ImmunityBio, Inc.$7.32 at publication $50m a quarter. 366 patients. $3 or $46.

01Summary

ANKTIVA sells $50.7m a quarter¹ after eight consecutive quarters of growth, at roughly 99% gross margin, with a permanent J-code, NCCN listings in two settings, and a narrowing operating loss. The product works. The investment question is how fast it can scale, and against what.

Four conclusions from this work differ from both the standard bull and bear framings of the name.

First, what limits this business is evidence and timing, not BCG supply. ANKTIVA is approved in the US only in combination with BCG, so supply constrains the business the drug has today. "BCG-unresponsive" is a status that requires documented adequate prior BCG, so the shortage constricts the eligible pool for every second-line agent. But it does not gate the growth thesis. ANKTIVA in the BCG-naive setting is given on the same BCG schedule the patient already receives, so it adds no BCG demand of its own. Urologists also prioritize high-risk naive patients during shortages. What actually decides 2028 is whether QUILT-2.005's full readout holds up and how fast the FDA moves.

Second, the competitor that matters is not the one with the head start. AstraZeneca won the first approval in the BCG-naive setting in May 2026, on randomized Phase 3 data, in the same first-line population ANKTIVA is filing for. But ten weeks on, NCCN has declined to list IMFINZI plus BCG at all, citing the absence of an overall-survival benefit⁴⁹; AstraZeneca withdrew its European filing in the same setting⁵⁰, and its own Q2 call gave the launch a single passing mention. Infusion in an oncology suite against instillation by the urologist at the same visit as BCG is one of four headwinds it carries, and not the largest. The pressure is coming from the other direction. INLEXZO is ramping faster than ANKTIVA did from an equivalent launch point, reporting sales above $30m in its second full quarter and nearly one in three eligible patients starting on the regimen a quarter later¹⁴², helped by needing no BCG and by an installed sales infrastructure. It sells into the same second-line pool. That is the competitive risk worth watching, and Section 03 sets out what would confirm it — two consecutive quarters in which ANKTIVA’s sequential adds fall below the modeled band while INLEXZO’s insertions keep growing.

Third, the Gulf approvals matter for more than lung cancer. ANKTIVA's subcutaneous approvals in Saudi Arabia and the UAE put an injectable, systemically-administered vial on a pharmacy shelf in jurisdictions where off-label prescribing is lawful — which makes every future NSCLC approval a lawful supply channel for immune reconstitution in a far larger population. That is examined in Section 12.

Fourth, the company is closer to self-funding than its going-concern⁴¹ language implies. Operating cash flow turns positive in Q1 2028 with a cash trough near $113m — tight but not forced. The larger issue is the $480m convertible maturing December 31, 2027³⁶, one quarter ahead of that trough, examined in Section 16.

02Unit economics and how the drug gets paid for

ANKTIVA (nogapendekin alfa inbakicept) is an IL-15 superagonist approved April 2024³ for BCG-unresponsive NMIBC with carcinoma in situ, administered intravesically in combination with BCG. That combination requirement is the most consequential fact about the franchise and recurs throughout this report.

Bladder cancer is a disease of the old: median age at diagnosis is 73, and roughly three-quarters of new cases are diagnosed after 65.¹⁰⁹ The payer mix follows from the age. Across 163,949 high-risk patients in the National Cancer Database, 62% were covered by Medicare and 32% commercially, with Medicaid at 3% and the uninsured at 2%.¹¹⁰ That is why the analysis below turns on Medicare mechanics rather than commercial contracting — for the bladder cancer business in the US, Medicare largely is the market.

The pattern: percentage growth decelerates while the absolute dollar adds hold steady. The last four quarters added $5.4m, $6.5m, $5.9m and $6.5m, against $9.9m in the quarter before them. Constant adds imply a constant rate of new patient starts, since each patient contributes across several quarters of dosing, so a steady intake produces a straight line rather than a curve. Why the intake is steady rather than accelerating is the question Section 05 takes up.

The burn tells a cleaner story. It improved $13.2m year-on-year, from $79.7m to $66.5m, on a quarter-standalone basis derived from the half-year statements. Revenue over the same period rose $24.3m, which is nearly twice the improvement in burn — so cash operating costs grew by roughly $11m even as the company moved closer to breakeven. The improvement came from revenue rather than from cutting costs.

Unit economics#

WAC is $35,800 per 400mcg dose¹⁵, unchanged since launch — FY2025²'s growth was volume, not price. At the reported ~17% gross-to-net, net realization is approximately $29,714 per dose. The US label permits up to 30 doses over 37 months (roughly $891k net lifetime), but QUILT-3.032 persistence implies an expected 12.8 doses, or about $380k net per patient. Q2 2026 revenue therefore represents roughly 1,700 doses, or in quasi-steady state some 130–200 new patient starts per quarter — 550–800 annually. The BCG-unresponsive pool is not directly counted, and the derived range runs from roughly 7,000 to 13,000 depending on how strictly the label definition is applied.¹³⁶ Penetration is therefore somewhere between 4% and 11%, and most of the approved population is untreated on any reading.

The gross margin, and why it persists#

ANKTIVA reports cost of sales of $0.298m against $50.7m of revenue in Q2 2026 — a gross margin above 99%. No commercial biologic sustains that on normal economics, and the reason matters for the cost line in this model.

Manufacturing costs incurred before FDA approval were expensed to research and development rather than capitalized, because inventory was not yet probable of future economic benefit. ImmunityBio completed and released drug substance sufficient for 170,000 doses in May 2024.⁴⁸ At the current consumption rate of roughly 1,700 doses a quarter, that is on the order of twenty-five years of bulk supply carried at essentially zero cost. Reported cost of sales therefore reflects fill-finish and packaging, not drug substance.

The practical consequence is that a model normalizing cost of goods to the 10–15% typical of large biologics would be wrong. ANKTIVA is dosed at 400 micrograms against a $35,800 list price; the protein mass is trivial and the substance behind it is already expensed. Product cost of sales should stay in the low single digits as a percentage of revenue for years, and this model holds it there.

Two details bound how long that advantage runs. Finished drug product carries a three-year shelf life, extended from two in September 2024, while bulk substance holds a multi-year life before it is filled.⁹² The company reserves $0.2m against excess and obsolescence and has recorded no inventory write-downs since approval. At roughly 1,700 doses a quarter, most of the 170,000-dose stockpile will never be filled and sold before it lapses. That costs nothing in reported earnings, since it was expensed to zero before approval. It does mean the margin advantage ends when current supply is consumed rather than when the stockpile is exhausted.

The offsetting consideration is that the fixed cost of the manufacturing base is not in cost of sales at all. The Dunkirk fill-finish facility spans 400,000 square feet with capacity for a million vials annually⁴⁸ against current demand of roughly 7,000 doses a year — utilization below one percent. Under ASC 330-10-30-7 abnormal idle capacity must be expensed currently rather than capitalized, so that burden sits in research and development and selling, general and administrative expenses. Bring-up and validation costs will roll off as the facility reaches commercial status, and idle-capacity charges partly replace them. The accounting can say what the plant costs; what it is for is a different question, taken up in Section 10. The model treats this as approximately neutral and does not credit an opex reduction. Capital spending offers no relief either. It fell from $30.6m in 2023 to $3.8m in 2025 as the original build finished, then turned back up: $5.3m in the first half of 2026 against $2.2m a year earlier. A lease amendment effective January 2026 commits the company to at least $40.0m of capital spending by the end of 2028, and management has referenced roughly $50m of fill-finish equipment for the site.¹³⁴

How the practice gets paid, and why the price is hard to move#

ANKTIVA is a buy-and-bill drug. That single fact shapes the competitive analysis in this report more than any clinical result does, because it determines who can administer the drug and what they earn for doing so. J9028 is the permanent billing code assigned to ANKTIVA, and having one is what lets a practice be paid for the drug rather than only for the procedure.

Two consequences follow. The practice buys the vial and bills the payer, so its margin is a percentage of the price — roughly $2,150 a dose, which a list-price cut would reduce proportionally.¹³⁸ And that drug margin runs between eight and twenty times what Medicare pays for the instillation itself.¹⁰⁶ A urologist who administers ANKTIVA keeps it; one who refers the patient out for an infusion earns the referral and nothing else.⁵⁸ BCG inverts the arithmetic exactly: at roughly $141 a vial the add-on is about eight dollars, and the procedure fee is the entire economics. That asymmetry is what Section 04 rests on.

The same arithmetic erodes the price slowly without anyone deciding anything. Average sales price is a trailing average of the manufacturer’s own net realizations, so every discount, rebate, chargeback and 340B unit feeds back into it two quarters later and lowers the reimbursement ceiling for everyone. A deliberate cut accelerates a ratchet that only turns one way.

Two further constraints make a cut costly to undo. A list-price reduction propagates into average manufacturer price, Medicaid best price and the 340B ceiling,⁵⁹ resetting floors that cannot easily be raised again, which is why manufacturers reach for copay support and rebates rather than list price. None of that argues against cutting. It makes the cut one-way, so the test is whether the volume it buys is worth a ceiling that cannot be raised afterwards — a question about the size of the response, not about the direction. And the practice buys at wholesale acquisition cost and is reimbursed at ASP plus six percent, so the spread it earns is set two quarters in arrears against a price it paid today.

A 99% gross margin, twenty-five years of drug substance already expensed and a plant running below one percent of capacity together suggest an obvious move: cut the price, expand volume, fill the factory.

The clearest thing a cut would do is lower what the patient owes. At 20% coinsurance the first dose costs roughly $7,160 today, and this population is largely elderly and on a fixed income.

What a cut would not do is change who gets treated. When Medicare cut reimbursement for androgen-deprivation therapy in 2003, urologists reduced the use for which there was no evidence of benefit by more than thirty percent and did not reduce appropriate use at all.¹⁰⁸ Margin moves which agent a practice reaches for among comparable options, and whether it administers rather than refers. It does not appear to move clinical judgement about who needs treatment,¹⁰⁷ and that is the shape of the effect assumed here.

The case where a cut reduces adoption is narrower still. It requires a cut deep enough to push ANKTIVA’s retained margin below that of a bladder-sparing competitor the same practice could administer instead. The sharper risk sits upstream of prescribing: a practice that cannot cover its acquisition cost stops buying the drug, or stops taking Medicare patients altogether, and the patient never reaches the conversation.

Coverage friction has largely resolved. The code is permanent, pass-through runs to September 2027, NCCN carries Category 2A listings in two settings, and the company reports commercial and government programs covering more than 240 million lives.¹¹¹ That figure rose from 100 million three months after approval and has held since January 2025.

Cost-sharing has not resolved. Medicare Part B coinsurance is 20%, roughly $7,160 on the first dose and above $100,000 across a full course. The manufacturer copay card cannot be used by Medicare, Medicaid or VA patients under federal anti-kickback rules. The median patient therefore depends on Medigap, a Medicare Advantage out-of-pocket cap, or the company’s own assistance program.¹¹² The residual exposure falls on the roughly one in eight traditional-Medicare beneficiaries with no supplemental cover at all. That is a barrier for an individual patient and a small one for the forecast.

Price moves volume outside US buy-and-bill. Gulf sales are tender and cash-pay without ASP mechanics, and a hypothetical inpatient sepsis indication would be bundled into a DRG where a $35,800 dose is unviable. Both are separate price tiers, not arguments for a single lower global price — and publishing a low international price carries its own hazard, discussed under risks. The model therefore holds US price constant and treats capacity as optionality on indication breadth rather than on price.

