challengeable protection. ANDA sameness requirements for parenterals give a cyclodextrin lyophilate more teeth than an oral tablet would have, but a 505(b)(2) alternative formulation or a Section viii carve-out of the breast-cancer method-of-use are both live routes. ASSUMPTION, not fact.

Stacked with IRA negotiation landing in the same 2033–2035 window, the effective commercial life of this asset is roughly eight to nine years, not the sixteen the 2042 patent superficially implies. And note the timing collision: VIKTORIA-2’s first-line readouts arrive end-2028 and 2030. If they work, the first-line franchise launches with only a fraction of its natural commercial life protected by composition-of-matter. Management understands this — it is exactly why a subcutaneous formulation program with a fresh patent application was announced in May 2026.

4.4 The commercial asymmetry

Celcuity is commercializing alone. No US or ex-US partner has been announced. The sales force was fully hired and onboarded as of May 2026 — averaging 24 years in pharmaceutical sales and 16 in oncology, sized to cover ~9,000 HCPs with focus on ~2,000 high-volume breast oncologists. Headcount has never been disclosed. Total company headcount is 155.

The asymmetry is stark. The representative calling on those same 2,000 physicians is AstraZeneca carrying Truqap and Enhertu and Datroway and (pending) camizestrant in one bag; or Lilly with Verzenio and Inluriyo; or Roche with Itovebi and the HER2 franchise. Celcuity carries one intravenous drug. In Greenwald’s framework this is a negative economy of scale — identical fixed distribution cost amortized over a fraction of the revenue.

Offsetting it: the Part B buy-and-bill margin (Section 3.3) gives the practice an economic reason to infuse, and NCCN Category 1 preferred status does much of the selling that a large sales force would otherwise have to do.

4.5 Verdict — no durable firm-level advantage

Running Greenwald’s taxonomy honestly:

VERDICT: Celcuity has no durable firm-level competitive advantage. It owns a time-limited, revenue-shared legal monopoly on one differentiated molecule in a segment that is currently uncontested but sits inside the most heavily capitalized oncology indication in the world. That is a valuable asset and a strong product position. It is not a moat around a business. The relevant test is whether a financial outcome would deteriorate absent the moat — and here the answer is that pricing on gedatolisib deteriorates on a known calendar date regardless of anything management does. Every competitive question after roughly 2034 must be re-won from zero, with a molecule the company neither invented nor wholly owns.

5. Growth History and Forward Opportunities

5.1 There is no growth history — only spending history

Celcuity has never recorded a dollar of product revenue in fourteen years of operation. The only thing that has compounded is cost: R&D from $25.8M (FY2021) to $145.0M (FY2025), a 54% annual rate, and SG&A from $2.6M to $27.2M, an 80% rate. Headcount went from 87 at end-2024 to 155 by Q1 2026. All of it was funded by issuing securities.

The single legitimate “growth” datapoint in the company’s history is not financial at all: the value of the gedatolisib asset, acquired in April 2021 for $5.0M in cash and 349,406 shares, and now the basis for a ~$4.7 billion market capitalization. That is an extraordinary return on an in-licensing decision. It is also not repeatable, not a business model, and not evidence about future revenue.

5.2 The forward opportunity, ranked by how much it is worth and when

1. Second-line PIK3CA wild-type — approved, launching late Q3 2026. ~22,900 US patients a year, of whom ~18,800 had no approved targeted option. Base case ~$1.1B of US peak revenue. This is the only revenue line that exists today, and the next four quarters of launch data will determine most of what the equity is worth.

2. Second-line PIK3CA mutant — sNDA planned Q3 2026, approval plausibly 2H 2027. ~14,100 US patients a year, less erosion from inavolisib pulling patients into first line. The data supporting it are strong — 11.1 vs 5.6 months and HR 0.50 against an active comparator, the first Phase 3 to beat another PAM inhibitor head-to-head. Base case adds ~$0.4B. Probability of approval is high (~85%), but ODAC risk is materially greater than for the wild-type filing, into a segment with three approved competitors.

3. First line — VIKTORIA-2, the real prize and the real wait. Study 1 (endocrine-resistant, n=440) topline by end-2028; Study 2 (endocrine-sensitive, n≈740) by 2030. Together they address the ~90,000 US women a year newly diagnosed with HR+/HER2− advanced disease. Base case adds ~$1.0B; the bull case is $5B+. The supporting evidence is a 31-patient single-arm Phase 1b reporting median PFS of 48.6 months and ORR 79% against a ~25-month historical benchmark for ribociclib plus letrozole. INTERPRETATION: that Phase 1b number is spectacular and close to uninterpretable. Thirty-one patients, no randomization, a cross-trial control, and a population selected by trial-entry criteria. Management deserves credit for choosing ribociclib as the comparator — the hardest available benchmark, with a demonstrated OS benefit — rather than something easier. But the readouts land in 2028 and 2030, against a composition-of-matter expiry in 2029 (~2034 with PTE). If first line works, it works with a truncated protected life.

4. Metastatic castration-resistant prostate cancer — optionality, not value. CELC-G-201 is a Phase 1b/2 with 38 patients, no randomization, a historical control, and no recommended Phase 2 dose set. Six-month rPFS of 67% and median rPFS of 9.1 months compare favorably to a ~40% historical benchmark, and tolerability was excellent (no G3 hyperglycemia, no DLTs, zero discontinuations for treatment-related AEs). Risk-adjusted, this is worth under $100M. There is also a negative read-across: AstraZeneca’s capivasertib succeeded in PTEN-deficient hormone-sensitive prostate cancer but failed a separate Phase 3 in prostate — PAM inhibition in this tumor is validated only in a narrow setting.

5. The subcutaneous formulation. In development, first patent application filed 2026, targeted to be available on roughly the same timeline as a potential first-line approval; FDA is expected to require a clinical-equivalence study. It addresses both the route disadvantage and the patent cliff. Correct strategy, years from delivery.

6. Ex-US. Roughly 30–40% of the global opportunity. Celcuity has no ex-US commercial infrastructure and no disclosed partner, while owing Pfizer a diligence obligation to commercialize in at least one international major market. With 155 employees, this is currently unmonetizable on a standalone basis and will require a partnership — or an acquirer.

5.3 Verdict

Quality of growth: unproven, front-loaded with risk, and structurally time-limited. The near-term opportunity is real, uncontested and NCCN-endorsed, but it is a single indication in a single country whose revenue depends on a duration of therapy the label suggests may be six months rather than nine. The large opportunity — first line — is genuinely large and genuinely uncertain, rests on a 31-patient Phase 1b, and does not report for two to four years, by which point the composition-of-matter patent has expired or is running on extension. What Celcuity has is not a growth curve; it is a sequence of dated binary events, each of which must be won, against a clock that started on 2026-07-14.

6. Financial Quality

6.1 There are no economics yet — only a cost structure

Celcuity has reported zero revenue in every fiscal year of its existence. Any discussion of “financial quality” is therefore a discussion of two things: how fast the money goes out, and how it was raised. Both are knowable and both matter.

Income statement, FY2021–FY2025 ($ thousands, per the 10-Ks):

Quarterly, most recent: Q1 2026 R&D $33.1M, SG&A $17.4M (from $6.4M a year earlier — $6.6M of the increase is commercial headcount and launch activity), total opex $50.5M, net loss $52.8M ($0.97/share), operating cash burn $55.1M. Headcount went from 87 at the end of 2024 to 155 by Q1 2026. The Q2 2026 10-Q is not yet filed.

INTERPRETATION. The shape of the cost curve is exactly what it should be. R&D peaked with VIKTORIA-1’s enrollment and is now flat-to-rolling-over quarter over quarter (Q1’26 $33.1M vs Q1’25 $32.2M) even as VIKTORIA-2 ramps; SG&A is where the money is going now, tripling year over year as the commercial organization was built. That is a company converting from a developer into a seller. The question the numbers cannot answer is whether $17.4M a quarter of SG&A is enough infrastructure to launch an infused oncology product against AstraZeneca and Roche — a full US oncology launch is normally a far larger commitment. Either Celcuity is running an unusually capital-efficient targeted launch into ~2,000 high-volume breast oncologists, or SG&A has a great deal further to climb. This is the single largest uncertainty in the 2027 burn.

6.2 Balance sheet: strong liquidity, essentially no equity

FACT. On 2026-06-08 Celcuity paid a Payoff Amount of approximately $137.5M and terminated the Oxford/Innovatus loan agreement outright, funded from the $575M convertible issue. The retired facility carried an effective rate in the 11–12.5% range and a 30% minimum-liquidity covenant (steppable to 50%). What replaced it is $575M of 0.250% unsecured paper due 2032, struck at a $124.53 conversion price — a 40% premium to the reference price — plus the existing $201.3M of 2.750% notes due 2031 struck at $51.30. Cash interest falls from roughly $16M a year to about $7.0M. There are now no financial covenants of any kind and no maturity before August 2031.

Runway. Against pro-forma liquidity of roughly $746M, the burn math is:

Management’s last public runway statement — “through 2027,” in the March 2026 10-K and repeated on the May call — predates the $575M raise and is stale and conservative. Celcuity does not have a financing problem in the window that matters.

But it does not have equity, either. Book value per share at 3/31/26 was $1.107, giving a P/B of roughly 78x at $86.38 — not the 45.1x the AZI screen reports, which is built on the stale FY2025 equity figure spread over a larger share base. Pro forma for the Q2 loss and the ~$9.6M extinguishment charge, GAAP book equity is approximately zero to slightly negative, because the last $776M of funding came in as debt and convertibles, which add liquidity but no equity. The honest statement is that P/B is a meaningless metric for CELC and that the screen’s 93rd-percentile reading understates rather than overstates how little asset backing sits under this price. Accumulated deficit stands at $501.7M.

6.3 Dilution — the real cost of the last five years

The economic share count is up 433% since the end of 2020 — a 38% annual compounding of the share base over 5.4 years. Add the if-converted shares from both notes (3.92M at $51.30 and 4.62M at $124.53) and the count reaches 63.5M; add all options, RSUs and remaining warrants and it reaches roughly 69.6M. On top of that sits an undrawn ~$400M ATM and a new 2026 equity plan with a 1%-per-year evergreen through 2036. Total equity overhang is approximately 9.6M shares, or 17.6% of the economic count.

6.4 Quality-of-earnings and accounting flags

There is no revenue to manipulate, so the flags are narrower — but not zero:

• R&D capitalization: none. All R&D expensed; license upfronts and milestones expensed on incurrence. Correct treatment. ✅
• Going concern: never. No substantial-doubt language appears in any 10-K, FY2021 through FY2025. ✅
• Clinical accruals — the one real soft spot. Prepaid clinical trial costs went $6.8M → $18.9M (+179%) in FY2025 while R&D rose 39%; accrued clinical trial costs were flat at ~$16.7M across the same period, then halved to $8.8M in Q1 2026. The accrual as a multiple of that quarter’s R&D fell from 0.50x (4Q24) to 0.27x (1Q26). This is consistent with VIKTORIA-1 completing enrolment and winding down — the benign explanation, and probably the right one — but management names “prepaid or accrued clinical trial costs” as one of only two critical accounting estimates, and a flat-then-halving accrual against rising expense is precisely the line item that goes wrong in biotech. OPEN QUESTION.
• Auditor. Boulay LLP of Minneapolis, engaged since 2017, re-ratified in 2026. The FY2025 opinion states “we determined that there were no critical audit matters.” A small regional firm auditing a ~$4.7 billion issuer and reporting zero CAMs on a company whose single largest estimate is the clinical accrual is a governance flag, not a fraud allegation.
• Supply commitments. Only $5.5M of non-cancelable clinical purchase commitments at 12/31/25, with clinical contracts “generally cancelable with advance notice.” No committed drug-substance or drug-product supply obligation is disclosed anywhere for a product that received FDA approval in July 2026 — and whose launch was delayed on drug supply. Either unusually flexible contracting or under-disclosure. OPEN QUESTION.
• Disclosure error. The Q1 2026 10-Q’s reclassification note describes the direction of a ~$2.5M R&D/SG&A reclass backwards relative to its own restated figures. Immaterial to totals; sloppy.
• One-timer to normalize. Q2 2026 will carry a ~$9.6M loss on extinguishment of the term loan.