03Competitive landscape

Cross-trial comparison in this field is unreliable. Every pivotal study is single-arm with different eligibility, biopsy protocols, response definitions, and follow-up. The tables below present the data with those differences visible rather than ranked, and flag BCG dependency — the axis on which ANKTIVA is structurally disadvantaged. Both ImmunityBio and Johnson & Johnson have published sponsor-funded indirect comparisons favoring their own product; these are marketing-adjacent rather than evidence.

Four things are worth carrying out of the tables.

First, two competitors need no BCG and ANKTIVA does. TAR-200 is a gemcitabine-releasing device placed in the bladder. Cretostimogene is an oncolytic virus. Neither depends on a supply that has been short for years, and ANKTIVA cannot be given without it. That is the structural disadvantage, and the reason Section 05 spends so long on supply.

Second, IMFINZI is the only competitor approved in the US in the BCG-naive setting. That setting is the largest single line in this model, sized in Section 04. IMFINZI’s weaknesses there matter more than its headline numbers do.

Third, the response rates are not comparable across rows. They sit in one table for convenience. The highest number does not win.

The fourth is not in the tables at all, because tables hold profiles rather than trajectories. INLEXZO is ramping faster than ANKTIVA did. J&J reported sales slightly above $30m in Q1 2026, its second full quarter. After a permanent J-code took effect on April 1, its Q2 call reported nearly one in three eligible patients starting on the regimen, with new patient insertions up roughly 75% over the prior quarter.¹⁴² Some of that pace is machinery rather than product. J&J ran SunRISe-1, the registrational TAR-200 trial in BCG-unresponsive disease, at 142 sites across fourteen countries, and opened SunRISe-2, its muscle-invasive companion study, at 272 locations, against 32 US sites for QUILT-3.032, so at approval hands-on familiarity with the device already existed at several times the number of institutions that had ever touched ANKTIVA. Its representatives had also been calling on US urologists since ERLEADA’s US approval in 2018.¹⁷⁰ The pace still stands despite the worse tolerability the table shows, which says prescribers are not weighting grade ≥3 rates the way the profile comparison implies. The second-line share assumptions already absorb competitive decay; what would break them is two consecutive quarters in which ANKTIVA’s sequential revenue adds fall below the modeled band while INLEXZO’s insertions keep growing. J&J does not disclose INLEXZO revenue separately, so the insertions commentary on its earnings calls is the number to watch.

Two things follow. Including the BCG its US label requires, ANKTIVA costs three to six times what the two cheaper alternatives cost, and the patient’s twenty percent share is roughly $91,000 against $26,000 for ADSTILADRIN and $15,000 for KEYTRUDA. INLEXZO appears to cost more, though its figure is estimated: the pivotal paper publishes no median, so nine systems is the midpoint of what the schedule and the disposition counts allow. The ranges overlap heavily in any case. ANKTIVA runs from $76,000 to $1.14m depending on how long a patient stays on it. That is a wider bracket than INLEXZO’s. The comparison holds at the median rather than at either end. And it takes fifteen visits in the first year where ADSTILADRIN takes four. In a population with a median age of 73, visit burden bears on adherence as directly as price does. The obvious response to a gap that size is to cut the price. Section 02 sets out why the arithmetic there is not the ordinary one. Average sales price ratchets in one direction, and under buy-and-bill a lower price also shrinks what the practice retains for administering it.

The comparison has limits. INLEXZO’s pivotal publication does not report how many systems patients received, so its course is a bracket rather than a figure: five systems for a patient stopping at the first assessment, fourteen at the label maximum. That range spans ANKTIVA’s course rather than sitting to one side of it, which is the honest answer until the dose count is published. The same platform is being extended beyond this tier: Erda-iDRS, the erdafitinib-releasing system formerly TAR-210, reported final phase 1 results at the EAU congress in March 2026, with an 89% complete response rate and a median response duration of 18 months in intermediate-risk disease carrying fibroblast growth factor receptor alterations, and 83% recurrence-free survival at twelve months in the high-risk, BCG-experienced, papillary-only cohort.¹⁶¹ The intermediate-risk tier sits outside the pools this model counts, but that second cohort is the population behind ANKTIVA’s papillary supplemental application, approached with a device rather than a vial and in patients selected by mutation rather than by risk category. Two agents are priced at average sales price and two at list, because INLEXZO has held a permanent billing code only since April 2026.¹⁴⁰ Those bases are not interchangeable, and the patient-share column inherits whichever its row uses.

04The BCG-naive share question

The single most consequential assumption in this model is how much of the BCG-naive population ANKTIVA captures. More of the 2032 valuation turns on that one number than on any clinical result.

The pool moves in two directions, and the net is mildly favorable over the forecast. Age-adjusted incidence has fallen about 1.5% a year since 2005 as smoking rates decline. The absolute number of new cases is projected to grow 1.5% to 1.8% a year through 2040, because the population is ageing into the disease faster than the rate is falling.¹¹³ A model running to 2032 should assume a slowly growing eligible population, not a shrinking one.

It is largely a question of workflow rather than data. Community urologists manage these patients, and what a practice can administer in its own rooms decides more than what a trial reported. IMFINZI has three further obstacles beyond that one, and the interim durability figures need discounting before they can carry a forecast. The base case is anchored on ANKTIVA’s own second-line launch. The bull case requires something different in kind — not a faster launch, but a change in what the guidelines recommend.

Where these patients are treated#

Site of care decides more than the data. BCG-naive high-risk NMIBC is managed almost entirely by community urologists. IMFINZI requires intravenous infusion every four weeks for up to thirteen cycles — which roughly two-thirds of US urology practices cannot deliver in-house⁵², forcing a referral to medical oncology and the loss of both the patient relationship and the revenue. ANKTIVA is instilled in the same visit as the BCG the patient is already receiving, by the same staff, and bills under J9028¹⁰ with buy-and-bill margin retained, plus CMS pass-through status through September 2027. The margin is not frictionless. Clearing-house claims data for January through August 2026 show 31.55% of J9028 claims denied and an average of 25.8 days to payment, so the practice fronts a five-figure drug cost into a roughly one-in-three chance of a denial cycle.¹⁷⁵

The precedent is unusually clean. Pembrolizumab was approved by the FDA in BCG-unresponsive NMIBC in January 2020 with no locally administered on-label competitor at the time. A claims analysis of Optum Clinformatics data covering January 2020 to December 2021 identified just 126 patients²⁷ initiating treatment over two years, median age 80, median time on therapy 6.8 months. An intravenous systemic agent essentially failed to penetrate a urology-managed setting despite having the field to itself.

What else stands in IMFINZI’s way#

Three obstacles matter more than where the drug is given. First, and most consequentially, NCCN has declined to list the regimen. The Bladder Cancer panel reviewed both POTOMAC and the sasanlimab CREST data and chose not to include either for BCG-naive high-risk disease, noting that although both improved disease-free survival, no differences in overall survival were observed.⁴⁹ POTOMAC's survival data remains immature — hazard ratio 0.80, confidence interval 0.53 to 1.20, at roughly 14% maturity. Without an NCCN listing, payer coverage and prescriber confidence both lag.

Second, toxicity reads differently in curative-intent disease. Clinicians surveyed in late 2025 reported minimal use of systemic immunotherapy in NMIBC and described checkpoint-inhibitor toxicities as feeling disproportionate for a non-muscle-invasive setting.⁵¹ POTOMAC's 21% grade 3/4 rate sits against roughly 4% for BCG alone.

Third — and this is the one most easily missed — IMFINZI does not escape the BCG shortage, because it is given with BCG. The same clinicians report that the requirement for concurrent BCG has itself limited uptake, since scarce supply is rationed to naive patients rather than to combination regimens.⁵¹ Every constraint the shortage imposes on ANKTIVA in this setting, it imposes on IMFINZI too.

Two further signals from the first ten weeks. AstraZeneca withdrew its European filing for the BCG-naive indication⁵⁰, and its Q2 2026 earnings call spared the launch one sentence, with no patient-start or channel metrics disclosed. Read both carefully. It is far too early for uptake data, and silence at this stage is ambiguous. What can be said is that the observable record so far shows a launch tracking the structural obstacles rather than outrunning them.

Weighing the evidence#

Safety carries more weight here than it would elsewhere. This is a curative-intent, bladder-preservation population of otherwise-well patients. A 21% grade 3/4 rate and 27% immune-mediated AEs — thyroid dysfunction, hepatic events, dermatitis, with 45% of affected patients requiring high-dose corticosteroids — reads very differently in first-line organ preservation than in metastatic disease. The evidence nonetheless indicates safety is secondary to workflow.

How the penetration assumption is built. It is constructed as two factors rather than one blended rate: the fraction of eligible patients who receive any branded add-on, multiplied by ANKTIVA's share of that add-on pool. ANKTIVA likely wins a high share of the second factor on workflow and economics. The first factor stays modest, because most BCG-naive patients will remain on BCG alone — it is cheap, familiar, effective for the majority, and payers facing a $750k–$1.1m add-on across roughly 31,000 first-line patients a year will impose prior authorization and probable step-through.

Building the base case#

The base case therefore carries roughly 4.5% of the eligible pool in launch year 2028. That figure is anchored on ANKTIVA's own second-line launch rather than on a generic band. Two named comparators bound it, both measured from the date a working reimbursement code existed rather than from approval. That distinction matters here: ANKTIVA's permanent J-code took effect on January 1, 2025, roughly nine months after approval, and the pre-code quarters are not representative of anything.

The J-code effect is the reason the pre-code period must be excluded. The first quarter carrying J9028 delivered $16.5m against $7.2m the quarter before — a 129% increase, with unit volume up 150% and a single quarter's units exceeding all of fiscal 2024.⁴³ Averaging that suppressed period into a "first year post-approval" figure understates what the current configuration can do. A BCG-naive launch would inherit the working code, the salesforce and the prescriber relationships, and would skip the friction ANKTIVA itself paid in 2024 — though an organization sized for the second-line niche would still have to broaden across a pool four times larger, a qualification taken up in Section 05.

Against that reference class, 4.5% for a naive launch is central-to-conservative rather than aggressive. It assumes a first-line add-on roughly matches what the same company achieved in the harder second-line setting. The workflow fit is better, because the drug is given alongside BCG the patient already receives. Against that sits payer resistance across a pool four times larger: roughly 31,000 newly-treated high-risk patients a year, against a derived 7,000 to 13,000 becoming BCG-unresponsive.¹³⁶⁵⁷ Years two and three ramp toward 7–12% and 12–20%, gated on a clean full readout, guideline inclusion and coverage.

What the bull case requires#

05What BCG supply does and does not constrain

Merck is the sole US supplier of TICE BCG, allocating a global capacity of 600,000–870,000 vials¹⁴ annually on the basis of historical purchasing patterns — a formula that structurally excludes small and newer practices regardless of patient need. A company-cited survey found 57% of US urologists unable to treat patients for want of it. The question is what that actually constrains.

It constrains the second-line funnel. "BCG-unresponsive" is a definitional status²⁸ requiring documented adequate prior BCG — generally five of six induction instillations plus two of three maintenance. Patients who never receive enough BCG never enter the category, so the shortage narrows the second-line funnel for every agent in it, including the BCG-free ones. This is the correct reading of ANKTIVA's linear ~$6m quarterly adds: it is partly a market-wide phenomenon, not solely company-specific execution.

It does not constrain the BCG-naive opportunity. This point carries the entire 2028 forecast. In the naive setting ANKTIVA is administered alongside the same BCG schedule the patient receives as standard of care — six weekly induction instillations, then three-week maintenance courses, identical to the control arm of QUILT-2.005. Adding ANKTIVA adds no incremental BCG. Moreover, AUA/SUO shortage guidance directs that scarce BCG be prioritized for exactly these high-risk naive patients, and dose-splitting lets one vial treat three. Even at an aggressive 2,500 naive patients a year, ANKTIVA-associated BCG use is a low single-digit percentage of current global supply — and that BCG is consumed with or without ANKTIVA.

One residual friction survives in the second-line setting: because ANKTIVA there does consume BCG for a patient who has already failed it, a rationing urologist has an argument for withholding scarce supply from that combination. That pressure does not apply to first-line use.

Why TICE relief is modeled as an unlock rather than a wash#

ImmunityBio has a second source of BCG. Recombinant BCG, or rBCG, is a genetically modified version of the same bacterium, manufactured by the Serum Institute of India. It is not approved in the United States. The FDA authorized it in February 2025 under an expanded-access program, a mechanism that lets patients receive an investigational drug outside a clinical trial when no adequate alternative exists.⁹⁰ For a company whose only approved product must be given alongside a rationed input, a second supply looks like the obvious answer to the constraint this section describes.

There is a third source further out. Tokyo-172 is a different BCG strain, made by the Japan BCG Laboratory, to which ImmunityBio secured exclusive US rights in May 2026. An NCI-sponsored trial of roughly a thousand patients found it non-inferior to TICE on high-grade recurrence-free survival. Unlike rBCG it would be a licensed product rather than an expanded-access one, so it carries a billing code and a margin. ImmunityBio would have to file for it, and approval before 2029 looks unlikely.

Two supply sources plus Merck's Durham expansion look like enough to settle the BCG problem. The billing mechanics say otherwise, and they govern the largest single upward revision in this model.

A urology practice administering TICE bills two components: the instillation procedure under CPT 51720, and the drug itself under J9030, the code for BCG, at a national Medicare rate of $2.821 per milligram, or roughly $141 for a 50mg vial.⁴⁶ Under an expanded-access program rBCG is investigational, carries no J-code, and is supplied free or at cost recovery — so the drug line disappears. The procedure fee survives: reporting of 51720 at full value is appropriate regardless of dose.⁴⁷

There is a second constraint, and it is regulatory rather than economic. The rule against promoting an investigational drug is explicit, and charging under an expanded-access program leaves it in force.⁹¹ ImmunityBio may tell a practice the program exists and supply the protocol; marketing rBCG as it markets ANKTIVA is closed to it. So the gap between roughly two hundred practices registering and around fifty-eight becoming active is a regulatory ceiling rather than a marketing failure, and only a license application lifts it. None has been filed.

So the direct economic loss is small — roughly $141 of gross billing per instillation, and at ASP plus 6% only single-digit dollars of margin. That is precisely what makes the friction decisive. To preserve a $141 line item a practice must accept institutional-review-board oversight, informed consent, adverse-event reporting to the sponsor and FDA, live-biologic storage and handling, and staff training. The cost-benefit is indefensible for any practice with an alternative.

The conversion data confirms it. ImmunityBio reported nearly 200 urology practices registering for the rBCG program in May 2025.⁴² By early 2026 the program had roughly 58 active sites. Three-quarters of registering practices did not proceed.

The US core line accordingly decays through Q2 2027 as the current niche saturates, then re-accelerates from Q3 2027 as relief arrives — sequential adds returning to roughly $7m rather than fading toward $2m. FY2028 US core moves from $364m to $413m, and the FY2028 total from $841m to $890m.

The risk to this is timing rather than direction. Merck's guidance is that supply increases "gradually over time following local market review and approvals," and the gating sequence — prior-approval supplement, pre-approval inspection, a manufacturing cycle exceeding three months, then lot release — could push relief into 2028. A one-year delay removes most of the FY2028 benefit without changing the terminal picture.

Commercial capacity is not what binds#

A young commercial organization facing a competitor with an established urology field force looks like it should accelerate as its own team matures. The evidence shows that catch-up has already happened.

ANKTIVA's inflection came with reimbursement, not headcount. The step change set out in Section 04 — revenue more than doubling in the first quarter the permanent J-code was effective — arrived without any corresponding change in the size of the field organization. By late 2025 the company described adoption spanning leading research centers and community urology clinics including rural areas.⁴⁴

The J-code has now been effective for eighteen months. If commercial immaturity were still suppressing demand, sequential adds would be rising. They have held between $4.9m and $6.5m for six consecutive quarters.

The spending pattern makes the same point from the other direction. Selling, general and administrative expense fell in 2025, to $150m from $168.8m, as the company insourced commercial functions. It then stepped up sharply: $45.8m in the first quarter of 2026 and $51.8m in the second, against a 2025 quarterly range of $32.7m to $42.3m, with management attributing the increase to personnel and commercial costs.⁷¹ Roughly a third more commercial spend produced no change in the sequential adds. And the buildout is still running: as of August 2026 the company is recruiting uro-oncology business managers for individual territories — San Antonio, St. Louis, Indianapolis — alongside an area sales director, a health-systems director for Eastern-US hospital networks, and field reimbursement managers.¹⁷¹ The calibration for what coverage takes is public: UroGen covered the US urology call point for Jelmyto at launch with a field force of about 48 representatives.¹⁷² An organization still extending its reach while the adds hold flat strengthens the conclusion rather than weakening it. More field capacity is not what moves the number, the flat line is set by something other than selling effort, and no catch-up acceleration is modeled.

Judged against the reference class rather than against intuition, the cadence is mid-pack. ANKTIVA’s 2025 ran ahead of ADSTILADRIN’s first full US year — roughly $77m at then-current exchange rates, capped by Ferring’s own manufacturing ramp rather than by demand — while INLEXZO is faster than both, for the reasons Section 03 sets out.¹⁷³ Flat sequential adds are what a supply-gated intravesical launch looks like from the middle of its class.

One caveat sits under all of it. Revenue is recognized on shipment to specialty distributors rather than on administration to a patient, and four customers account for the overwhelming majority of it — 41%, 20%, 18% and 17% in the most recent quarter that discloses the split.⁸⁴ The company reports unit volume growth alongside revenue, which is the better proxy, but it does not disclose channel inventory or sell-through. Eight quarters of roughly $6m sequential adds are therefore consistent with steady end-demand and also, in principle, with a distributor building stock. Nothing in the disclosure separates the two. The unit-volume series and the reorder pattern argue for demand from patients rather than shelves, and a channel build large enough to explain two years of adds would be unusual, but the accounts cannot settle it.

Two caveats belong with that. ImmunityBio discloses no field headcount, territory count or ordering-account count, so the conclusion rests on what the spending and the revenue did rather than on a disclosed organization chart. It is also a conclusion about the second-line niche, where the call point is a few hundred high-volume practices. No public dataset can check the account picture yet: the CMS provider-level file that would count the practices billing J9028 runs only through calendar 2023, before the code existed, so the first year it can count arrives around 2027. Until then the quarterly Part B drug-spending series and the state Medicaid utilization files, on roughly six-month and one-quarter lags, are the only outside reads.¹⁷⁴ Section 04 assumes a BCG-naive launch inherits that salesforce. Against a pool four times larger and less concentrated, inheriting the organization is not the same as covering the population, and the first year of that launch would carry a coverage lag this one no longer does.

06Revenue model

We build from Q2 2026 actuals using a cohort and penetration approach rather than extrapolated dollar adds: $29,714 net per dose, 12.8 expected doses per patient, and a new-start ramp modulated by BCG availability. The systemic line uses a different figure. ResQ201A carries no treatment-duration cap — maintenance runs to the end of study, where the pembrolizumab and nivolumab lung labels stop at twenty-four months. At a median overall survival near fourteen months in this population, consumption is bounded by survival rather than by protocol: six to twelve doses a patient, not the thirty-odd an open-ended schedule implies.⁹⁵ Papillary is treated as acceleration of an existing trend rather than a discrete new market, because NCCN Category 2A already permits reimbursed use. BCG-naive enters late, small, and supply-gated. NSCLC contributes nothing before 2029 and is excluded from the base case entirely.

Consensus at roughly $1.1bn for FY28 now sits level with our bull case rather than above it. The Street reaches that figure principally through the BCG-naive opportunity, modeled with an approval arriving on time and a launch curve that assumes the full readout confirms the interim. We reach a similar number by a different route — a slower bladder ramp offset by the lymphopenia and Gulf channels. The agreement on the total conceals a disagreement about composition, and the two paths carry different risks.

What would break this model, in order of how much it would move: a QUILT-2.005 full readout that loses significance, which removes the BCG-naive line entirely; net price falling below roughly $27,000 a dose through most-favored-nation reference pricing, which cuts every year proportionally; and BCG supply relief arriving materially later than mid-2027, which pushes the naive ramp right without changing its shape. The first is binary and the other two are continuous, which is why the spread in Section 16 is wide at the tails rather than around the middle.

07Mechanism, and the limits of the durability data

ANKTIVA is an IL-15 mutant (IL-15N72D) bound to a dimeric IL-15 receptor-α sushi domain fused to IgG1 Fc. That scaffold holds the cytokine the way the body holds it, with one cell presenting IL-15 to another rather than releasing it freely. The arrangement is called trans-presentation, and copying it is what drives proliferation of NK cells and CD8+ memory T cells through the IL-2/IL-15 receptor βγc complex. Its differentiator versus IL-2 is that IL-15 does not support regulatory T cells — it expands cytotoxic effectors without expanding immunosuppression.

The hypothesis that this "trains" durable antitumor immunity is plausible and is consistent with the long response durations observed. But two qualifications matter for valuation.

There is no published human immune-monitoring data from the QUILT bladder trials demonstrating memory T-cell generation, epitope spreading, or trained-immunity markers. The mechanistic backbone is a rat NMIBC model and systemic-setting data in other diseases. Durable clinical responses are consistent with immune training but do not prove it.

Intravesical administration produces no measurable systemic exposure. The FDA label states systemic concentrations were below the limit of quantitation³ (<100 pg/mL) in all patients. A bladder-cancer patient receiving ANKTIVA is therefore not receiving a body-wide immune benefit, and prescribing on that rationale would have no scientific basis. The systemic thesis lives entirely in the subcutaneous program — a different route, a different set of trials, and a different risk profile.

Finally, the framing that chemotherapy "wears out" while immunotherapy trains is too clean. Gemcitabine is a recognized inducer of immunogenic cell death, depleting myeloid-derived suppressor cells and enhancing antigen cross-presentation — which is why J&J is testing TAR-200 with a checkpoint inhibitor. Neither agent has proven durable immune memory in human bladder tissue.

The competitive frame is wrong outside the bladder#

The competitive sections above treat checkpoint inhibitors as rivals for the same patient. Inside the bladder that is right. Outside the bladder the mechanism runs the other way.

Checkpoint inhibitors do not kill tumors. They release a brake on T cells that are already present. They fail when there are too few functional effector cells for that release to matter — an immunologically cold tumor, an exhausted repertoire, or a lymphopenic patient. An IL-15 superagonist does precisely the thing that deficit calls for. On that reading IMFINZI, pembrolizumab, sasanlimab and cetrelimab become the agents whose failures ANKTIVA is designed to rescue.