6.5 Verdict

Do economics improve with scale? Unknowable — because there is no scale yet, and the structural ceiling is lower than it looks. What can be said: liquidity is strong and was made materially stronger eight weeks ago; the cost curve has the right shape, with R&D plateauing and SG&A stepping up; there has never been a going-concern qualification; and the accounting is conservative where it counts. What must be said against it: five years of financing compounded the share base 38% a year, GAAP book equity is now approximately zero, $776 million of convertible principal sits above the common, and every future revenue dollar carries a low-to-mid-teens royalty to Pfizer plus up to $330 million of remaining milestones. Celcuity’s steady-state gross margin will be structurally below that of a company that owns its molecule, by roughly the royalty. This is a strong liquidity position attached to a business whose profitability has never been demonstrated and whose eventual margin is capped by a contract it cannot renegotiate.

7. Capital Allocation

For a company that has never earned revenue, capital allocation is the management track record. There is nothing else to grade. Celcuity’s record splits cleanly into three decisions.

7.1 Decision one: the Pfizer in-license — outstanding, and the reason the company exists

In April 2021 Celcuity paid $5.0 million in cash and 349,406 shares for exclusive worldwide rights to gedatolisib for all diseases, exclusive “including as to Pfizer” — a shelved Wyeth/Pfizer molecule that had cleared Phase 1b and gone nowhere. Five years later that asset is an FDA-approved drug with NCCN preferred Category 1 status and is the sole basis for a ~$4.7 billion market capitalization. Celcuity may terminate for convenience on 90 days’ notice; Pfizer may not terminate for convenience at all.

This is among the highest-return in-licensing decisions in recent small-cap biotech, and it deserves to be said plainly. It required a genuinely non-consensus judgment — that comprehensive PAM-pathway blockade would work where a decade of single-node inhibitors had produced withdrawn indications and abandoned programs, and that the toxicity that killed the earlier pan-PI3K compounds could be engineered around with dosing and prophylaxis. Management was right.

The price of being right is permanent: tiered royalties in the “low to mid-teens” percent of net sales, running country-by-country until the later of twelve years from first commercial sale, expiry of all exclusivity, or expiry of the last valid claim — i.e. plausibly to 2042 — plus up to $335 million of milestones, of which $5 million was paid in January 2026 on NDA acceptance and an undisclosed approval milestone is very likely payable in Q3 2026. At $2 billion of revenue the royalty alone is roughly $260–300 million a year. Celcuity will never earn the margin of a company that owns its molecule.

7.2 Decision two: five years of financing — bad early, excellent late

Celcuity has raised approximately $1.389 billion across its life and repaid $137.5 million. The record divides sharply at 2025-07-28.

Before the data — capital raised on the buyers’ terms.

The December 2022 PIPE is the one that cost shareholders most. Roughly $100 million was raised at $5.75 with warrants over 6,956,450 shares attached at $8.05 — warrants that were later exercised for $42.1 million of cash in 2025 alone, against a market price running from $40 to $140. Add the October 2023 pre-funded warrants at $8.699 and roughly twelve million share-equivalents were issued below $9. That is the arithmetic behind a 433% increase in the economic share count since 2020 — a 38% annual compounding of the share base.

To be fair to management: these were the terms available. Celcuity was a pre-data, single-asset developer inside the worst biotech bear market since 2008, and it financed without ever incurring a going-concern qualification. But “the terms available” is an explanation, not an excuse — and the cost is permanent.

After the data — genuinely excellent execution.

The June 2026 refinancing is the single best capital-allocation decision in the company’s history and it is being under-appreciated. Celcuity replaced an 11–12.5% secured facility carrying a 30% minimum-liquidity covenant with 0.25% unsecured paper struck 40% above market, cut cash interest from roughly $16 million a year to about $7 million, extinguished every financial covenant, pushed the nearest maturity to August 2031, and walked away from $220 million of milestone-contingent Term E/F capacity it no longer needed. A company three weeks from an FDA decision, financing at a quarter of a percent, at a forty-percent premium, into a 26% drawdown, is doing something right.

The caveat that keeps this from being unambiguous: no capped call or bond hedge was purchased on either convert. The July 2025 loan amendment expressly permitted capped calls; none appears in the 8-K, the 424B5, the 10-K or the Q1 10-Q. On $776.3 million of convertible principal, the dilution is entirely unhedged. And the Jefferies ATM was upsized from $50 million to $400 million on 2026-01-09 alongside an automatic shelf, all of it still available — 8.4% of the current market cap, issuable at management’s discretion with no marketing period.

7.3 Decision three: commercialize alone rather than partner

Celcuity is launching an intravenous oncology product in the United States by itself, with 155 total employees, against AstraZeneca, Roche, Novartis and Lilly. It has no US partner and no ex-US partner, while owing Pfizer a diligence obligation to commercialize in at least one international major market.

INTERPRETATION. This is the highest-variance decision on the list and the jury is out. The case for it: the target universe is only ~2,000 high-volume breast oncologists, NCCN Category 1 preferred status does much of the selling, the buy-and-bill margin gives practices a reason to infuse, and retaining 100% of US economics on a potential billion-dollar franchise is worth far more than a partner’s upfront. The case against it: SG&A went from $6.4 million to $17.4 million a quarter and is still climbing, the launch has already slipped a quarter on supply, and roughly 30–40% of the global opportunity sits stranded in territories the company cannot reach. The launch delay is the first real evidence on this decision and it went the wrong way.

7.4 Incentives and governance — the ownership is real, the plan is not

The alignment evidence that counts is behavioral. CEO and co-founder Brian Sullivan owns 8.62% of the company, has never sold a share, put $1.5 million of his own money into the 2022 PIPE at $5.75, and exercised warrants in September 2025 and held. Neither the CSO nor the CFO has ever sold. Directors and officers together own 13.33%.

The plan design is weak. The annual incentive is “milestones, approved by our Compensation Committee, that advance our core business strategies,” at 40–60% of base salary. The milestones themselves are not disclosed, and there is no threshold/target/maximum scale. There is no total-shareholder-return metric, no return metric, and no revenue metric anywhere in the plan. FY2025 payouts were made at 40–60% of base “due to achievement of milestones target” — a statement no outside shareholder can audit. FY2025 CEO compensation was $16.4 million, of which 94.2% was option grant-date fair value, at a company with zero revenue — up 9.4x from 2023.

Two option repricings in four years is the clearest governance negative in the file. In May 2022 a “one-time stock option repricing” cut the exercise price of every outstanding option held by every officer and director to $5.50, cancelling strikes as high as $27.94. In January 2025 management repriced again, for 44 employees. The new 2026 Stock Incentive Plan — approved 2026-05-14, with 3,000,000 shares and a 1%-per-year evergreen running through 2036 — finally prohibits repricing without stockholder approval. That fix arrived after both repricings. Total equity overhang is roughly 9.6 million shares, or 17.6% of the economic share count. The March 2025 Change in Control & Severance Plan gives the CEO 3x base plus target bonus and full single-trigger vesting of all outstanding equity on a change of control — which matters given how live the M&A question is.

Grant timing, to management’s credit, is clean. The large 2025 NEO option grants were made on 2025-08-18 at a $51.57 strike — three weeks after the positive topline, on the post-gap price rather than the pre-gap price. No pre-announcement grant timing appears anywhere in the Form 4 record.

And then there is the selling. Two facts sit uncomfortably against management’s own conviction:

• Director David Dalvey sold 109,325 shares for ~$4.81 million on 2025-07-28 — the very day of the VIKTORIA-1 topline release — under a 10b5-1 plan, and a further 60,000 shares into the run to $140, for total 2025–26 sales of 169,325 shares / ~$14.0 million. His direct holding fell to 65,000.
• Baker Bros. Advisors — the largest shareholder at 19.99% — sold 3,100,000 shares at $102.50 on 2026-07-14, the day of FDA approval, for approximately $317.8 million. The Form 4 carries no 10b5-1 flag. That block cut Baker’s common position by 39% and represented roughly 10% of the entire 29.9-million-share float, executed 18.7% above today’s price.

INTERPRETATION. The Baker Bros. sale is a material fact and it cuts two ways. It is a mechanical, non-fundamental explanation for a meaningful part of the 17.6% decline on 2026-07-15 — a block that size has to clear somewhere. It is also the most informed seller in the security choosing the day of approval to take a third of a billion dollars off the table. Neither reading cancels the other, and both belong in the analysis.

Also worth noting: no open-market purchase has been made by any officer or director since 2023-05-22. Insiders have not bought this stock at $86, or at $145, or anywhere in between.

7.5 Verdict

Mixed, with a sharply improving recent record and a governance tail that has not been cleaned up.

The in-license was outstanding. The last twelve months of financing were top-decile — equity at the top of a gap, converts at a 40% premium and a quarter of a percent, and the retirement of an expensive covenanted facility. Ownership alignment at the executive level is genuine and unusual: a founder-CEO with 8.62% who has never sold a share is worth more than most compensation-committee language.

Against that: five years of pre-data financing compounded the share base 38% a year and permanently transferred a large share of the eventual outcome to PIPE investors and warrant holders; every future revenue dollar carries a low-to-mid-teens royalty to Pfizer; $776 million of convertible dilution is unhedged; $400 million of ATM capacity sits undrawn; the incentive plan contains no measurable metric; management repriced options twice; and the two largest informed sellers chose the topline day and the approval day to sell.

On the specific Section 7.6 question — has management allocated capital intelligently? — the answer is yes recently and no historically, and the recent record is the one that governs from here. The single most consequential allocation decision is still ahead of them: whether $746 million funds a self-commercialized launch that works, or funds an expensive proof that a 155-person company cannot out-detail AstraZeneca.

8. Changes and Headwinds — Last Two Years

Two years ago Celcuity was a $15 stock with a fully-enrolled Phase 3 and no data. Today it is an $86 stock with an approved drug, a salesforce, $675 million of convertible debt, and a third of its float sold short. Almost everything about the company changed, and the sequence matters.

Clinical and regulatory (the value creation).

Strategic and development changes. In May 2026 management redesigned VIKTORIA-2, the first-line program, in three ways: it added a second study covering endocrine-sensitive patients (≈60,000 of the ≈90,000 US women newly diagnosed with HR+/HER2− advanced disease annually); it dropped PIK3CA-status-based primary endpoints in favor of full intent-to-treat analyses, which cut Study 1’s sample size from 638 to 440 patients; and it set ribociclib-based regimens as the control arms. Topline for Study 1 (endocrine-resistant, 440 patients) is guided to end-2028; Study 2 (endocrine-sensitive, ~740 patients) to 2030. The design was pre-agreed with the FDA at a Type B meeting. Separately, Celcuity began developing a subcutaneous formulation and filed its first patent application on it in 2026.

INTERPRETATION on VIKTORIA-2. Expanding into endocrine-sensitive first-line is the difference between a $1–2 billion drug and a $4–5 billion drug, and choosing ribociclib as comparator is honest — it is the hardest available benchmark, with a demonstrated survival benefit and roughly 25 months of median PFS. But the readouts are 2028 and 2030, and the composition-of-matter patent expires December 2029. The first-line franchise, if it works, arrives with a fraction of its natural commercial life protected by the strongest form of patent. That timing mismatch is the least-discussed structural problem in this equity.

Capital structure changes (the dilution and leverage). The company financed aggressively into strength, which is the correct instinct, but the cumulative effect is a materially different balance sheet:

• July/August 2025 (within 72 hours of the +167% gap): equity at $38.00 (1,836,842 shares + 335,526 greenshoe + 400,000 pre-funded warrants; net $91.6M) and $201.3M of 2.750% converts due 2031 (net $194.9M) struck at a $51.30 conversion price.
• September 2025: $30.0M Term D drawn under the Oxford/Innovatus facility.
• June 2026 (the morning after the −25.7% ASCO session): $575.0M of 0.250% Convertible Senior Notes due 2032 — $500M base plus a fully-exercised $75M greenshoe; net ~$557.0M — struck at $124.53, a 40% premium. No capped call or bond hedge was purchased on either convert, so the dilution is unhedged.
• 2026-06-08: the entire Oxford/Innovatus term loan was paid off at a Payoff Amount of ~$137.5M and the loan agreement terminated. Cash interest falls from ~$16M/yr to ~$7.0M/yr, and every financial covenant — including the 30% minimum-liquidity test — is extinguished. Q2 2026 will carry a ~$9.6M loss on extinguishment.
• Economic share count (including the 6,147,787 pre-funded warrants, which the company itself counts in basic EPS) rose from ~10.3M at end-2020 to 54.9M — +433%, a 38% annual compounding of the share base over 5.4 years.
• Shareholders approved a new 2026 Stock Incentive Plan on 2026-05-14: 3,000,000 shares plus a 1%-of-shares-outstanding evergreen every year through 2036. It does, finally, prohibit option repricing without stockholder approval — after management had repriced twice (a blanket reset of every officer and director option to $5.50 in May 2022, and a 44-employee reset in January 2025).

Leadership and governance. The board expanded from seven to eight with the appointment of Charles (Chip) R. Romp in February 2026, and the company built out a full commercial leadership layer — a Chief Commercial Officer (Eldon Mayer) and a payer/national-accounts organization — over the preceding 24 months. Founder Brian Sullivan remains Chairman and CEO.

Headwinds.

1. The launch slipped. Management spent two quarters describing itself as launch-ready — “we have since hired and onboarded all of our oncology sales specialists,” Q1 2026 — and then, on approval day, guided commercial launch to late Q3 2026 on drug-supply readiness. Leerink’s Andrew Berens called the delay “not anticipated, given prior commentary on launch readiness.” This is the credibility event of the last two years, and it is why an approval produced a 17.6% decline.
2. Tolerability moved to the foreground. The approved label’s 72% stomatitis rate (22% Grade 3) is materially more prominent than the 2.3–3.1% discontinuation rates management emphasizes. Both are true; the label is what community oncologists will read.
3. Securities-litigation noise. Multiple plaintiff firms (Pomerantz, Kessler Topaz, Bragar Eagel, Kirby McInerney) issued “investigation” releases between 2026-07-20 and 2026-07-30. INTERPRETATION: these are near-automatic after a large single-day decline and carry little information; they are logged for completeness, not weighted.
4. The short base grew into the good news. Short interest rose from 8.85M shares (2025-12-31) to 11.15M (2026-07-15) — 34.4% of the 29.9M-share float, 6.85 days to cover.

Verdict. The last two years strengthened the asset and weakened the story’s margin for error simultaneously. Every clinical and regulatory question was answered favorably — twice, in two cohorts, plus a broad label and NCCN Category 1. Meanwhile the share count grew a third, $675 million of convertible debt was added, the first-line readouts were pushed to 2028/2030 against a 2029 composition-of-matter expiry, and management’s launch-readiness credibility took a real hit at exactly the moment credibility became the operative variable. On net: thesis-strengthening on the science, thesis-complicating on the execution and the capital structure.

9. Risk Analysis

Catastrophic-loss risk. Low in the next twelve months and non-trivial thereafter. With roughly $800 million of pro-forma liquidity against a ~$55–70 million quarterly burn, Celcuity is not going to run out of money soon. A total loss requires something closer to a safety withdrawal or a comprehensive commercial failure. The more realistic severe case is a 60–75% de-rating — which this security has done before (−76.3% peak-to-trough between September 2021 and May 2022) — if the launch curve disappoints across two or three quarters while $675 million of converts sits above the equity.

10. Valuation Discussion — Embedded Expectations

No price target and no recommendation appears in this section. The question here is only: what must be true for $86.38 to be the right price?

10.1 The capital structure, built by hand

Celcuity’s share count is wrong on every screen. The company has 6,147,787 pre-funded warrants exercisable at $0.001 that sit outside “shares issued and outstanding” but are economically shares — both the FY2025 10-K and the Q1 2026 10-Q state they are included in basic EPS “as the pre-funded warrants are exercisable for nominal consideration.” The arithmetic confirms it: Q1 2026 net loss of $52,841K ÷ $0.97 per share = 54.48M weighted basic = 48.35M common + 6.15M warrants.

The two convertible tranches — $776.3M of face value:

No capped call or bond hedge was purchased on either issue. The July 2025 loan amendment merely permitted capped calls; none appears in either 8-K, the 424B5, the 10-K or the Q1 10-Q — and the physical-settlement election on the 2031 notes is itself inconsistent with a hedge. The dilution is unhedged.

The enterprise value bridge:

We anchor on the hybrid ~$4.91B. The 2031 notes are 68% in the money and the company has irrevocably elected physical settlement — they are equity in economic substance, not a $201M cash liability. The 2032 notes are 31% out of the money and are a genuine cash claim unless the stock rises 44%. Screens quoting ~$4.2B understate the enterprise value by 16%.

A fact that matters more than it first appears: net cash is NEGATIVE ~$30 million — $746M of pro-forma cash against $776.3M of convertible face. Celcuity has no downside cash cushion. This is a genuine structural difference from the clinical-stage names it is often compared with; Apogee, for instance, carried roughly $1.26 billion of net cash (~$16.8 per share) into its takeout. We return to the consequence in Section 10.7.

Pro-forma cash of ~$746M is an estimate: $387.1M filed at 3/31/26, plus ~$557.0M of net convert proceeds, less the ~$137.5M term-loan payoff, less ~$60M of Q2 burn. The Q2 10-Q is not yet filed. The Oxford/Innovatus facility is terminated — no covenants, no minimum-cash test, cash interest down from ~$16M to ~$7.0M a year, with a ~$9.6M Q2 extinguishment loss to normalize out.

On the screens. ROIC.ai’s market cap of $5,518M and EV of $5,455M are marked to the 3/31/26 price and predate both the June convert and the term-loan payoff — rejected. AZI’s percentiles carry almost no information here: P/E is null (TTM EPS −$3.8444); P/S is undefined against zero product revenue; and P/B computed on the actual 3/31/26 equity of $53,525K over 48,347,390 shares is $1.107 per share — a P/B of ~78×, not the 45.1× AZI reports (AZI uses the stale FY2025 book of $100.6M). Pro forma for the Q2 loss, book equity is approximately zero to negative, because the last $776M of funding was entirely debt. P/B is not a meaningful metric here, and the 93rd-percentile reading understates rather than overstates the situation.

10.2 Revenue per patient — the correction that drives everything

This is the single largest correction to a naive model. The label reports a median duration of exposure of 6.2 months (triplet) / 5.7 (doublet) against a 9.3-month median PFS, with 12% / 9% permanent discontinuation for adverse reactions — not the 2.3% / 3.1% management quotes — plus 64% / 37% interruptions and 41% / 22% dose reductions. Modeling duration off median PFS overstates revenue per patient by roughly 50%.

10.3 Margin structure and the conditional-NPV constant

The Pfizer royalty is a permanent tax on the top line:

The NOL shield is real but thinner than it looks: federal NOLs of $83.3M and a deferred tax asset of $100.3M before a full valuation allowance; adding 2026 losses gives roughly $150M by year-end, which shields only about 1.2 years of peak EBIT. A Section 382 limitation is a live risk given the ownership churn.

Effective commercial life is roughly 8–9 years, not 16:

Deriving k, the present value per $1 of peak sales — two methods.

Method A — scale the standard constant. The conventional single-asset framework produces 0.795× peak on a 34% after-tax margin. Celcuity’s royalty-adjusted after-tax margin is 26.35%: 0.795 × (26.35/34) = 0.62. This is conservative but imports a clinical-stage timing drag into an asset that is launching now.

Method B — bottom-up, CELC-specific. At a 13% discount with a mid-year convention, valuation date mid-2026, launch late Q3 2026 and peak in 2032:

Adopted: k = 0.62 (low) / 0.80 (central) / 0.91 (high). Method B is the better estimate for an approved, launching asset, and we use 0.80 as the central case. Note this works against the bearish conclusion below — a higher k lowers the peak sales the price requires — and the conclusion survives it anyway.

Present value of remaining development spend: ~$0.45B. VIKTORIA-2 Study 1 (440 patients, topline end-2028) and Study 2 (~740 patients, by 2030) at $130–150M a year tapering, discounted at 13%, gives $0.52B gross and $0.33B incremental over maintenance; the present value of cumulative operating losses to breakeven is ~$0.41B. $0.45B is the blend.

10.4 The revenue build

¹ Approved, so technical probability is 100%; the 85% is a commercial discount.

The ~37,000-patient US second-line pool is a FACT (10-K FY2025, independently corroborated by an AstraZeneca 6-K of 2026-05-27), split roughly 60% wild-type / 40% mutant. This $1.85B risk-adjusted global peak converges with the independently-built $1.94B in Section 3.1 — two different routes to the same number.

10.5 The reverse-DCF — and the finding that settles the section

Solving EV + PV(development) = PoS × k × Peak:

$4.911B + $0.45B = $5.36B → required probability-weighted peak = $5.36B ÷ k

And here is the finding that makes the range academic:

One hundred percent of the entire 37,000-patient US second-line population, at $113,500 of net revenue per treated patient, is $4.20 billion of peak sales.

That is 37% below the $6.70 billion the current enterprise value requires — and below the requirement at every value of k in the plausible range, including the generous $5.89 billion.

Therefore: at $86.38, the approved indication cannot support the enterprise value at any market share whatsoever. The price necessarily embeds first-line success — VIKTORIA-2 Study 1 (topline end-2028) and Study 2 (by 2030) — at high probability. Nothing in the approved label or in the VIKTORIA-1 data speaks to that.