The preclinical support is more specific than a general appeal to synergy. Trial registries and papers written before approval call the drug N-803, and this report uses that name where the sources do. In anti-PD-L1-refractory models, N-803 plus anti-PD-L1 outperformed either agent alone. The effect required both CD8+ T cells and NK cells to be present, and neither arm alone accounted for it.⁵³ Two further details make the relationship look less like addition and more like interdependence. N-803 itself upregulates PD-L1, so the cytokine creates the target the checkpoint inhibitor blocks. And IL-15 preferentially expands TCF1+ progenitor-exhausted T cells — which are the specific population that responds to PD-1 blockade, rather than the terminally exhausted cells that do not.⁵³ Each agent supplies what the other requires.

The position taken here is that complementarity is a credible mechanism, an unproven clinical proposition, and already priced. The NSCLC and lymphopenia lines in the sum-of-the-parts are precisely this option — carried at roughly 30–35% and 15–20% probability respectively rather than at franchise value. ResQ201A is what converts it from mechanism to evidence, or ends it.

08Second-line NSCLC — the larger prize and the harder fight

ANKTIVA holds accelerated approvals for metastatic NSCLC in Saudi Arabia (January 2026)³¹ and the UAE (July 2026)³² via the subcutaneous route — the first subcutaneous IL-15 superagonist approvals anywhere. There is no US or EU approval. The confirmatory Phase 3, ResQ201A²³ (ANKTIVA + tislelizumab + docetaxel versus docetaxel, 2:1, 507 patients, overall survival primary), began October 2025 with estimated primary completion September 2028 — implying a realistic US approval window of 2029–2030.

The TAM argument is sound: 2L+ NSCLC is worth multiples of the entire BCG-unresponsive bladder market, and a 15% share would exceed ANKTIVA's total current opportunity. The difficulty is what will already be established by 2029.

The more realistic commercial path may be combination rather than competition. ANKTIVA's mechanism — restoring NK and T-cell competence — is logically complementary to ADCs and bispecifics rather than rivalrous with them. ImmunityBio is not currently running such a registrational combination, and that is a strategic gap worth watching.

The route the program has not taken#

ANKTIVA’s one approval rests on regional delivery. Instilled in the bladder it reaches the tumor at high concentration while systemic exposure stays below the limit of quantitation, which removes the dose-limiting toxicity that has defeated cytokine after cytokine in the blood. Two things about that are worth knowing. Interleukin-2 was instilled in bladders from the early 1990s and given into the abdomen for ovarian cancer, with documented activity and not one registration to show for it.¹¹⁷ And ANKTIVA did not arrive there by retreat. The systemic and intravesical programs ran together from 2013 and 2014, and the systemic monotherapy in solid tumors produced no objective responses at all.¹¹⁸ What separates this company is a molecule that could use that route, and a willingness to keep going in the compartment that worked. Every lung program runs the other way: QUILT-2.023, QUILT-3.055, ResQ201A and the Lung-MAP substudy all give N-803 systemically.⁶⁷

The lung is the one other organ where the bladder logic transfers, because it can be dosed directly. Nebulized IL-15 has been tested in dogs with spontaneous pulmonary metastases, where fourteen days of monotherapy produced clinical activity. Dogs are a better model than mice here, because the tumors arise on their own in an intact immune system. Two qualifications belong with that. The molecule was recombinant human IL-15 rather than a superagonist, so it is the route that transfers rather than the drug. And no inhaled IL-15 trial in humans exists or is planned, by ImmunityBio or anyone else, which makes this an argument about a road untaken rather than a program in waiting.⁹⁴ Baseline lymphopenia correlated with benefit, which is the same association the company reports from its own NSCLC data.⁶⁶

This is an observation about strategy rather than a revision to the forecast. No inhaled IL-15 has been tested in humans, ANKTIVA has no such presentation, and nothing here changes a number in this model. But it sharpens what the systemic program is betting on. ResQ201A must win in the compartment where this class has repeatedly lost, using the route that produced cytokine release syndrome in more than half of patients in one prior IL-15 study, while an approach closer to the one that actually worked sits untested. If the readout disappoints in 2028, the question worth asking is whether the molecule failed or the route did.

Why 14.1 months cannot be compared to 10.5#

The reasoning matters more than the conclusion here. QUILT-3.055 is single-arm, n≈79–86, with median overall survival of 14.1–14.6 months. The comparators it is measured against are randomized: docetaxel plus ramucirumab produced median overall survival of 10.5 months in REVEL (n=1,253). A single-arm median cannot be set against a randomized median and read as superiority.

The reasons are structural rather than technical. QUILT-3.055 enrolled patients who had responded to prior checkpoint therapy and then progressed with no intervening treatment — a favorable-prognosis, checkpoint-sensitive group, not REVEL's all-comers. Single-arm designs additionally carry performance-status enrichment, eligibility restriction, and era effects: supportive care has improved since REVEL enrolled in 2010–2013. The oncology methods literature is consistent that single-arm results systematically overestimate benefit relative to the randomized trials that follow.

The ALC-responder analysis has the same problem in sharper form. Responders showed median overall survival of 16.2 months against 11.8 months for non-responders. But patients well enough to mount a lymphocyte response are patients well enough to live longer, and they are also on treatment longer. That is reverse causation, and only randomization separates it from drug effect.

The drug may well work. What is unknown is the magnitude, and ResQ201A exists precisely to find out. The defensible position is that ANKTIVA is a promising option in a setting with poor alternatives, and its differentiation is unproven, rather than demonstrably the best. What would settle it is a single number: the ResQ201A overall-survival hazard ratio in 2028. A confidence interval crossing one ends the systemic thesis and takes this line to zero; clear separation would move the pipeline from option value to a forecast.

09Immune reconstitution: lymphopenia, NK cells and sepsis

Everything in this section rests on one proposition: that a depleted immune system is itself worth treating. ANKTIVA raises lymphocyte counts, and the founder has argued since the 2017 IND that doing so should be as routine as correcting anemia or neutropenia — a product the company’s Founder’s Vision document goes as far as naming, proposing ANKTIVOGEN on the model of EPOGEN and NEUPOGEN.¹⁴³ If that view ever prevails, the addressable population is not a cancer indication. It is a fraction of everyone who has been through chemotherapy, radiation, transplant, sepsis or old age.

The scale is the reason to keep the possibility in view at all. In the general adult population, one in five US adults carries a lymphocyte count at or below 1,500 per microliter and three percent sit at or below 1,000.⁹⁷ Narrow it to people receiving cancer treatment and the numbers are still enormous: more than half a million Americans receive radiotherapy each year, immunosuppression follows in up to seventy percent of them, and severe lymphopenia in roughly a quarter.⁹⁸ Add chemotherapy, transplant recipients, the chronically immunosuppressed and post-sepsis survivors and the population runs to millions. No drug is approved anywhere to treat any of it.⁹⁹

That last sentence carries the argument in both directions. A market with no incumbent is either an extraordinary opportunity or a category that has resisted creation for good reason.

The size of that opportunity sits outside the valuation. The bear, base and bull cases in Section 16 run to 2032 and carry lymphopenia at thirty million, five hundred million and fourteen hundred million dollars, which are probability-weighted revenue lines rather than market sizes. What follows is the market size — a ceiling rather than a forecast, reachable only on a label that no regulator has granted anyone.⁸⁹ It is set out so the reader can apply their own probability and reach their own number, which is the only honest way to carry an option this far from resolution.

There is no approved therapy and therefore no price. The nearest guide is intravenous immunoglobulin, the one established market that sells immune replacement rather than a cancer treatment. Medicare pays roughly ninety-one to ninety-nine dollars a gram.¹¹⁹ Replacement dosing for primary immunodeficiency runs 0.4 to 0.6 grams per kilogram every three to four weeks, which for a typical adult is thirty to fifty thousand dollars a year in drug cost. The neurologic indications, dosed at immunomodulatory levels, cost between $108,016 and $136,892 per patient-year.¹²⁰ Those are drug costs alone, and infusion and facility fees are charged on top. A hundred thousand dollars a course is the figure used below, at the upper end of what immune replacement already commands.

The two columns differ by a factor of ten, and the difference is one question: whether a label would cover a deficit that resolves on its own. In stage III lung cancer, 59% of patients developed a grade 3 or worse deficit during chemoradiation and 89% had recovered within three months.¹²⁴ A therapy indicated for the persistent remainder addresses tens of thousands of Americans a year. One indicated during treatment addresses a quarter of a million. The regulatory record is silent on which, and our search of the FDA, EMA and WHO registers found neither granted.⁸⁹

Outside the United States the population is larger and the market is not. Europe sees 2.1 times the American case count and Asia 4.6 times,¹²⁵ and neither system reimburses supportive care at American specialty prices. At this price it would be a cash product abroad. Such markets do exist: self-purchased drugs account for 43.9% of cancer medical costs in China,¹²⁶ and commercial CAR-T there sells at comparable prices to patients paying for themselves. How much converts is an assumption this report declines to make, and the ex-US row is left empty rather than filled with a guess.

One thing is unresolved and bears on all of it. ANKTIVA raises the count; the NK programs supply the cells. Whether immune reconstitution turns out to be a drug, an infusion, or the two together changes the cost of goods, the site of care and the margin. The record does not settle it, and a market this size with the product undefined is the shape of an option rather than a business.

Three things decide whether any of it arrives: what a label would require, the cell assets meant to exploit it, and the one trial carrying an endpoint a regulator already accepts. Each is taken in turn.

What it would take to treat lymphopenia#

The obstacle is not the biology. It is that no regulator has ever approved a therapy for an acquired lymphocyte deficiency.⁹⁹ Every comparable approval either rested on a clinical event or attached to a disease with a defined cause. Filgrastim was approved to decrease the incidence of infection as manifested by febrile neutropenia, not to raise the neutrophil count. Immunoglobulin replacement for primary immunodeficiency rests on serious bacterial infection rates. Interferon gamma-1b for chronic granulomatous disease rested on time to serious infection. Romiplostim was approved on a platelet surrogate, but for immune thrombocytopenia — a defined autoimmune disease rather than a low count.⁹⁹

Two further precedents sharpen what the obstacle is and is not. It is not mechanistic uncertainty. The FDA has approved immunoglobulin for chronic inflammatory demyelinating polyneuropathy, multifocal motor neuropathy and dermatomyositis on labels stating that the mechanism has not been fully elucidated, and each of those approvals rested on a functional endpoint: a disability score, grip strength, a composite of clinical measures.¹⁶⁷ The agency has never required a mechanism. It has required an outcome, and ANKTIVA’s mechanism is better characterized than immunoglobulin’s.

Nor would approval settle payment. Immunoglobulin in secondary immunodeficiency is the closest structural parallel to lymphopenia — acquired, measurable, and secondary to another illness or its treatment. Medicare covers it only where a low immunoglobulin level is accompanied by documented recurrent or serious infection, and the laboratory value alone has never been sufficient.¹⁶⁸ A lymphopenia label would face the same question twice, once at the agency and again at the payer, and a count restored without an event prevented answers neither.

The counter-example is exact. Immunoglobulin lowered plasma amyloid-beta in Alzheimer’s disease and missed both the cognitive and the functional endpoint in a 390-patient Phase 3. The program was abandoned.¹⁶⁹ A biomarker that moves while the clinical endpoint does not is the outcome this thesis has to outrun.