Expressed as penetration: the base second-line case (wild-type plus mutant) is $882M. The residual $5.82 billion must come from first line, ex-US and prostate at close to 100% probability — requiring roughly 20% share of first-line endocrine-resistant and 22% of first-line endocrine-sensitive, both trials succeeding, with no probability discount at all. At a blended $150K per patient that is ~44,700 patients a year — roughly 37% of the entire US HR+/HER2− advanced treated population across all lines of therapy.

Management’s own number, decoded. “Peak revenue of up to $2.5 billion annually” for second line alone, against a stated “>$5B second-line TAM,” is a 50% share of the entire 37,000-patient pool on management’s own $135K per patient — roughly 60% on ours.

10.6 The disconfirming base rate — no targeted agent has ever reached $1B in this setting

The verified year-three bracket in this indication is $400–750 million. Note also that the $505M Piqray “peak” is blended with the Vijoice/PROS indication from FY2022 — breast cancer alone peaked below it.

And a warning that is only ninety days old, in the same line of therapy. Arvinas and Pfizer’s VEPPANU (vepdegestrant) was approved on 2026-05-01 for second-line HR+/HER2− ESR1-mutant breast cancer. Neither partner will commercialize it — they are jointly seeking a third-party partner. Arvinas’s enterprise value is approximately negative $89 million. Two sophisticated owners have just told the market what they think a newly-approved second-line HR+/HER2− targeted agent is worth.

10.7 Scenarios

Probability-weighted: (0.30 × $9) + (0.45 × $20) + (0.25 × $85) ≈ $33 per share.

The sharpest single output of this section: on a strict 13%-discount risk-adjusted NPV, the bull case roughly justifies today’s price and offers little above it.

There is no busted-biotech floor here, and this is the most under-appreciated risk in the structure. The usual convention — cash at failure × 0.85, since busted biotechs historically transact at 0.7–1.0× net cash — assumes there is net cash. Celcuity’s net cash is negative $30 million. On a VIKTORIA-2 failure in 2028, cash would be roughly $400M against $776.3M of converts — the 2031 notes revert to being debt the moment they fall out of the money — leaving roughly $376 million of net debt, and that failure arrives only after another ~$500M has been spent. Late failure is the expensive kind, and here it lands on a leveraged balance sheet. What genuinely cushions the downside is not cash but the asset itself: an approved, NCCN Category 1, wholly-controlled franchise is saleable in a way a failed clinical program is not.

10.8 Comparable companies

Recently-approved and early-launch single-asset biotechs (enterprise values at 2026-07-31):

Two observations. The de-risked band is EV = 1.5–3.5× street peak, and CELC at 2.0–2.65× sits low-to-middle — which is the strongest argument the bulls have. But note the second: pre-approval oncology assets clear at the same multiples (COGT 2.0×, SRRK 2.7×, RLAY 3.0×). Celcuity is receiving no premium at all for being approved — which tells you the market’s peak-sales number, not its risk assessment, is doing the work.

Cross-check against the banker band. Centerview’s fairness opinion in the GSK/Nuvalent transaction ran a “selected public companies” analysis at EV / 2031E revenue of 2.0–4.0× — and Celcuity is itself a named constituent of that peer set, alongside Cogent, Crinetics, Cytokinetics, Insmed and Revolution Medicines. Precedent transactions ran at 3.0–6.0× five-year-forward revenue. CELC’s 2031E revenue on our base ramp is ~$1.75B, giving 2.8×; on a conservative ~$1.2B, 4.1×. Celcuity sits in the upper half to the top of the banker-sanctioned band.

M&A comparables — just-approved single- and dual-asset oncology and rare-disease takeouts:

Median roughly 3× street peak; premiums of 23–95% for approved single assets, with a 35–50% median. At the 3× median, our $1.85B risk-adjusted peak implies ~$5.7B of enterprise value, and management’s $2.5B implies ~$7.5B — the latter being roughly where the sell-side sits (11–12 analysts, targets $155–178, average ~$163.50).

Three caveats on the takeout thesis that are rarely stated. An acquirer must price the change-of-control conversion-rate adjustment on $776.3M of converts, and the March 2025 Change in Control & Severance Plan single-triggers full vesting of all outstanding equity. The academic base rate cuts against the lazy version: for small firms, mergers are primarily an exit for companies in financial trouble, and a higher Tobin’s q reduces the probability of being acquired — and Celcuity has just funded itself through launch with no distress catalyst. And most simply: a takeout is a scenario, not a valuation method.

10.9 The sector-calibration counter-test

A reverse-DCF that declares a stock several times too expensive should be checked against the sector, because the failure mode of a 13% discount rate is that it condemns everything. Running the identical back-solve on peers:

The absolute discounted-cash-flow method and the relative multiple disagree by roughly 4×, and the disagreement is sector-wide rather than specific to Celcuity. Every commercial-stage biotech in this comparison trades above a 13%-discount intrinsic value.

The reason is structural. The peak-sales-multiple convention implicitly assumes near-certain success, no discounting for the years to peak, and a long post-peak tail. For Celcuity, that third assumption is specifically broken by the December 2029 composition-of-matter expiry.

Conclusion: Celcuity is expensive on an absolute basis — the price is roughly 2.6× the probability-weighted risk-adjusted NPV — and mid-to-rich on a relative basis. Its gap is worse than Insmed’s and Travere’s, and far better than Relay’s.

Cross-read against comparable situations. The closest structural analogue is Travere: a single asset with a third-party royalty off the top (16% to Ligand/BMS versus Celcuity’s 13–15% to Pfizer) and a finite, legible exclusivity horizon. Travere Therapeutics trades at a $5.36B enterprise value on $536M of trailing revenue; Celcuity trades at $4.91B on zero revenue, and screens richer — comparable enterprise value, no commercial track record, a shorter runway (December 2029 versus April 2033), and a first-line dependency Travere does not carry. Against Insmed, Celcuity screens cheaper on EV/peak (2.0–2.65× versus 3.2–5.0×) — but Insmed has a real, observable, accelerating launch curve and no royalty burden, while Celcuity’s peak is entirely modeled. Against Apogee, which carried ~$1.26B of net cash into its takeout, Celcuity’s negative $30M makes the downside profiles simply not comparable.

10.10 Verdict on embedded expectations

What the market is underwriting correctly. The wild-type data are genuinely best-in-class (HR 0.24, mPFS 9.3 versus 2.0 months) and gedatolisib is the only approved targeted agent with a Phase 3 win in a population no competitor addresses. The mutant cohort beat alpelisib head-to-head, so the sNDA is high-probability. NCCN Category 1 preferred status removes the formulary obstacle. And the balance sheet funds the launch, with no maturity before August 2031 and no covenants.

What the market is underwriting incorrectly, in order of magnitude.

1. The price cannot be justified by the approved indication at any penetration. One hundred percent of the 37,000-patient second-line pool is $4.20 billion of peak sales against the $5.9–8.6 billion the enterprise value requires. First-line success is not optionality — it is load-bearing — and it reads out in end-2028 and 2030.
2. Duration of therapy. The label’s 6.2-month median exposure, 12% discontinuation, 41% dose reduction and 64% interruption are materially worse than the 2.3%/3.1% management quotes. A model built on the 9.3-month median PFS overstates revenue per patient by roughly 50%.
3. The patent cliff is 3.4 years away, not fifteen. Composition-of-matter expires December 2029 (~2034 with extension), and the first-line trials carrying most of the embedded value read out in 2028 and 2030 — so the first-line franchise would launch into the last few years of exclusivity. That timing mismatch is the genuine variant perception.
4. The Pfizer royalty — 13–15% of net sales plus ~$330M of remaining milestones — is a permanent ~14-point haircut that gross-margin comparisons to owned-asset peers miss entirely. It is what takes k from ~1.0 to ~0.80.
5. There is no cash floor. Net cash is negative $30 million against $776.3M of converts, and a 2028 failure arrives after another ~$500M has been spent.
6. Form factor. Thirty-minute infusions on Days 1, 8 and 15 of every 28-day cycle — ~39 a year, buy-and-bill under Part B, with no permanent J-code before 2026-10-01 or 2027-01-01 and heavy 340B exposure — competing against once-daily orals, with 72% all-grade and 22% Grade 3 stomatitis requiring prophylactic mouthwash for at least eight weeks.
7. No price has been announced. Every revenue estimate in the market, including ours, rests on a modeled $26,000–29,000 WAC.

One note on price formation. With 34.4% of the 29.9M float short at 6.85 days to cover, 95.7% institutional ownership, and Baker Bros. having sold 3,100,000 shares — roughly a tenth of the float — at $102.50 on approval day with no 10b5-1 flag, the quoted price carries less information than usual about what informed holders actually believe.

11. Variant Perception

11.1 What consensus believes

Consensus on Celcuity is not really a view about the science any more — VIKTORIA-1 settled that in two cohorts — it is a view about the slope of the launch curve. The sell-side position, as of the approval, runs roughly: gedatolisib is a genuinely differentiated, first-and-only option in an uncontested segment; NCCN Category 1 preferred status arrived unusually fast; the mutant sNDA roughly doubles the market; VIKTORIA-2 is the free option on a multi-billion-dollar first-line franchise; and Celcuity is an obvious strategic target into the strongest oncology M&A market since before Covid. Jefferies called the approval a “differentiated foothold” and flagged potential for “near-term rapid uptake.” H.C. Wainwright raised its target on approval. Evaluate put gedatolisib in its top-ten 2026 launches at $2.1B by 2032. Management’s own framing — “>$5 billion” of second-line TAM and “up to $2.5 billion” of peak revenue — anchors the upper end, and eleven to twelve covering analysts carry targets of $155–178, averaging roughly $163.50. Decoded, management’s number is a 50% share of the entire 37,000-patient second-line pool on its own revenue-per-patient assumption.

The dissenting consensus, which showed itself in the price on 2026-07-15, is narrower and more specific: the launch slipped a quarter on drug supply after two quarters of launch-ready messaging; the label’s 72% stomatitis rate is what community oncologists will actually read; and Leerink’s Andrew Berens put the crux plainly — “the ability of doctors to effectively manage the inflammation and keep patients on drug will be critical to commercial success.”

What the tape says about positioning. This is not a crowded momentum trade and it is not a value name. Roughly 89% of CELC’s return variance is idiosyncratic (specific volatility 72.5% annualized against 76.8% total; the best-fitting factor model explains only ~27% of variation), no style factor survives the model’s L1 penalty, and the momentum loading is negative (−0.60) despite a +120.6% twelve-month return — because that return arrived in two overnight gaps rather than an accumulable trend. Meanwhile the biotech industry factor was positive at 21, 63, 126 and 252 days while CELC fell 31% over three months, so the drawdown is entirely stock-specific. The book is crowded on both sides: 95.7% institutional ownership against short interest of 34.4% of the 29.9M float at 6.85 days to cover, and short interest has risen 26% since December. The honest description is a post-catalyst de-rating on a binary asset that has just turned commercial — with the first real launch datapoint set to move the stock violently in whichever direction it points.

11.2 The strongest bull case

A first-and-only drug in a genuinely empty segment, endorsed by the guidelines before it even launched, at a price that has already given back 40%.

1. The white space is real and unusually well-documented. Roughly 18,800 US patients a year — PIK3CA wild-type, no AKT1/PTEN alteration — had no approved targeted option. The proof is not a slide; it is the control arm: fulvestrant monotherapy produced a median PFS of 2.0 months. Every competing PAM-pathway agent is biomarker-restricted away from this population.
2. The label came out better than the trial. Approval requires only ≥1 prior line of endocrine therapy in the metastatic setting — not prior CDK4/6 inhibitor exposure, which the trial required — and caps no line of therapy. That widens the eligible pool by an estimated 15–25% over the studied population. There is also no gating companion diagnostic, so the drug is not held hostage to a test approval.
3. NCCN preferred Category 1 in sixteen days, before launch. In US oncology, guideline listing is the effective payer gate. This is the fastest, cheapest de-risking of commercial access available, and Celcuity got it before it shipped a vial.
4. The mutant data are better than the wild-type data on the metric that matters. The mutant cohort beat an active, approved comparator — alpelisib — at 11.1 vs 5.6 months, HR 0.50, ORR 48.9% vs 26.0%. This is the first Phase 3 in history to show one PAM inhibitor superior to another. An sNDA follows in Q3 2026.
5. The toxicity trade is favorable where the class has actually failed commercially. Hyperglycemia is what killed Piqray — 64–79% incidence with 26% discontinuation, requiring endocrinology co-management. Gedatolisib’s hyperglycemia runs ~9–11%; in the head-to-head, Grade 3+ hyperglycemia was 2.6% versus alpelisib’s 14.5%. Trading a metabolic toxicity for a mucosal one managed with a mouthwash is a real, physician-visible advantage — and it is a direct product of FDA’s Project Optimus.
6. The IV route is a commercial asset, not just a liability. Under Part B buy-and-bill at ASP+6%, a $28,000 cycle pays the administering practice roughly $1,680 of gross margin per patient per cycle. An oral competitor pays the practice nothing. That is a genuine agency-relationship incentive at the exact point the prescribing decision is made.
7. The balance sheet was fixed at the perfect moment. $575M of 0.250% paper struck at a 40% premium, used to retire an ~11–12.5% secured facility and extinguish every covenant. Roughly $746M of liquidity, nothing due before August 2031, and cash interest cut from ~$16M to ~$7M a year.
8. The strategic bid is live and expensive. Novartis paid $2.0 billion upfront in March 2026 for a Phase 1/2 PI3Kα asset in this exact indication. Roughly 77% of 2025 biopharma deal value went to commercial-stage or Phase 3 assets. Pfizer — which sold Celcuity this molecule — faces Ibrance’s 2027 loss of exclusivity and has publicly targeted eight blockbuster oncology medicines by 2030.
9. And the price has already corrected 40% from a May high, into an approval, with a third of the float short.

11.3 The strongest bear case

A six-month drug sold by a 155-person company against Big Pharma, on a patent that expires in 2029, with a low-to-mid-teens royalty off the top — priced as a franchise.

1. No targeted agent has ever reached $1 billion in this setting — and the direct analog is now in decline. Piqray, the approved PI3K-pathway drug in HR+/HER2− breast cancer, peaked at $505 million in FY2023 (a figure itself blended with the Vijoice/PROS indication), fell to $449M in FY2024 and $382M in FY2025, and Novartis has now reclassified it out of Oncology into “Established Brands.” Capivasertib is at $728M with AstraZeneca’s own H1-2026 disclosure saying it has “achieved peak share in 2L biomarker-altered mBC.” Elacestrant is rolling over at ~$650M, with its royalty receipts guided down. The verified year-three bracket in this indication is $400–750 million. 1b. And a warning that is ninety days old, in this exact line of therapy. Arvinas and Pfizer got VEPPANU (vepdegestrant) approved on 2026-05-01 for second-line HR+/HER2− ESR1-mutant breast cancer — and neither partner will commercialise it, having jointly gone looking for a third-party partner instead. Arvinas’s enterprise value is approximately negative $89 million. Two sophisticated owners have just told the market what they think a newly-approved second-line HR+/HER2− targeted agent is worth.
2. The label says six months, not nine. Median duration of exposure in the pivotal was 6.2 months (triplet) / 5.7 (doublet) against a 9.3-month median PFS, with 12%/9% permanent discontinuation, 64%/37% dose interruption and 41%/22% dose reduction. Management quotes 2.3%/3.1%. Both are technically true — the company uses the treatment-related subset, the FDA uses all-cause — but a revenue model built on the press-release framing overstates duration of therapy by something like a third, and duration of therapy is the most sensitive input in the entire build.
3. The wild-type hazard ratio was earned against a straw man. Fulvestrant monotherapy at 2.0 months is not US standard of care post-CDK4/6i; everolimus combinations and ADCs are. There is no head-to-head against everolimus. Note what happened when the comparator was active: against alpelisib the hazard ratio moved from 0.24 to 0.50 — right on top of capivasertib’s 0.50.
4. The patent is the problem nobody is discussing. Gedatolisib is a legacy Wyeth/Pfizer molecule. US composition-of-matter expires December 2029 — 3.4 years post-launch — and even with a maximum Patent Term Extension reaches only ~December 2034. Beyond that the franchise rests on formulation and method-of-use patents, which are weaker and challengeable, with IRA negotiation landing in the same 2033–2035 window. Effective commercial life is roughly eight to nine years, not sixteen. And the first-line readouts that would justify the current valuation do not arrive until end-2028 and 2030 — after the composition-of-matter patent has already expired.
5. Pfizer takes a low-to-mid-teens royalty forever, plus up to ~$330M of remaining milestones. At $2B of revenue that is $260–300M a year off the top, permanently capping the operating margin roughly 13–15 points below an owned-asset peer. Every peer multiple applied to CELC without that adjustment is wrong.
6. The ADCs are compressing the line from above. Enhertu’s DESTINY-Breast06 label covers HER2-low and ultralow after ≥1 endocrine therapy — the same line, no PIK3CA restriction — capturing ~85% of HR+/HER2− tumors by IHC, on ~$4.5B of global revenue. Gedatolisib is not fighting other PI3K inhibitors; it is fighting for sequence position against a $4.5-billion ADC.
7. A 155-person company is detailing an intravenous drug against AstraZeneca’s bag, which carries Truqap and Enhertu and Datroway on the same call. This is a negative economy of scale: identical fixed distribution cost, a fraction of the revenue to amortize it over. The launch already slipped a quarter on supply — the first hard evidence on this decision, and it went the wrong way. The J-code gap adds one to two more quarters of buy-and-bill friction at $28,000 a cycle.
8. The capital structure has compounded against holders, and there is no cash floor. The economic share count is up 433% since 2020. There is $776M of unhedged convertible principal (no capped call on either issue), an undrawn $400M ATM — 8.4% of market cap, issuable at will — and a new equity plan with a 1%-per-year evergreen through 2036. GAAP book equity is approximately zero and net cash is negative ~$30 million. The usual busted-biotech backstop — 0.7–1.0× net cash — does not exist here; on a 2028 VIKTORIA-2 failure the company would carry roughly $376 million of net debt, having spent another half-billion getting there.
9. The most informed sellers chose the best days. A director sold on the day of the topline release. Baker Bros. — the 19.99% largest holder — sold 3,100,000 shares at $102.50 on the day of FDA approval, for ~$318 million, on a Form 4 with no 10b5-1 flag, cutting its common position by 39%. No officer or director has bought a share on the open market since May 2023.
10. Overall survival is immature and the class has a history. VIKTORIA-1 OS was ~25% mature and never formally tested, and fatal adverse reactions occurred in 2.3% of patients in both gedatolisib arms. Idelalisib, duvelisib, umbralisib and copanlisib all lost indications; ODAC voted unanimously in 2022 that PI3K approvals in heme malignancies should require randomized data, citing poor OS in comparator arms. Celcuity’s own 10-K concedes that toxicity “played a significant role” in halting prior pan-PI3K/mTOR programs.

11.4 The three to five assumptions that actually matter

Strip everything else away and the outcome turns on five numbers, in descending order of leverage:

1. Real-world duration of therapy. Six months or nine? The entire revenue model swings roughly 50% between them, and the answer is already latent in the label.
2. Peak penetration of the wild-type pool. Fifteen percent, thirty, or forty? NCCN Category 1 and an empty segment argue high; a first-time commercial organization selling an infusion argues low.
3. Net price. Undisclosed. A $24,000 versus $28,000 cycle moves peak revenue ~15%, before a gross-to-net assumption that is itself unobservable and 340B-exposed.
4. Whether VIKTORIA-2 works, and whether it works in time — which on the arithmetic above is really assumption number one. First line is where the $5-billion outcomes live, and it is the only place the current price can come from. It reads out in 2028 and 2030 — one and four years after composition-of-matter expiry.
5. Whether exclusivity survives past 2034. A granted PTE plus a defensible formulation patent is the difference between an eight-year and a fifteen-year asset, and terminal value is where most of the disagreement about this stock actually resides, even though almost nobody discusses it.

11.5 Where consensus is most likely offsides

INTERPRETATION. Our best guess at the genuine variant perception is not that the bulls are wrong about the drug or that the bears are wrong about the risks. It is that both sides are arguing about the wrong time horizon.

The bull case is being underwritten on the label — its breadth, the guideline listing, the emptiness of the segment — all of which are real, dated, and already public. The bear case is being underwritten on the launch delay and the stomatitis rate — also real, also public, and largely a first-year issue. Both camps are arguing about the second-line launch, and the arithmetic says the second-line launch is not what is being priced. One hundred percent of the entire second-line population, at the label-implied revenue per patient, is $4.2 billion of peak sales against an enterprise value requiring $6.7 billion. Whatever happens in 2L — triumph or disappointment — the current price is a bet on first line, from trials reading out in end-2028 and 2030, against a composition-of-matter patent that expires in December 2029. A company whose large opportunity arrives after its strongest patent has expired is a fundamentally different asset from the one being debated, and the market is pricing the near-term launch with far more precision than it is pricing that collision.

Second, and more tractably: the positioning is doing work that the fundamentals are not. Roughly 10% of the float changed hands in a single block on approval day, at $102.50, from the largest and best-informed holder. A third of the remaining float is short into a launch that has not started. That combination means the next two prints will produce a move whose magnitude is a function of the share register rather than of the revenue number — which is precisely the condition under which a fundamental investor either gets paid for patience or gets run over by other people’s mechanics.

The falsifying evidence, for each side, is the same and arrives on the same date: the implied revenue-per-patient in the Q3 and Q4 2026 filings. Everything else in this section is prologue.

12. Fact vs. Interpretation

13. Open Questions

These are the things we could not resolve from public sources, ordered by how much they would move the analysis.

1. What is the actual wholesale acquisition cost? Undisclosed as of the report date. Every revenue figure scales close to linearly with it; the difference between $24,000 and $28,000 per cycle moves peak US wild-type revenue by roughly 15%. Management said only that pricing would be “at a premium to currently available therapies.” All figures used here are third-party estimates.

2. What will real-world duration of therapy actually be? The model is more sensitive to this than to penetration. The label’s 6.2-month median exposure, 64% dose-interruption rate and 12% permanent-discontinuation rate argue for well below the 9.3-month median PFS; NCCN Category 1 preferred status and a 2.0-month alternative argue the other way. The first two or three quarters of launch data will settle it.