The nearest thing to a precedent is a cautionary one. Idiopathic CD4 lymphocytopenia is a recognized disease with its own diagnostic code, studied at the NIH for three decades, and no therapy has ever been approved for it anywhere.¹⁰⁰ Interleukin-7 raised CD4 and CD8 counts reliably in that setting and went no further.¹⁰⁰ The same drug met its lymphocyte endpoint in glioblastoma patients with treatment-related lymphopenia, a 148% rise against 17% on placebo. A decade later it remains unapproved, with the registrational trial and the breakthrough designation both still outstanding.¹⁰¹ Raising the count has been achievable for a decade. Converting that into a label has not.

Three things distinguish ImmunityBio’s position from that record. The drug is already approved and selling, so the safety database and the manufacturing exist. The RMAT designation and the pan-tumor expanded-access authorization are further than any prior entrant reached. And the position is not a retrofit: the founder has argued it consistently since 2017, before there was oncology data to explain away. None of that is a regulatory path. It is what separates an ambition with institutional footing from one that has none.

What the path actually requires is a clinical endpoint in a defined population — fewer infections, fewer treatment delays, or longer survival in patients made lymphopenic by chemotherapy or radiation. Not the count. That is a multi-year undertaking, no such trial has been disclosed, and this model prices it accordingly.

How a lymphopenia market would actually start#

One fact deserves more weight than the off-label framing gives it. In February 2025 the FDA granted Regenerative Medicine Advanced Therapy designation to ANKTIVA plus PD-L1 t-haNK for reversal of lymphopenia in patients receiving standard-of-care chemotherapy or radiotherapy. RMAT is available only for conditions the agency accepts as serious or life-threatening. In granting it, FDA implicitly acknowledged lymphopenia — in that defined population — as a serious condition worth treating.

That is materially more than any prior entrant in this space achieved. Interleukin-7 programs from NeoImmuneTech and RevImmune spent more than a decade demonstrating they could raise lymphocyte counts without ever securing a comparable regulatory foothold or reaching market. ANKTIVA arrives with an approved product, a manufacturing base and commercial infrastructure. It also holds an FDA designation carrying rolling review, intensive agency interaction and eligibility for accelerated approval.

The condition also has an unusual profile for category creation: it is measured on every routine complete blood count, so no new diagnostic is required; its prognostic harm is documented across a dozen tumor types; and no therapy is approved for it in any jurisdiction.⁸⁹ A recognized serious condition with a validated biomarker, a free universal test, and no competitor is a configuration that rarely holds all at once.

RMAT accelerates the path without solving the endpoint problem. It does not change the requirement that ImmunityBio demonstrate a clinical benefit — fewer serious infections, fewer treatment delays, or improved survival — rather than ALC correction alone. Nor does it extend beyond the designated population and combination. The bull case below therefore assumes RMAT converts into an approval in a defined chemotherapy- or radiotherapy-induced lymphopenia population, not a broad immune-reconstitution label.

On demand dynamics. Physicians do not currently treat lymphopenia — no NCCN, ASCO or ESMO guideline recommends it, and ALC is reported on every blood count but acted on by almost no one. Neutropenia became a multi-billion-dollar market because it causes acute, visible, life-threatening infection and because filgrastim gave oncologists something to do about it. Lymphopenia has neither driver: its harm is a statistical survival decrement visible only in retrospect. This is a demand-push market, and adoption will be slower than the epidemiology alone suggests.

So an approval, if one comes, opens the first gate of four. Guidelines would have to name lymphopenia as something to act on before prescribing becomes routine; payers would have to cover a supportive-care biologic; and the price would have to travel from where it was set — $35,800 a dose for a market of thousands — to somewhere a market of hundreds of thousands can clear. The Dunkirk build is sized for the far side of all four gates at once.

The exception, and the logical beachhead, is radiation oncology. That community already pays to modify practice — proton therapy, reduced field size, vertebral-body sparing — specifically to protect lymphocytes. Clinicians who spend money on technique to preserve ALC are clinicians who believe the problem is worth solving. A pharmacological option would be new to them, and the conviction should transfer.

The evidence they are acting on is unusually consistent. Across fifty-six studies and 13,223 irradiated patients, 37% developed lymphopenia, with an overall-survival hazard ratio of 1.70.⁸⁰ The seminal cohort found 43% severely and persistently lymphopenic two months after starting chemoradiation, with a death hazard ratio of 2.1 across gliomas, pancreatic and lung disease.⁸¹ Counting only the tumor types where clinicians already alter technique, roughly 80,000 US patients a year develop severe radiation-induced lymphopenia — a larger annual population than the entire BCG-unresponsive bladder opportunity, in a specialty already spending money on the problem.

One shortcut does not exist, and it is the one most likely to be assumed. Medicare must cover off-label uses of anticancer drugs listed in a recognized compendium, which is how many oncology drugs earn reimbursement outside their labels. That rule does not reach here. The statute covers drugs "used in an anticancer chemotherapeutic regimen," and raising a lymphocyte count is supportive care rather than tumor-directed treatment. The precedent points the same way: filgrastim, pegfilgrastim, epoetin and darbepoetin each obtained their own FDA supportive-care indication rather than relying on a compendium listing.⁸³ ANKTIVA would have to do the same. There is no cheap route from expanded-access data to reimbursed US use, which is why the bear case for this line is close to nothing rather than merely small.

The NK cell assets#

The scientific logic is sound and is the strongest version of the platform argument. IL-15 is the dominant survival and proliferation cytokine for NK cells. Infused NK cells characteristically fail because they do not persist. Pairing an IL-15 superagonist with an NK infusion to sustain those cells in vivo is a coherent design, and ANKTIVA drives roughly 20-fold NK expansion with increased cytotoxicity in preclinical work.

If that held in patients it would produce a razor-and-blades business: every NK infusion would pull ANKTIVA volume behind it. The human evidence does not yet support it. In haploidentical NK therapy, N-803 support accelerated host T-cell rejection of the infused cells rather than sustaining them, and in ovarian cancer the NK expansion it drove was not maintained beyond two weeks despite continued dosing.¹⁰⁴ The mechanism is sound and the clinical demonstration is missing, which is the same sentence this report writes about most of the platform.

The economics are also better than CAR-T. An off-the-shelf NK-92-derived product follows a one-donor-to-thousands-of-doses model. Autologous CAR-T lists at $370,000–$530,000 with cost of goods around $48,000–$115,000 per dose; allogeneic estimates run near $40,000 falling toward $10,000–$20,000 at scale. ImmunityBio holds GMP capacity in El Segundo and Dunkirk, New York.

The manufacturing base is built and paid for, and it is the part of the platform claim that does not depend on a future trial. The company reported more than five trillion clinical-grade NK cells manufactured since 2017 with over 2.7 trillion held cryopreserved, and its memory cytokine-enriched program yields up to ten doses from a single apheresis — one session drawing and separating a patient’s blood cells — within twelve days.¹⁰² A bank of that size is not a press release; it is capital already spent.

What it buys is bounded by a fact of the cell line. NK-92 derives from a lymphoma patient and is tumorigenic, so every dose must be irradiated before infusion. Irradiated cells cannot proliferate, persist about two days, and form no memory — which mandates repeat dosing and rules out the one-infusion economics that make CAR-T valuable despite its cost.¹⁰³ The competitors that avoid this constraint use induced pluripotent or donor-derived cells, and they are the ones the field backed.

Sepsis — the most endpoint-friendly indication in the pipeline#

Sepsis-induced immunoparalysis is driven by apoptotic depletion of T and NK cells, and late-phase sepsis deaths are dominated by secondary infection in immunosuppressed patients. Persistent lymphopenia on day four³⁰ independently predicts 28-day and one-year mortality. The global burden is 48.9m cases and 11.0m deaths²⁹ annually — roughly a fifth of all deaths worldwide.

ImmunityBio has registered NCT07578558²², a randomized open-label Phase 2 of ANKTIVA plus standard of care versus standard of care alone in critically ill adults with sepsis and persistent lymphopenia. Roughly 50 patients, estimated start July 2026, primary completion August 2027.

This matters disproportionately for one reason: sepsis supplies a hard clinical endpoint that the lymphopenia indication lacks. Twenty-eight-day mortality is unambiguous, FDA-acceptable, and fast to read out. It sidesteps the surrogate-endpoint problem that makes an oncology lymphopenia approval so difficult. A biomarker-enriched population — patients selected for persistent lymphopenia — also addresses the heterogeneity that sank previous immunomodulatory sepsis trials.

Against that: sepsis is a graveyard. Interleukin-7, GM-CSF, interferon-gamma and checkpoint blockade have all been trialed with equivocal results. And there is a specific negative signal — an IL-15 superagonist in burn-wound sepsis expanded lymphocytes but failed to improve bacterial clearance or survival. Expanding cells is not the same as improving outcomes, and that study is the most directly relevant preclinical datapoint available.

10Whether this is a platform or a bladder-cancer company

The case for valuing ImmunityBio as more than a bladder-cancer company rests on the subcutaneous program: NSCLC, and beyond it the lymphopenia concept — restoring absolute lymphocyte count in patients depleted by chemotherapy, radiation, age, or infection. The population runs to hundreds of thousands a year in the United States alone, sized in Table 12, and low ALC is robustly associated with worse outcomes across settings.

Two things are settled here and should be said plainly. That ANKTIVA raises the count is not the open question — the agency conduct described in Section 09 presupposes it, naming reversal of lymphopenia in a designation and authorizing an access program that doses patients against a lymphocyte threshold. And the absence of any lymphopenia label to date is not evidence the door is closed, because supportive-care labels are not written for counts. Filgrastim’s was granted for cutting infection in a defined chemotherapy population, not for raising neutrophils; the count is the mechanism, the outcome is the indication.¹⁴⁹ The open question is the translation: whether a restored count means fewer serious infections, fewer interrupted treatments, or longer lives, and in which named population. Behind it stands the rest of the stack — a route, a prescriber, a payer, a price.

The distinction here is what ALC is being asked to do. In every oncology trial in this class it has been a surrogate — a count that rises reliably while the clinical endpoint does not follow. Bempegaldesleukin expanded every immune subset measured and missed every endpoint; nanrilkefusp alfa raised CD8+ and NK counts and was discontinued for insufficient efficacy.⁶⁵ Lymphocyte expansion is the dependable part of this class and clinical benefit the unreliable part, which is why raising a count proves less than it appears to.

In lymphopenia the count is not standing in for the disease. It is the disease. That changes what the same pharmacology has to prove, and it is the strongest structural argument for the indication. It does not remove the regulatory obstacle: ALC remains a biomarker rather than a clinical outcome for approval purposes. The FDA accepts surrogate endpoints for accelerated approval only where they are reasonably likely to predict clinical benefit, and ALC has not been validated as such. Approving "treatment of lymphopenia" as an indication would in practice require demonstrating a clinical benefit — fewer infections, fewer chemotherapy delays, improved survival — in a defined population. No registrational lymphopenia trial with a clinical endpoint has been disclosed. This is a multi-year, high-risk endeavor, and the probability of a standalone approval on ALC restoration alone is low. The shape of the path, though, is known, because filgrastim walked it: a randomized trial in a named chemotherapy population, an infection endpoint the agency accepted — fever arriving on a collapsed neutrophil count, in 40% of treated patients against 76% on placebo — and a 1991 approval written on the outcome, not the count.¹⁴⁹ Run the same sequence here — endpoint agreement, registrational trial, two to four years of enrollment and follow-up, review — and the earliest plausible lymphopenia label lands around 2030, unless the sepsis readout in August 2027 or a lung approval shortens the argument. RMAT compresses the meetings, not the evidence. A first systemic US label would also be the approval of the syringe here. The US label is intravesical only and forbids subcutaneous dosing.³ The subcutaneous presentation already exists, approved in Saudi Arabia and the UAE for lung cancer, as Section 12 sets out. A supportive-care drug taken at population scale is a subcutaneous product — filgrastim again — not a chair in an infusion suite.