3. What does the final VIKTORIA-1 overall survival analysis show? Immature at ~25% events and never formally tested. This is the single largest tail risk in the file — an adverse readout would simultaneously damage NCCN positioning, payer coverage, the mutant sNDA and the entire VIKTORIA-2 first-line program.

4. Will the FDA convene an ODAC for the PIK3CA-mutant sNDA? Materially higher risk than the wild-type filing: HR 0.50 against an active comparator, into a segment with three approved options. The camizestrant precedent (6–3 against, 2026-04-30) shows the committee is currently willing to reject PFS-only breast-cancer packages. The head-to-head superiority design is the strongest mitigant.

5. What actually caused the launch delay? Management said “adequate drug supply.” Celcuity discloses only $5.5M of non-cancelable purchase commitments and no committed drug-substance or drug-product supply agreement anywhere in its filings, for a product that has just been approved. A single-source CMO constraint at launch would be a material operational finding.

6. How much of the “wild-type” pool is genuinely wild-type rather than simply untested? NGS testing penetration in metastatic breast cancer was below 40% through 2023, and the label requires no FDA-authorized test — selection is by absence of a detected mutation. This is simultaneously a volume tailwind (a permissive label) and a real-world-evidence risk (misassigned mutant patients would underperform).

7. How large is the Pfizer FDA-approval milestone, and what are the royalty tiers? The 10-K aggregates up to $335.0M without splitting development from commercial tranches or disclosing sales thresholds, and describes royalties only as “low to mid-teens.” Both hit the launch P&L immediately and both are material to any discounted-cash-flow build.

8. When does the permanent HCPCS J-code arrive — 2026-10-01 or 2027-01-01? During the gap, MACs price claims manually from invoices, raising denials and slowing cash conversion. This determines the length of a genuine one-to-two-quarter launch drag.

8b. Where do Roche’s giredestrant (evERA, PDUFA 2026-12-18, not ESR1-restricted) and INAVO121 land relative to the launch window? This is the largest unmodelled competitive variable in the file. A non-biomarker-restricted oral SERD approved in December 2026 would compete directly for the second-line slot REVTORPYK has to itself today.

9. What is the ex-US strategy? Celcuity has no ex-US commercial infrastructure and no disclosed partner, while owing Pfizer a diligence obligation to commercialize in at least one international major market. Roughly 30–40% of the global opportunity is currently unmonetizable on a standalone basis.

10. How large is the sales force? Never disclosed — only the ~9,000-HCP universe and the ~2,000 high-volume target. Total company headcount is 155, which bounds it.

11. Was there a 10b5-1 plan behind the Baker Bros. sale of 3,100,000 shares at $102.50 on approval day? The Form 4 carries no flag.

12. Why did clinical accruals behave the way they did? Prepaid clinical trial costs rose 179% in FY2025 while accrued clinical trial costs stayed flat, then halved in Q1 2026 — accrual coverage falling from 0.50× of quarterly R&D to 0.27×. The benign explanation (VIKTORIA-1 winding down) is probably correct, but management names this as one of only two critical estimates, and the auditor reported zero critical audit matters.

13. Will a Patent Term Extension be granted on the composition-of-matter patent, and to which patent will Celcuity elect to apply it? Not disclosed. The difference between a granted PTE and none is roughly five years of the franchise’s protected life.

14. What Must Be True

The bull and bear cases here pivot on the same hinge: whether patients who start REVTORPYK stay on it long enough, in large enough numbers, to convert an uncontested label into a franchise. Everything else is second order.

The bull case — what must be true

The bear case — what must be true

The hinge, stated once. B1 and R1 are the same question asked from opposite sides, and they will be answered by arithmetic, not argument: divide the first two full quarters of REVTORPYK revenue by the number of patients on therapy. If the implied course is eight months or better, the bull case is intact and most of the bear case is noise. If it is six months or worse, no amount of label breadth or guideline positioning rescues the current valuation. That number becomes observable in the Q3 2026 and Q4 2026 filings — roughly November 2026 and March 2027.

15. Source Appendix

See Appendix B — Source Appendix below, which lists every primary and secondary source relied upon: the trailing 60-month SEC corpus (352 filings, CIK 0001603454), the REVTORPYK prescribing information, FDA and DailyMed records, the VIKTORIA-1 publications and conference presentations, SEER and Flatiron/Foundation Medicine epidemiology, CMS and IRA policy documents, competitor labels and financial disclosures, the AZI price series and FactorsToday factor model, and the published peer data consulted. It also records, for completeness, the sources sought and not obtained — chiefly the undisclosed wholesale acquisition cost, the split of the Pfizer milestone package, the royalty tiers, and the not-yet-filed Q2 2026 Form 10-Q.

APPENDIX A — Standard Diligence Questionnaire

Celcuity Inc. (NASDAQ: CELC) — 2026-08-01

Supplemental to the analysis above. Answers are labeled FACT / INTERPRETATION / ASSUMPTION / OPEN QUESTION where the distinction matters. Where a question does not map to a pre-revenue, newly commercial biotechnology company, the correct sector analog is given instead.

General

What thoughtful questions have other investors asked about this company?

The sell-side Q&A on the Q1 2026 call was unusually narrow and unusually revealing — nearly every analyst asked about the subcutaneous formulation rather than the launch. Andrew Berens (Leerink) asked for a precedent for converting an IV small molecule to subcutaneous and pushed on whether lower Cmax would reduce mucositis; Kalpit Patel (Wolfe) asked whether gedatolisib’s antitumor effect is Cmax- or AUC-driven; Brad Canino (Guggenheim) asked whether the formulation was complete and what device the injection volume implies; Gil Blum (Needham) asked whether subcutaneous dosing could change the schedule. INTERPRETATION: the smartest investors in the room had already concluded that the weekly-IV route and the stomatitis rate are the binding constraints on commercial value, and were probing how fast they can be removed. Management answered none of them with specifics.

The other genuinely thoughtful questions on record: Tara Bancroft (TD Cowen) asked what explains the wide historical range of alpelisib-plus-fulvestrant performance (5–7 months) across BYLieve and SOLAR-1 — i.e. is your comparator arm going to behave?; Stephen Willey (Stifel) asked how first-line share is influenced by long-term OS data and how VIKTORIA-2 Study 2 is powered against ribociclib, which has an OS win; and Silvan Turkcan (Citizens) asked about payer engagement and dossier timing. Chase Knickerbocker (Craig-Hallum) asked directly about launch readiness — and management said it had “identified a launch ready date that’s before PDUFA.” That answer did not survive contact with the approval.

The questions investors should be asking and largely are not: what is the actual WAC; what is the real-world discontinuation rate against the label’s 12%/9%; why is the composition-of-matter patent expiring in 2029; and why did Baker Bros. sell 3.1 million shares on approval day.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Neither — there are no earnings. FY2025 net loss was $177.0M; Q1 2026 net loss was $52.8M. The company has never recorded revenue. The correct analog for “cyclical position” is position in the asset lifecycle: Celcuity is at the single most favorable inflection point a development-stage biotech reaches — approval granted, launch imminent, all development risk on the lead indication retired — and simultaneously at the point of maximum execution risk, with zero commercial track record.

Driven by the external environment or internal actions? Overwhelmingly internal. The factor data quantify this: roughly 89% of return variance is idiosyncratic (specific volatility 72.5% annualized against 76.8% total), the All-Factors model R² is only 0.195, and the biotech industry factor was positive at 21d, 63d, 126d and 252d while CELC fell 31% over three months. The drawdown is company-specific, not sector beta.

How stable are revenues? Not applicable today. Prospectively, unstable by construction. Revenue will be a function of patients on drug at any moment, and each patient’s course is bounded by progression — median PFS 9.3 months on the triplet, but median duration of exposure in the pivotal was only 6.2 months. The entire patient base must be replaced roughly every year. This is the opposite of a subscription business.

Outlook for products/services? One product, one approved indication, launching late Q3 2026. Near-term catalysts: first revenue disclosure (Q3 2026 10-Q, ~November), the PIK3CA-mutant sNDA (filing Q3 2026, approval plausibly 2H 2027), permanent J-code assignment (2026-10-01 at the earliest, realistically 2027-01-01), and the final VIKTORIA-1 overall-survival readout. Long-dated: VIKTORIA-2 Study 1 topline by end-2028, Study 2 by 2030.

How big will this market be — growing, shrinking, domestic or international? The addressed market is domestic only today. US second-line HR+/HER2− post-CDK4/6i is roughly 37,000 patients a year, of whom ~22,900 are PIK3CA wild-type and ~18,800 had no approved targeted option before 2026-07-14. The market grows slowly with breast-cancer incidence (~322,000 new US female cases in 2026) but is being compressed from above by antibody-drug conjugates — Enhertu’s DESTINY-Breast06 label covers HER2-low and ultralow after ≥1 endocrine therapy, capturing ~85% of HR+/HER2− tumors by IHC, and did ~$4.5 billion globally in FY2025. International (EU5 + Japan) is roughly 1.17x US incidence and is currently unmonetizable — Celcuity has no ex-US infrastructure and no disclosed partner, stranding roughly 30–40% of the global opportunity.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More. At the line level, roughly twelve approved options already touch second-line-and-beyond HR+/HER2−, with six to eight late-stage programs behind them. More importantly, capital is flooding the specific node: Novartis paid $2.0 billion upfront (up to $3.0 billion) in March 2026 for Synnovation’s SNV-4818 — a Phase 1/2 pan-mutant-selective PI3Kα inhibitor in this exact indication; Lilly paid ~$2.5 billion for Scorpion’s STX-478; Relay’s RLY-2608 is approaching Phase 3. Every dollar of that is aimed at the PIK3CA-mutant segment Celcuity’s sNDA is about to enter.

How profitable is the business (ROIC, ROE)? Deeply negative and not meaningfully computable. Accumulated deficit $501.7M; stockholders’ equity $53.5M at 3/31/26 and approximately zero to negative pro forma after the Q2 loss and the ~$9.6M debt-extinguishment charge. ROIC and ROE fail the test decisively. The forward-looking constraint that matters is structural: a low-to-mid-teens royalty to Pfizer permanently caps achievable operating margin relative to a company that owns its molecule.

How profitable is the industry — how many competitors, what barriers to entry? Oncology drug development is structurally unattractive: roughly $3.9 billion (2024 dollars) fully capitalized per approved compound, the lowest Phase-1-to-approval success rate of any therapeutic area at 3.4–6.7%, and aggregate returns that barely clear the cost of capital. Barriers to entry are patent-based and therefore expiring, not structural. The decisive evidence is market-share instability: Piqray led this category in 2019, was displaced by Truqap in 2023 and Itovebi in 2024, and in 2026 is the losing comparator arm in VIKTORIA-1. Meanwhile the profit pool sits with scale players — the first-line CDK4/6 inhibitor pool alone was $14.6 billion in 2025 against a second-line PAM-inhibitor pool of roughly $1.2–1.5 billion.

Can the business be easily understood? Yes, unusually so. One molecule, one approved indication, one country, one royalty obligation, one patent cliff with a known date. The difficulty is not comprehension but forecasting: nearly all the value depends on two unobservable quantities — real-world duration of therapy and penetration — plus one unresolved scientific question (mature overall survival).