The wider pipeline — pancreatic (QUILT-88), glioblastoma, HPV-positive head and neck, the M-ceNK cell-therapy platform, hAdV5 vaccine vectors — is broad on paper but largely Phase 1/2, single-arm, or not yet enrolling. Against $66.5m quarterly burn and a going-concern footnote, breadth is as much a risk as an asset: it spreads scarce capital across programs that may all end up underpowered.

Systemic IL-15 has a difficult history, and the closest precedent is closer than it looks. Sotio’s nanrilkefusp alfa is the same architecture — an IL-15 receptor-α sushi domain fused to IL-15 — given subcutaneously. It raised circulating CD8+ T and NK cell counts, spared regulatory T cells, and was well tolerated. Sotio discontinued it in October 2023 for insufficient efficacy.⁶⁵ Novartis halted NIZ985, an IL-15 and receptor-α heterodimer, short of enrolment. ANKTIVA's success has come precisely in the local, intravesical setting where systemic toxicity is avoided; the subcutaneous ambition moves it back into the arena where the class has repeatedly failed. The specific cytokine-plus-checkpoint precedent, and why the complementarity argument survives it, is examined in Section 07.

The factory as the tell#

Arguments about platforms are cheap, so weigh the physical evidence instead. Section 02 counts the Dunkirk fill line as a burden — a million vials of annual capacity against roughly 7,000 doses of demand, expensed as idle — and as accounting that is correct. As a statement of intent it reads differently. A million-vial line, plus El Segundo drug substance the company itself sizes at a million doses a year, is not scaled to a bladder-cancer franchise, and the company said as much at the moment of purchase: the 2022 release announcing the Dunkirk acquisition was titled for global pandemic response and preparedness, and counted nearly a million square feet of manufacturing across the US, South Africa and Botswana.¹⁴⁴

The demand ladder makes the mismatch precise. The approved indication needs 7,000–13,000 patients a year. Papillary and BCG-naive approvals raise that to tens of thousands. Even the bull case’s 2032 revenue of $4.6bn implies, at ANKTIVA course pricing, on the order of 120,000 doses a year — roughly an eighth of the Dunkirk line alone, before El Segundo. The first demand scenario that sizes the plant correctly is the one this section has been circling: the radiotherapy arithmetic in Section 09 alone implies over a hundred thousand severe-lymphopenia patients a year — more than a million doses at the median course — and the broader depleted population the section counts in millions would run both plants hot. The alternative that also closes the arithmetic is global BCG and rBCG supply against a shortage Merck caps at 600,000–870,000 vials, examined in Section 05.

The founder has named the product for that world. The company’s Founder’s Vision document proposes ANKTIVOGEN for the lymphopenia indication, explicitly on the model of EPOGEN and NEUPOGEN — agents used across all tumor types regardless of anatomy.¹⁴³ Read against that document, the March 2026 warning letter — treated in Section 17 as a conduct matter — records something more specific: a founder who does not draw the line between the vision and the product, corrected by the agency whose statutory job is that line. The FDA’s letter quotes him describing ANKTIVA as “approved for bladder cancer, but it actually can treat all cancers,” beneath a banner reading Cancer Therapeutic Vaccine (BioShield)⁴ — the ANKTIVOGEN future spoken in the present tense. The agency was right to object, and the objection settles nothing about whether the vision is achievable; it establishes that the product, today, is a bladder-cancer drug. The plant is being built for the other thing.

The record of building ahead of demand#

Whether that belief deserves capital depends on the base rate, and the record is checkable. Cancer MoonShot 2020, launched in 2016 with a promised cancer vaccine by 2020, produced little that its own deadline year could point to.¹⁴⁵ NantHealth listed at $14 in 2016; reporting that a $12m university donation had been structured to send $10m back to his own company took the shares down 23% in a single session, and the securities suits followed.¹⁴⁶ Verity Health, whose manager he acquired in 2017, filed Chapter 11 in August 2018 under more than $1bn of bond debt and pension liabilities.¹⁴⁷ NantSA, launched beside the South African president in January 2022 around a pledge of a billion vaccine doses a year by 2025, has disclosed no production at that scale.¹⁴⁸ Dunkirk itself sat largely idle from the 2022 purchase to the 2024 approval. Against the pattern sit Abraxane, which reached approval and made a fortune, and ANKTIVA’s own franchise — eight consecutive quarters of growth is not nothing. The base rate reads plainly: when this founder builds ahead of demand, the building arrives and the demand, so far, mostly has not — and the exceptions are exactly the products that survived a regulator.

The honest reading for a holder is that the capacity is the measure of the bet, not evidence the bet pays. Stated at its narrowest, the founder’s wager is about timing — that the gates open while the plant stands ready — so the operative question the concrete asks is when, not whether. It converts to an asset on evidence, and the evidence has a short list.

What protects any of it is the subject of the next section.

11Intellectual property and the technical moat

The bull case in this report rests on programs that do not yet earn revenue. What they are worth depends on how long ImmunityBio can keep them to itself, and that question has not been asked anywhere above.

A platform thesis is worth what its exclusivity is worth, and ANKTIVA’s runs on two clocks rather than one. The composition-of-matter patent on the superagonist complex dates from a September 2010 priority and expires in September 2031 on its nominal term.⁶² That is the shorter clock, and for a drug approved in the US in 2024 it is not the one that matters most.

The longer clock is the combination. ImmunityBio holds five US patents covering ANKTIVA administered with BCG — the intravesical regimen, the two-vial kit, and methods in non-muscle-invasive disease including the BCG-naive setting — with terms running to 2035.⁶³ Since the approved product is the combination, and since the BCG-naive indication that carries most of this model’s 2028 revenue falls inside those claims, this family rather than the composition patent is the operative protection.

Underneath both sits regulatory exclusivity. As a biologic licensed under section 351(a), ANKTIVA is eligible for twelve years of reference-product exclusivity, with no biosimilar application accepted for four. Measured from the April 2024 approval those dates would be 2028 and 2036. The Purple Book has not yet populated a date of first licensure, so the figures are read from the statute rather than the register, and no orphan exclusivity is listed.⁶⁴

The estate reaches further than a single-product wall in two respects. One is provenance. The superagonist was invented at Altor BioScience, which ImmunityBio acquired, and the claims were assigned through the NantCell and NantKwest entities into the current group, an assignment history recorded on the face of the patent.⁶² They sit inside the company rather than under license from a third party. The other is scope beyond IL-15. ImmunityBio holds granted US patents applying the same receptor-α sushi-Fc architecture to fusions with IL-12, IL-18, IL-7 and IL-21.⁶⁸ That is the closest thing to a platform claim in the estate. These are patents and preclinical constructs rather than programs. Nothing built on that scaffold beyond IL-15 has entered a trial, by this company or any other. There is also a reason to expect the architecture to travel poorly. Trans-presentation is how IL-15 physiologically signals, so the scaffold reproduces the cytokine's own biology; for IL-12, IL-18, IL-7 and IL-21 it is a delivery chassis rather than a biomimetic, and each of those cytokines carries an unsolved problem of its own.

There is a further signal in where the field is putting its effort. The approach most likely to widen the therapeutic window for a systemic IL-15 is to target it — tether the cytokine to an antibody so it acts where the tumor is rather than everywhere. Sotio is doing this with SOT201, a PD-1-targeted attenuated IL-15 in Phase 1, and Xencor licensed an IL-15 receptor-α Fc construct to Genentech.⁹⁶ ImmunityBio has no equivalent program. Its systemic bet remains an untargeted subcutaneous injection, which is the older design in a field that has moved on. The estate is worth holding as defense against anyone building the same constructs. It is not evidence of hidden clinical assets.

No inter partes review, opposition or third-party validity challenge touching these patents was found in the public record.⁷⁰ That is an absence of evidence rather than proof of none, but for a drug approved two years ago carrying a related-party financing structure, the absence itself is a fact about the record.

Two things this does not establish. The first is that 2031 is fixed. ImmunityBio filed for patent term extension in June 2024 on several US patents covering the composition, methods of use and methods of manufacture of N-803, and those applications remain under review at the USPTO and the FDA. Only one patent can be extended per approval, and the outcome remains unpublished. Take the composition date at face value, then, but as an open question. A grant would move it toward the middle of the decade.¹⁸⁷ The second is that the estate protects this molecule in this combination rather than the superagonist concept, which has prior art outside the company.

One item belongs in a buyer’s diligence and sits outside the financial statements. The earliest granted composition patent carries a government-support statement: the supporting research, it says, was carried out by the United States as represented by the Secretary of Health and Human Services. The statement is unusual in form. It gives no grant or contract number, and it drops the standard recitation that the government retains rights, which leaves what attaches ambiguous on the face of the document. The continuation third parties cite as describing the marketed molecule is silent on the question. The company’s own risk factors, separately, describe march-in rights over government-funded technology it licenses. The public record shows an open question rather than an encumbrance, and the answer sits in agreements held privately.¹⁸⁸

The competitive test is already visible. Hengrui’s SHR-1501 is an IL-15 and sushi-domain IgG1 Fc fusion in trials for non-muscle-invasive bladder cancer by the intravesical route⁶⁹ — the same architecture, the same route, the same indication. Whether the ANKTIVA-plus-BCG claims exclude it is a question of claim construction rather than of biology, and it is the practical measure of what this estate is worth.

12The Gulf opportunity, and who would pay for it

ANKTIVA holds the only NSCLC approvals anywhere — Saudi Arabia (January 2026) and the UAE (July 2026). The question is whether affluent patients with checkpoint-refractory disease, facing few remaining options, will fund treatment themselves. The arithmetic is seductive: at Gulf pricing, a few hundred patients would exceed a quarter of current US revenue. The channel deserves testing rather than dismissal.

Two distinct opportunities sit inside it, with very different feasibility.

Domestic and regional private pay#

Wealthy Gulf residents and regional patients already in the catchment, treated at private hospitals, paying cash. This bypasses the NUPCO tender system entirely and is the more realistic near-term channel. Wealth is abundant: Saudi Arabia hosts 350,000-plus millionaires and the UAE was the world's leading destination for relocating millionaires in 2025. The constraint is disease arithmetic.

Against that pool, the self-pay-capable and clinically-willing subset is perhaps 15–25%, and real-world persistence is short — median overall survival in this population is roughly 14 months, so most patients receive well under the 24-month maximum. Insurance will not close the gap: UAE mandatory plans cap at AED 150,000–250,000 and Saudi CCHI-mandated cancer cover at SAR 500,000, both below a six-figure course. Those caps bind expatriates, who are the large majority of the population. Emirati nationals sit in a different system: Thiqa, the Abu Dhabi program administered by Daman, states no annual benefit limit per member.¹⁵⁵ Dubai runs a parallel top-up above the basic cap for low-salary workers, and it covers three cancers – breast, colorectal and cervical.¹⁵⁷ Lung is not among them. That is the mandatory floor, and against the floor the conclusion is cash or nothing.