Can it be undermined by foreign low-cost labor? No. Manufacturing is 100% outsourced to CMOs and cost of goods is a trivial fraction of a $300,000-a-year drug. The relevant analog is generic entry, which is dated: composition-of-matter expires December 2029 (~December 2034 with a maximum Patent Term Extension), NCE exclusivity runs to July 2031 with Paragraph IV filings permitted from ~July 2030.

Do brands matter? No. Physicians prescribe on data and guidelines. The functional equivalent of a brand here is the NCCN preferred Category 1 listing, granted 2026-07-30 — sixteen days after approval and before launch. That is the single most valuable non-patent asset the company owns, and it is revocable if the evidence base deteriorates.

What is the nature of competition? Evidence-based and winner-take-segment within a biomarker slot. Competition is not on price, service, or relationship but on randomized data and guideline position. It is also sequential rather than head-to-head — gedatolisib’s real fight is for position in the treatment sequence against the ADCs, not for share against other PI3K inhibitors.

Customers’ switching costs? Essentially zero for the decision-maker. An oncologist can write a different regimen for the next patient at no cost; NCCN publishes the comparison; search costs are nil. There is genuine captivity at the individual patient level — you do not switch a responding patient — but with a 6.2-month median exposure, that annuity lasts under a year. That is a revenue characteristic, not a barrier to entry.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? Yes, and they are the whole company. The gedatolisib license is carried at essentially nothing — the $5.0M upfront was expensed to R&D on incurrence, as were all subsequent milestones. No intangible for the approved product appears on the balance sheet. GAAP book equity of ~$53.5M (approximately zero pro forma) against a ~$4.7B market capitalization is the measure of how much economic value sits off the balance sheet. Also unrecognized: the CELsignia diagnostics platform (6 US patents, 30 international, earliest expiry 2033), now dormant — the word does not appear once in the FY2025 10-K — and $501.7M of accumulated deficit representing NOLs whose deferred tax asset is presumably fully valuation-allowanced.

Off-balance-sheet liabilities? Modest but real. Up to ~$330 million of remaining Pfizer milestones (an approval milestone is very likely payable in Q3 2026; the amount is not disclosed — OPEN QUESTION), and the low-to-mid-teens royalty on all future net sales, which is an unrecognized claim on future revenue running plausibly to 2042. Operating leases are immaterial ($74K of remaining minimum payments). Purchase commitments are only $5.5M, all clinical and largely cancelable. OPEN QUESTION: no committed drug-substance or drug-product supply obligation is disclosed anywhere for a product that launched in July 2026 — and whose launch was delayed on drug supply.

How conservative is the accounting? Broadly conservative, with one soft spot and one governance flag. Conservative: all R&D is expensed with no capitalization; license upfronts and milestones are expensed on incurrence; there has never been a going-concern qualification in any 10-K from FY2021 through FY2025. Soft spot: prepaid clinical trial costs rose 179% in FY2025 ($6.8M → $18.9M) while R&D rose 39%, and accrued clinical trial costs were flat at ~$16.7M and then halved to $8.8M in Q1 2026 — accrual coverage falling from 0.50x of quarterly R&D to 0.27x. Management names clinical accruals as one of only two critical estimates. The benign explanation (VIKTORIA-1 winding down) is probably correct, but it warrants a direct question. Governance flag: the auditor is Boulay LLP, a small Minneapolis firm engaged since 2017, whose FY2025 opinion states “we determined that there were no critical audit matters” — zero CAMs at a ~$4.7 billion issuer whose largest estimate is the clinical accrual. Separately, the Q1 2026 10-Q states the direction of a ~$2.5M expense reclassification backwards relative to its own restated figures.

How CapEx-hungry is the business? Barely at all in the conventional sense — manufacturing is fully outsourced and leases are immaterial. The economic equivalent of capital expenditure is clinical development: R&D of $145.0M in FY2025, with VIKTORIA-2’s two studies (440 and ~740 patients across ~200 sites) running to 2028 and 2030. Working capital will become a genuine consideration at launch: buy-and-bill inventory plus the J-code gap, during which MACs price claims manually from invoices, will slow cash conversion for one to two quarters.

Capital Allocation & Management

How much FCF does the business generate, how does management use it, what is the philosophy? None — free cash flow is deeply negative. Operating cash burn was $153.3M in FY2025 and $55.1M in Q1 2026 (trailing twelve months: $172.5M). The philosophy is straightforward and, of late, well executed: raise opportunistically into strength, spend it on clinical development and now commercial build, and retain 100% of the asset. Pro-forma liquidity of roughly $746 million funds 9–13 quarters depending on launch burn, with no financial covenants of any kind and no maturity before August 2031.

Significant acquisitions recently? No corporate acquisitions ever. The one transaction that matters is the April 2021 in-license of gedatolisib from Pfizer for $5.0M cash plus 349,406 shares — arguably the highest-return capital-allocation decision in recent small-cap biotech, and the reason the company exists in its present form.

Buying back shares? No, and it would be inappropriate. The company has never generated cash and has never repurchased stock.

Issuing large amounts of new shares to insiders? Yes, materially. FY2025 stock compensation was $21.4M (up from $7.0M), including an $8.4M market-condition option tranche that vested when a stock-price threshold was hit. Options outstanding at 3/31/26 were 5,820,708 at a $20.37 weighted-average exercise price, with intrinsic value of roughly $468M at year-end 2025. The new 2026 Stock Incentive Plan adds 3,000,000 shares plus a 1%-of-shares-outstanding evergreen every year through 2036. Total equity overhang is roughly 9.6 million shares — 17.6% of the 54.9-million economic share count.

Compensation policy of directors/management? FY2025 CEO compensation was $16,373,868, of which 94.2% ($15.4M) was option grant-date fair value, at a company with zero revenue — 9.4x the 2023 figure. CSO $4.19M; CFO $3.41M; non-employee directors $140,000 each ($40,000 cash + $100,000 options). The annual incentive is “milestones, approved by our Compensation Committee, that advance our core business strategies,” targeted at 40–60% of base salary — the milestones are not disclosed, there is no threshold/target/maximum scale, and there is no TSR, return, or revenue metric anywhere in the plan. FY2025 payouts were made at 40–60% of base “due to achievement of milestones target,” which no outside shareholder can verify.

Two option repricings in four years is the clearest governance negative in the file. In May 2022 a “one-time stock option repricing” reset the exercise price of every option held by every officer and director to $5.50, cancelling strikes as high as $27.94. In January 2025 management repriced again for 44 employees. The 2026 Plan finally prohibits repricing without stockholder approval — after both events. The March 2025 Change in Control & Severance Plan gives the CEO 3x base plus target bonus and full single-trigger vesting of all outstanding equity on a change of control.

To management’s credit, grant timing is clean: the large 2025 grants were struck at $51.57 on 2025-08-18, three weeks after the positive topline — on the post-gap price, not the pre-gap price. No pre-announcement grant timing appears in the Form 4 record.

Motivations of management? Ownership, more than plan design. Co-founder and CEO Brian Sullivan owns 8.62% of the company, has never sold a share, put $1.5 million of his own money into the December 2022 PIPE at $5.75, and exercised warrants in September 2025 and held. Neither the CSO nor the CFO has ever sold. Directors and officers own 13.33% in aggregate.

Set against that, two informed sellers chose the best possible days. Director David Dalvey sold 109,325 shares for ~$4.81M on 2025-07-28, the day of the VIKTORIA-1 topline, and 169,325 shares for ~$14.0M in total through May 2026. Baker Bros. Advisors — the 19.99% largest shareholder — sold 3,100,000 shares at $102.50 on 2026-07-14, the day of FDA approval, for approximately $317.8 million, on a Form 4 carrying no 10b5-1 flag; that block was roughly 10% of the entire float. And no officer or director has made an open-market purchase since 2023-05-22.

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? None of these. Celcuity Inc. is a Delaware corporation (converted from a Minnesota LLC on 2017-09-15) with ordinary common stock listed on Nasdaq under CELC. Ordinary Form 1099 tax reporting; no K-1.

Dividend policy? No dividend has ever been paid and none should be expected. The company has no earnings, no free cash flow, and a $501.7M accumulated deficit.

How profitable is the business? Not profitable, and not close. FY2025 operating loss $172.2M on zero revenue. Even on a fully successful base case, profitability is several years away and permanently capped below owned-asset peers by the Pfizer royalty.

Is net income diverging from cash from operations? Modestly and explicably. FY2025 net loss $177.0M against operating cash burn of $153.3M — the gap is stock compensation ($21.4M) and non-cash interest. Q1 2026 inverted: a $52.8M net loss produced $55.1M of operating cash burn, because accrued clinical costs fell $8.1M and other accruals fell $6.9M. INTERPRETATION: cash burn running ahead of net loss on falling accruals is worth watching, but here it is consistent with a Phase 3 winding down rather than with earnings management. A one-time ~$9.6M loss on extinguishment of the term loan will hit Q2 2026 and should be normalized out.

Market data. At $86.38 (2026-07-31): market capitalization on the 54,914,075 economic share count (48,766,288 common plus 6,147,787 pre-funded warrants, which the company itself includes in basic EPS) is approximately $4.74 billion; the honest hybrid enterprise value — treating the deeply in-the-money 2031 notes as equity and the out-of-the-money 2032 notes as debt — is approximately $4.91 billion. Book value per share is $1.107 (P/B ~78x), heading to approximately zero pro forma; P/B is not a meaningful metric for this company. Float is 29.91 million shares against 95.7% institutional ownership and short interest of 11,152,374 shares — 34.4% of float, 6.85 days to cover. Average daily dollar volume is ~$166 million over 90 days; liquidity is not a constraint.

Risks & Downside

What factors would cause the stock to decline? In rough order of probability-weighted impact: (1) a slow launch — first revenue disclosure in the Q3 2026 10-Q against a Street model of only ~$36M for 2026, with the J-code gap and manual MAC pricing as a one-to-two-quarter drag; (2) real-world persistence below trial — the label’s 12%/9% permanent discontinuation, 64%/37% dose interruption and 6.2-month median exposure against a 9.3-month median PFS, with 72% stomatitis to manage in community practice; (3) a flat or adverse final overall-survival readout from VIKTORIA-1, which would simultaneously damage NCCN positioning, payer coverage, the mutant sNDA and the entire VIKTORIA-2 program; (4) ADC encroachment compressing the endocrine-therapy window from above; (5) VIKTORIA-2 failure in 2028/2030; (6) further dilution from the undrawn $400M ATM or unhedged conversion of $776M of notes; (7) positioning unwind — 34.4% of float short against 95.7% institutional ownership means the first hard launch datapoint moves the stock violently in whichever direction it points.

Risk of a catastrophic loss? Low over the next twelve months; non-trivial thereafter. With roughly $746 million of liquidity, no covenants and no maturity before 2031, insolvency is not a near-term scenario. The realistic severe case is a 60–75% de-rating if the launch curve disappoints across two or three quarters while $776 million of convertible principal sits above the common — and this security has done exactly that before, falling 76.3% peak-to-trough between September 2021 and May 2022.

Chance of a total loss? Very low in any reasonable horizon. Total loss would require something close to a safety withdrawal or a comprehensive commercial failure combined with exhaustion of the cash pile — and even in severe scenarios the approved asset, NCCN Category 1 listing and worldwide rights retain acquisition value. The relevant M&A context: ~77% of 2025 biopharma deal value went to commercial-stage or Phase 3 assets, and a Phase 1/2 PI3Kα program in this exact indication commanded $2.0 billion of cash upfront from Novartis in March 2026.

Recent News & Events

Has the business environment changed recently? Fundamentally, and twice in three months. On 2026-07-14 the FDA approved REVTORPYK (gedatolisib) — three days ahead of the PDUFA date, on a regular (not accelerated) approval, with no ODAC convened — converting Celcuity from a clinical-stage developer into a commercial-stage company. On 2026-07-30 NCCN added the regimen as a preferred Category 1 second-line-and-subsequent option, sixteen days after approval and before launch. Against those, commercial launch slipped to late Q3 2026 on drug-supply readiness, and the stock fell 17.6% on the day after approval on roughly six times normal volume.

Significant acquisitions? None by Celcuity. In the surrounding competitive environment: Novartis / Synnovation ($2.0B upfront, up to $3.0B, March 2026) for a Phase 1/2 PI3Kα inhibitor in this indication; Lilly / Scorpion (~$2.5B) for STX-478. Biotech M&A reached ~$149 billion across 49 deals in H1 2026 at a ~36% median premium — the strongest market since before Covid.

Change in accounting policies? No policy changes. One presentation change: the Q1 2026 10-Q reclassified approximately $2.5M of Q1 2025 expense between R&D and SG&A in connection with the FDA’s acceptance of the NDA for filing — with the note describing the direction incorrectly relative to its own restated figures. Immaterial to totals. Q2 2026 will carry a ~$9.6M loss on extinguishment of the retired term loan.

Recent changes — new markets, facilities, management?

• Capital structure, transformed. On 2026-06-08 Celcuity issued $575.0 million of 0.250% Convertible Senior Notes due 2032 (conversion price $124.53, a 40% premium; greenshoe fully exercised) and used part of the proceeds to pay off and terminate the Oxford/Innovatus term loan at ~$137.5 million, eliminating an ~11–12.5% secured facility and every financial covenant. Cash interest fell from ~$16M to ~$7M a year. No capped call was purchased on either convertible issue.
• Commercial organization built. Headcount went from 87 (Q4 2024) to 155 (Q1 2026); SG&A from $6.4M to $17.4M a quarter. All oncology sales specialists were hired and onboarded as of May 2026, averaging 24 years in pharmaceutical sales and 16 in oncology, targeting ~9,000 HCPs with focus on ~2,000 high-volume breast oncologists.
• Facilities. Corporate headquarters moved to 2800 Campus Drive, Minneapolis during 2026. All manufacturing remains outsourced.
• Development plan changed. VIKTORIA-2 was redesigned in May 2026 after a Type B FDA meeting: a second study covering endocrine-sensitive first-line patients was added, PIK3CA-status-based primary endpoints were removed in favor of intent-to-treat analyses (cutting Study 1 from 638 to 440 patients), and ribociclib was set as the control in both. A subcutaneous formulation program was disclosed with a first patent application filed.
• Board and plans. Charles (Chip) R. Romp joined the board in February 2026, expanding it to eight. Shareholders approved the 2026 Stock Incentive Plan (3,000,000 shares plus a 1%/year evergreen through 2036) and a restated ESPP on 2026-05-14.
• Litigation noise. Multiple plaintiff firms (Pomerantz, Kessler Topaz, Bragar Eagel, Kirby McInerney) issued “investigation” press releases between 2026-07-20 and 2026-07-30. INTERPRETATION: these follow essentially every large single-day decline and carry little information; noted for completeness, not weighted.

APPENDIX B — Source Appendix

Celcuity Inc. (NASDAQ: CELC) — 2026-08-01

All sources accessed 2026-08-01 unless otherwise noted. Primary sources are listed first within each category. Every non-obvious claim in the analysis above traces to an entry here.

1. Company primary sources — SEC filings (CIK 0001603454)

The trailing 60-month SEC corpus (352 filings, 2021-08-01 → 2026-08-01) was enumerated with scripts/edgar.sh since CELC 2021-08-01 and mirrored locally to output/CELC/sources/ via scripts/fetch_sources.sh. Form breakdown: 120× Form 4, 56× 8-K, 31× SC 13D/G, 22× Schedule 13G, 18× Form 144, 15× 10-Q, 12× D/A, 10× 424B5, 7× DEF 14A, 6× S-8, 6× DEFA14A, 5× 10-K, 4× S-3, 4× UPLOAD, 4× CORRESP, 4× EFFECT, 4× ARS, 3× FWP, 3× PRE 14A, 3× 8-K/A, 3× Form 3, 2× S-3/A, 2× 4/A, 2× 424B3, 2× 305B2, 1× S-3ASR, 1× 10-K/A.

Annual reports

Quarterly reports

Current reports (Form 8-K) — the material-event timeline

Proxy statements and registration statements

• DEF 14A, filed 2026-04-02 — Summary Compensation Table (CEO FY2025 total $16,373,868, 94.2% option grant-date fair value); annual-incentive language (“milestones, approved by our Compensation Committee… 40 to 60% of base salary”); director compensation ($140,000 each); beneficial ownership as of 2026-03-17; 2026 Stock Incentive Plan description (3,000,000 shares plus a 1%/year evergreen through 2036). output/CELC/sources/DEF_14A/2026-04-02_formdef14a.htm
• DEF 14A, filed 2025-04-01 — FY2024/FY2023 Summary Compensation Table.
• Form 424B5, filed 2026-06-05 — 2032 convertible notes: cover terms, conversion price $124.53, use of proceeds, capitalization and dilution. https://www.sec.gov/Archives/edgar/data/1603454/000149315226027395/form424b5.htm
• Form FWP, filed 2026-06-03/05 — pricing term sheet for the upsized $500M offering. https://www.sec.gov/Archives/edgar/data/0001603454/000149315226027183/formfwp.htm
• Shelf registrations: S-3 2021-11-17; S-3 2022-12-30; S-3 2023-11-14; S-3 2024-08-30 (S-3/A 2024-11-26, effective 2024-12-02); automatic S-3ASR 2026-01-09 (WKSI status), alongside the upsizing of the Jefferies ATM from $50M to $400M.
• Offering prospectuses (424B5): 2022-02-04, 2024-05-30, 2024-12-06, 2025-07-28 through 07-31 (four), 2026-06-03 and 06-05.

Insider filings

All 125 Forms 3/4/5 filed 2021-08-01 → 2026-08-01 were downloaded as raw XML and parsed to 301 transaction lines (output/CELC/2026-08-01/_scratch/insider_txns.tsv). Filings of particular note:

• 2026-07-16 — Baker Bros. Advisors LP: sale of 3,100,000 shares at $102.50 on 2026-07-14 (~$317.75M), in two blocks; no 10b5-1 flag. https://www.sec.gov/Archives/edgar/data/1603454/000119312526306299/xslF345X06/ownership.xml
• 2025-07-30 — David Dalvey: sale of 109,325 shares at $43.30–$46.05 on 2025-07-28, the day of the VIKTORIA-1 topline release; 10b5-1.
• 2025-09-10 — Baker Bros. Advisors LP: open-market purchases of 170,100 shares at $56.22–$58.02 (~$9.6M).
• 2025-08-20 — NEO option grants at a $51.57 strike (Sullivan 400,000; Laing 100,000; Hahne 80,000).
• 2022-05-19 — Form 4s documenting the one-time option repricing to $5.50 for every officer and director.

2. FDA, label and regulatory sources

3. Clinical publications and conference presentations

4. Epidemiology and market data

5. Pricing, reimbursement and policy

6. Competitor and industry financial data

7. News and trade press

8. Quantitative data sources and tooling

9. Methodological and comparative references

The risk-adjusted NPV in Section 10 follows a conventional single-asset biotechnology framework: a 13% discount rate with a mid-year convention, an operating margin at scale adjusted for the Pfizer royalty, a blended cash-tax rate reflecting the net-operating-loss shield, the present value of remaining development spend, a probability-of-success grid, and an algebraic back-solve of the probability-weighted peak sales that the current enterprise value implies. Probability-of-success anchors are drawn from the BIO / Informa / QLS study of 12,728 phase transitions (oncology likelihood of approval from Phase 1 of 5.3%; solid tumors 4.6%; biomarker-preselected 15.9%; likelihood of approval from Phase 3 in solid tumors 39.8%).

The cross-sectional multiple bands come from published sources:

• Centerview Partners’ fairness opinion in the GSK / Nuvalent transaction (Schedule 14D-9, June 2026) — a “selected public companies” analysis at EV / 2031E revenue of 2.0–4.0× whose peer set explicitly includes Celcuity, alongside Cogent, Crinetics, Cytokinetics, Insmed and Revolution Medicines; and a precedent-transactions analysis at 3.0–6.0× five-year-forward revenue. https://www.sec.gov/Archives/edgar/data/0001861560/000119312526280275/d26485dsc14d9.htm
• Public-company enterprise values in the comparables tables are computed from each company’s latest reported cash and shares outstanding at prices as of late July 2026; peak-sales figures are the published street or company estimates cited alongside them.
• Transaction values and premia are taken from the acquirers’ own press releases and merger filings, and from the aggregate H1 2026 M&A data cited in Section 6.

Third-party sell-side and academic industry primers were consulted as framework only and are all of 2004–2011 vintage, predating the CDK4/6 and PI3K-inhibitor era: Morgan Stanley, The US Healthcare Formula (2011); Deutsche Bank, Pharmaceuticals for Beginners, 3rd edition (2005) and 2010 edition; and Danzon, Epstein & Nicholson, Mergers and Acquisitions in the Pharmaceutical and Biotech Industries, NBER Working Paper 10536 (2004). These are third-party research, not the conclusions of this article, and none is current data.

10. Sources sought and not obtained

Recorded for completeness, because their absence bounds several conclusions in the memo:

• REVTORPYK’s wholesale acquisition cost. Not disclosed by the company as of 2026-08-01. Every revenue figure in Section 3.1 and Section 10 scales approximately linearly with it. All pricing figures used are third-party estimates ($24,000–28,000 per 28-day cycle), explicitly labeled as such.
• The split of the $335.0M Pfizer milestone package between development and commercial tranches, and the sales thresholds that trigger the commercial portion. Aggregated in the 10-K without detail. An approval milestone is very likely payable in Q3 2026; the amount is unknown.
• The Pfizer royalty tiers. Disclosed only as “percentages ranging from the low to mid-teens.”
• Sales force headcount. Never disclosed — only the ~9,000 HCP and ~2,000 high-volume-provider target universe.
• Any committed drug-substance or drug-product supply agreement. Only $5.5M of non-cancelable clinical purchase commitments appears at 12/31/25, with no take-or-pay disclosed, for a product whose launch was delayed on drug supply.
• Q2 2026 Form 10-Q. Not yet filed as of the report date (expected mid-August 2026). The pro-forma cash position of ~$746M and the pro-forma book equity of approximately zero-to-negative are therefore estimates built from Q1 2026 actuals plus the disclosed financing terms.
• Final VIKTORIA-1 overall survival. Immature at approval (~25% events) and never formally tested.
• Whether a Patent Term Extension will be granted on the composition-of-matter patent, and to which patent the company will elect to apply it.
• Whether a 10b5-1 plan underlay the Baker Bros. sale of 3,100,000 shares on 2026-07-14. The Form 4 carries no such flag.