The layer above the floor changes the reason rather than the answer. Top-tier domestic and international plans carry annual limits an ANKTIVA course does not approach.¹⁹¹ At that tier the constraint stops being the cap and becomes the experimental-treatment exclusion, which the reinsurance industry applies on a guideline basis: on-label use is not experimental, off-label use inside NCCN or ESMO guidance should not be classified as experimental, and off-label off-guideline use is excluded.¹⁹⁰ The lung indication is on-label in Riyadh and Abu Dhabi and appears in neither guideline. International wordings add a second test: that treatment be consistent with generally accepted standards of medical practice in the country where it is received.¹⁹¹ A single-arm approval struggles to pass that test. Saudi domestic insurers hold the most discretion, since no CCHI rule limits cover to registered or formulary drugs, but six-figure oncology is pre-authorized against cheaper guideline options; UAE cover is gated by the Dubai formulary and the Abu Dhabi threshold this section reaches below. We found no GCC insurer listing, covering, pre-authorizing or denying the drug. So the conclusion survives softened rather than reversed. For lung, perhaps a tenth of the eligible pool at the outside could route a course through an enhanced or international plan — a slice the funnel’s 15–25% self-pay subset already spans — and an insured market would clear at a lower net price than a cash one, because insurers bring the bargaining the cash channel avoids. For bladder the test cuts the other way: that indication is FDA-approved and sits in the NCCN guideline as of March 2026,⁴⁹ so it clears the exclusion lung fails, and insurance is a live channel rather than a workaround.

The approved posology fixes the unit of account. The Emirates label gives ANKTIVA as a fixed 1 mg subcutaneous dose once every 21 days for the duration of checkpoint therapy — one vial per dose, roughly seventeen doses across a full year on treatment.³² Twelve doses is what median survival in this population typically allows, and twelve is what the arithmetic here runs on.

The optimistic case has a ceiling, and the arithmetic sets it tightly. Take the top of the funnel — ninety patients, twelve doses each, US list price with no discount — and the answer is roughly $39m a year. That is an enormous number against a base case of $3m for 2027, and it is why this line carries more variance than anything else in the model. It is also a long way below a hundred million, which would need twenty-eight doses a patient, more than twice what survival in this population allows. The constraint is the disease, and it does not move.

The asymmetric case is the one next door. Lymphopenia reaches a population two orders of magnitude larger, and the gate is regulatory rather than epidemiological — a formal Gulf indication converting lawful off-label use into reimbursed prescribing. That is the difference between the $12m base and the $130m bull in Table 13, and unlike the NSCLC ceiling it is not capped by how many patients exist. Whether it happens is unknowable; the point is that the upside lives there rather than in private-pay lung volume.

Why travel is the only route for NSCLC#

Patients with money and a terminal diagnosis do relocate for treatment — that much is not in doubt. The question is whether they must travel to get it, and for this indication they mostly must.

The obvious workaround would be for a US oncologist simply to prescribe ANKTIVA off-label — it is already FDA-approved here, and off-label prescribing of approved drugs is legal under 21 U.S.C. § 396. We tested that route and it is foreclosed on formulation grounds.

Three things stand in the way of a US physician reconstructing the Gulf dose from intravesical vials. The concentration differs. The US product is filled and tested for instillation rather than injection, and the label prohibits that route. And reimbursement fails at the first test: ANKTIVA appears in no CMS-recognized compendium for NSCLC, which is what Medicare's off-label anticancer rule under Social Security Act §1861(t)(2)(B) requires. Self-pay would run roughly $1.87m a year at three vials per dose every three weeks, using the wrong product by a prohibited route with no liability cover. We found no documented case of independent US off-label use.

The inbound case, stated as a breakeven#

Table 15 counts residents. That is the right denominator for a domestic market and the wrong one for this drug, because the argument above establishes that a patient in Boston, Munich or Tokyo who wants the studied regimen has three options, and two of them are trials. The subcutaneous presentation is on-market in two countries and nowhere else. Every checkpoint-refractory lung cancer patient on earth who can pay and travel is therefore addressable in principle by those two countries, which is a different proposition from the funnel in Table 15 and is not modeled anywhere in this report.

Sizing it by assumption would produce a number with no defensible provenance, so the arithmetic is run backwards instead. At the US list price of $35,800 a dose and the twelve doses persistence allows, a patient-year is about $430,000. A hundred million dollars of annual Gulf NSCLC revenue is therefore roughly 230 international patients on therapy; a quarter of a billion is roughly 580. Set against world lung cancer incidence of 2,637,005 in 2024,¹⁵⁴ the first of those is about one patient in eleven thousand, and the second about one in forty-five hundred.

That framing is the useful one because it moves the question off epidemiology. The pool is not the constraint at these volumes. Three other things are, and none of them is currently knowable. Conversion: whether a patient progressing on checkpoint therapy will relocate for an injection every three weeks for up to two years, which is a different commitment from flying somewhere for a surgery. Throughput: the centers and the infusion capacity exist, since Cleveland Clinic Abu Dhabi reports more than 22,000 chemotherapy infusions since its oncology center opened in 2023¹⁵⁸ and Burjeel Medical City holds ESMO accreditation for integrated oncology.¹⁵⁹ What remains unestablished is narrower and more specific: whether either would run a two-year course at three-week intervals for non-resident patients, and at what volume. And the window: an approval anywhere else closes the arbitrage entirely, with the confirmatory readout expected around September 2028 the most likely closing date.

Why NSCLC approvals matter beyond lung cancer#

There is a structural consequence of the Gulf NSCLC approvals that the lung-cancer framing obscures. The subcutaneous 1.2 mg presentation is now a registered and commercially launched product in Saudi Arabia (registration 78609-1572-260487 in the UAE; SFDA registration certificate with pricing issued in Saudi Arabia). Launched carries its qualifiers: distribution runs through two named distributors with initial patients identified, the certificate prices the product but the price is unpublished, no units or treated patients have been disclosed, and through the June quarter the 10-Q attributes all product revenue to US sales.¹ Off-label prescribing is a lawful physician prerogative in both jurisdictions — peer-reviewed surveys describe it as common practice in the Kingdom, and nothing in the UAE framework prohibits it for a registered biologic.

The consequence: a physician in Riyadh or Dubai can lawfully purchase an injectable ANKTIVA vial and prescribe it off-label to raise the lymphocyte count of a patient who does not have lung cancer. That is impossible in the United States, where only the intravesical presentation exists and the label forbids the route. Every jurisdiction that approves subcutaneous ANKTIVA for NSCLC therefore simultaneously creates a lawful supply channel for off-label immune reconstitution. The NSCLC filing footprint is the gating variable for a much larger addressable population.

Dosing follows the registered label — a fixed 1 mg subcutaneously every three weeks, one vial per dose.³² At an assumed Gulf net price of $20,000–$40,000 per vial, chronic use runs $350,000–$700,000 per patient-year; a several-month adjunctive course runs $100,000–$250,000. Those are plausible luxury-medicine numbers for a small number of patients, and the self-pay argument holds here in a way it does not in the United States: where the product is purchasable, money solves the access problem.

What the Gulf approvals unlock beyond lung cancer#

Section 09 sizes this market and sets out the regulatory obstacle. The Gulf is where the off-label version is already possible, and the base case carried there from 2027 assumes exactly that: a small line, immaterial to 2028, that compounds and requires no US regulatory event. A formal Gulf lymphopenia indication would convert it from off-label to on-label, and ImmunityBio has said it is pursuing label expansion for chemotherapy-induced lymphopenia and has opened discussions with both Gulf regulators about indications beyond lung and bladder.

What makes that path shorter than it looks is revealed preference. The SFDA granted the world’s first NSCLC approval for this drug, and the world’s first subcutaneous approval, in January 2026 — ahead of the FDA and the EMA, on single-arm data.³¹ That is not a general claim about Gulf regulators moving quickly. It is a specific demonstration that one regulator will act on this asset before the agencies that usually go first. A Gulf lymphopenia indication would face the same surrogate problem the FDA has not resolved, but it would face it in a jurisdiction that has already shown where it sets the bar.

Neither fact justifies raising the estimate. No Gulf revenue has been recognized, prescriber willingness in Table 17 is moderate rather than high, and the surrogate problem is unresolved in every jurisdiction. What has changed is the probability attached to the bull case rather than the value of the base — and since that bull case is roughly five times the base by 2032, the distinction matters more than the arithmetic suggests.

Would the Saudi state pay?#

If state health programs fund treatment rather than individuals, the binding constraint shifts from household wealth to state budget — a materially larger and faster-moving opportunity. We tested this and the evidence came back thinner than the narrative suggests.

The state-funding pathway also has gates. A high-cost drug must clear SFDA pricing (done), mandatory health technology assessment, CCHI formulary listing, and NUPCO tender adoption — a process that since 2026 applies a "Made in Saudi" scoring weight favoring local manufacture. That process also has a calendar, which governs not whether the option pays but when. Public-sector medicines are bought centrally, and a product must carry a NUPCO catalog code before it can enter the main tender at all;¹⁵⁰ until then the route is hospital-level direct purchase, which is what private distribution currently substitutes for. The main tender is placed early in the year, and awards settle slowly: the complementary pharmaceutical tender of 2024 was announced in March and reached final results the following June.¹⁵¹ Catalog coding in one year therefore produces awarded volume in the next, which places the earliest meaningful state volume in 2027 or 2028. Saudi Arabia has funded high-cost novel therapies before: KFSHRC has treated more than 200 CAR-T patients and localized manufacturing, cutting cost per treatment from roughly SAR 1.3m to SAR 250,000. But those therapies carried US or EU approval.

And would the UAE?#

The UAE has the deeper pocket for its own citizens and the harder gate for this particular drug, which is an awkward combination for the thesis. Thiqa carries no annual limit, so for an Emirati patient the household-wealth constraint that governs the private-pay funnel does not apply at all. The constraint moves instead to what Abu Dhabi will agree to buy.

Since June 2025 that question has a published answer with a number attached. The Department of Health appraises high-cost technologies against a baseline cost-effectiveness threshold of 0.75 times GDP per capita for one quality-adjusted life year, with multipliers for severity, health gain and rarity, the largest of which is rarity at up to three times.¹⁵⁶ At an Abu Dhabi GDP per capita near $74,400 that baseline is roughly $55,800 a QALY. ANKTIVA at twelve doses of US list runs about $430,000 a patient-year, so clearing the baseline would take a gain near 7.7 quality-adjusted life years, and clearing it with the full rarity multiplier would still take about 2.6. Median overall survival in the trial that supports the indication was 14.6 months. The gap is not a rounding difference and no plausible discount closes it.

The rarity multiplier carries a second sting. It is keyed to diseases classified as rare by the EMA or the FDA, and the reliance pathway that speeds Emirates Drug Establishment review works the same way, crediting prior approval by a stringent authority. So the UAE has written the question of Western approval into its own rules rather than leaving it as precedent. Section 12 records that no case was found of Saudi state funding for a therapy lacking US or EU approval; the same search for the UAE returns the same answer, and here the criterion is codified rather than merely unbroken. What that means in practice is that the EMA filing is the swing factor for UAE reimbursement, more than any Gulf commercial development is.

Does supply gate the ramp?#

No, and the reason belongs on the record because the vial is not the US vial. The subcutaneous presentation is a distinct manufactured product — double the US concentration, a different fill, a different route — but it comes off the same chain: drug substance in El Segundo, fill and finish at Dunkirk, whose stated capacity is a million vials a year against released drug substance already sufficient for 170,000 doses.⁴⁸ The modeled ramp peaks at 25–40 patients on therapy, roughly 425–680 vials a year at the label’s cadence. What could bind instead is logistics: lot release for an imported biologic, cold chain, and the import cadence that runs through Modern Pharmaceutical Company in the Emirates and the company’s Saudi subsidiary with its two distributors.³² A larger vial, a new presentation or an added manufacturing site would route through the SFDA’s post-registration variation process and the Emirates establishment’s reliance pathway rather than through fresh approvals; the review clocks for such variations are not established from published guidance, which is worth knowing before assuming they are fast. Saudi localization scoring is a pricing incentive rather than a gate, and nothing requires local manufacture of an imported innovative biologic. The constraint on this line is demand and payment, not production.

What we have modeled, and what we have not#

The base case now carries an explicit Gulf private-pay NSCLC line, building from $0.5m in Q4 2026 to $8m per quarter by Q4 2028 — roughly $26m across 2028, on the order of 25–40 patients on therapy at any time at Gulf reference pricing. Combined with MENA bladder revenue, the region contributes about $60m across the forecast window. That bladder line is built from willingness and ability to pay rather than from a count of eligible patients, which is why the UAE indication being broader than the US one does not move it. Widening the label from CIS to all BCG-unresponsive disease multiplies the eligible pool, since papillary-only is the larger share of the two. It does not change who pays for it. The binding constraint in the Gulf is the payment route — no NUPCO listing, no ministry purchase, insurance caps below the cost of a course — and a wider label meets the same wall with more patients behind it. That lifts FY28 from $700m to $722m, or 1.8%, which is not a re-rating.

We have not put state funding in the base case. Sized separately: if the Saudi state funded the full eligible NMIBC and NSCLC population under a managed-entry agreement at a 40–60% discount to US list, it would add $40–80m annually from 2028 on these assumptions — more than private pay, because volume at a discount beats a handful of patients at list. That discount is a direction set by the pricing regime rather than a figure chosen for convenience. The SFDA prices by external reference to a basket narrowed to sixteen comparable countries in 2022,¹⁵² and has required a formal economic evaluation for new registrations since July 2025.¹⁵³ Both push the same way: in the state channel the upside is earned on volume, not price. That is the upside case, and it is a call option with identifiable triggers rather than a forecast. State funding of the approved indications and a formal lymphopenia indication are independent of each other, and not mutually exclusive. If both arrive, the second dominates, being the one not bounded by how many lung and bladder patients exist. No number is carried for the pair. The failure modes are independent too: the state can decline while the indication arrives, or the reverse, so tracking one of them is tracking half the option.

13Europe, and why the UK went private-pay

Ex-US contributes roughly $159m of FY2028, about 19% of the total. Three structural facts govern that line, and the first determines whether it should be modeled at all.

How European revenue is recognized#

The decisive question is whether ImmunityBio books end-market sales in Europe or a thin royalty from a licensee. The evidence points to the former. Accord Healthcare is described as a distribution partner²⁰ deploying an 85-person sales force within 100-plus commercial, medical and marketing staff across the UK, EU and EFTA — a services model, not a license. ImmunityBio simultaneously opened a Dublin subsidiary in February 2026 explicitly to support European distribution, and post-authorization EMA records list ImmunityBio Ireland Limited¹².

On that structure, European revenue should be modeled at ImmunityBio-captured net sales — end-market price less a distributor margin of 25–35% — rather than a royalty of a few percent. That is a large multiple of the alternative, and it is why the EU line carries weight here.

Two advantages Europe has#

BCG supply is materially better. The US depends on a single supplier. Europe has access to six approved BCG substrains¹³, including Medac's Danish 1331, Serum Institute's, and the Japanese Tokyo strain. Since ANKTIVA is dosed with BCG, the constraint that suppresses US uptake is substantially absent in Europe. This is an under-appreciated positive and argues for a faster European penetration curve than the US launch produced.

The competitive field is far emptier. ANKTIVA is the first and currently only authorized therapy in the EU for BCG-unresponsive NMIBC with CIS. ADSTILADRIN received a CHMP positive opinion only in March 2026 and is not yet authorized; INLEXZO is not EU-approved; cretostimogene is earlier still. That is a first-mover window of perhaps twelve months or more in a market where the US equivalent is already crowded with four agents.

How an approval becomes revenue#

The ex-US line built in Section 06 rests on markets that have cleared a regulator but not yet a payer. Six jurisdictions have authorized ANKTIVA, and one reimburses it. Between a regulator's approval and a dollar of recognized revenue sit four further steps — a price, a payer listing, commercial availability, and an invoice. The United States has completed all four. Elsewhere most of that sequence is still ahead, and the table below shows how far each market has come.

The company's own filing settles the question. The Form 10-Q for the quarter ended June 30, 2026 states that product revenue "was generated from U.S. sales of ANKTIVA."⁶¹ The geographic table carries a single non-US line — $518k for the quarter, $542k for the half, booked as other revenue rather than product sales. That line is not ANKTIVA. Other revenue is disclosed as bioreactor consumables and license royalty income,⁷⁹ which is legacy NantKwest business predating the drug — laboratory hardware and non-exclusive licenses on the NK cell lines. What the filing does not disclose is which of the two the European figure represents. Against $50.7m of quarterly product revenue it is immaterial, and it is the only ex-US line in the accounts.

The "approvals in roughly 34 countries" figure resolves mostly to arithmetic: the European authorization covers 27 member states plus Iceland, Liechtenstein and Norway, which is thirty of the thirty-four. The separately adjudicated jurisdictions number six. We verified no approval in any other Gulf state, in Latin America, Canada, Australia, Japan or Switzerland, and found none announced as filed.

Saudi Arabia is the furthest advanced and the most instructive. Approval came in January 2026 and commercial availability followed in April, inside three months — far quicker than any European timetable. Availability and a state contract are different things, though: SFDA pricing, a CCHI or MOH listing and a NUPCO award are each still outstanding, and no Saudi revenue has been separately disclosed. The Gulf opportunity in Section 12 should be read as a channel that has opened rather than one that has started paying.

What conversion would look like#

Each of the following would mark a market moving from approval into revenue, and none currently sits on a published calendar. A German launch or the start of a G-BA benefit assessment; a French accès précoce decision; any national price publication; a NUPCO award or CCHI listing in Saudi Arabia; a MOHAP price or emirate formulary listing in the UAE. The cleanest signal needs no interpretation at all — a separately disclosed ex-US product-revenue line in a future 10-Q. Until one of these appears, the ex-US line in this model rests on system norms rather than on anything the company has scheduled.

The UK is the one market where the company has declined a reimbursement route, and the reasoning follows. A NICE termination for non-submission means the product is not routinely funded in England and Wales. The company chose not to submit evidence. That is a commercial judgement rather than a regulatory setback, and a defensible one against a payer currently rationing standard-of-care oncology. It also avoids setting a reference price that most-favored-nation rules could pull back into the US line, a risk set out in Section 17. Private-pay supply remains available. The practical effect is that the UK contributes almost nothing until a resubmission, which was never a large part of this forecast.

Asia — modeled near zero, but not worthless#

ANKTIVA's only Asian approval is Macau, granted March 2026. No filings have been disclosed for Japan, China, Korea, Taiwan, Australia or ASEAN, so a near-zero 2028 assumption is correct.

One disclosure bears on the terminal value. The Oberland revenue-interest agreement is defined on worldwide net sales excluding China, with China defined to include the PRC and Hong Kong. A carve-out of that specificity implies China economics are structured separately and may sit with a partner or affiliate outside the main entity. Asian bladder-cancer incidence is roughly 215,700 cases a year, about 35% of the global total, with China near 92,800 and Japan near 34,600 — and Japan has both the region's highest incidence and the deepest BCG-use tradition.

The model carries a small rest-of-world line rising to $2.5m a quarter by end-2028, reflecting Macau and opportunistic named-patient revenue. The honest position is that 2028 is right and the terminal number is understated: an activated Japan filing or an identified China partner would be worth materially more than anything else in this section.

14Ownership, governance and the founder dependency

ImmunityBio is a controlled company, and almost every distinctive feature of its capital structure follows from that. Patrick Soon-Shiong and his affiliated entities held roughly 62.5% of the outstanding shares at the 2026 record date.⁷² Beneficial-ownership tables show a higher figure, near 66%, because they add the shares he could acquire within sixty days by converting the note discussed in Section 16. Both are accurate on their own basis.

The control is purely economic. There is no dual-class structure and no super-voting stock: one share, one vote, and no cumulative voting. His influence tracks his exposure exactly, which is a materially different position from a founder controlling a company through a supervoting class while holding a fraction of the economics.

What the structure permits, and what the company does anyway#

Nasdaq rules let a controlled company opt out of certain governance requirements, and ImmunityBio formally claims two: that a majority of the board be independent, and that it maintain an independent nominating committee.⁷² It then satisfies both voluntarily. Six of nine directors are independent, a governance committee exists, and the audit committee is fully independent, which the exemptions never covered in any case.

Two features are worth naming rather than characterizing. The founder chairs the nominating committee, so the controlling shareholder participates in selecting the directors who oversee him. And the compensation is substantial: for fiscal 2025, roughly $12.7m to the chief executive and $12.3m to the executive chairman, mostly in options and stock, against a net loss of $351.4m.⁷³ The most recent say-on-pay vote passed with about 99% support, which a 62% holder largely determines.

The related-party web#

The convertible note is the largest related-party arrangement but not the only one. The company licenses laboratory and manufacturing space from a NantWorks entity, leases further property from Soon-Shiong-controlled entities, operates under a shared-services agreement with NantWorks, and pays a clinical trial operator owned by a company officer.⁷⁴ Each is individually modest against a $351m loss. Collectively they mean a meaningful share of the cost base is paid to entities the controlling shareholder also controls. A dedicated related-party committee reviews these under a written policy.

The company itself came out of the same web. What trades as ImmunityBio is the former NantKwest, a public natural-killer-cell company, which on March 9, 2021 completed a stock-for-stock merger with NantCell, a private company then named ImmunityBio; the combined entity took the ImmunityBio name and the IBRX ticker. Both sides were Soon-Shiong-controlled, and he held approximately 81.8% of the shares outstanding immediately afterward, since diluted to the figures above. The bioreactor and royalty income booked as other revenue is the legacy of the NantKwest side of that merger.¹⁸⁵

Where the promotion went wrong#

In March 2026 the FDA sent a warning letter over promotional claims, citing a television advertisement and a podcast in which the founder said ANKTIVA could prevent cancer following radiation exposure. The agency responded that it was not aware of data supporting claims that the drug can cure cancer.⁷⁶ Two untitled letters had preceded it, in September 2025 and January 2026. Shares fell about 21% when the letter was published, and a securities class action followed in the Central District of California covering the period from the podcast to the disclosure.⁷⁷

The episode is small in dollars and large in signal. It says the gap between what the evidence supports and what the founder says in public is wide enough for a regulator to act on, and that is the same gap this report navigates in every section. It is also a governance matter rather than a marketing one, because the speaker controls the company.

Who owns the float, and what changed#

The controlling stake makes every ownership percentage ambiguous unless the denominator is stated. Against 1,059,836,273 shares outstanding at July 31, 2026,¹⁶² institutions held 177.9m at the end of March — under a fifth of the company. Against the roughly 405m shares not held by Soon-Shiong and his entities, the same holding is close to half the tradable float. Both figures appear in circulation and they describe the same position.

What the filings show is three different kinds of money moving in different directions. Index managers buy because the stock is in an index, not because of anything in this report. Market makers and statistical-arbitrage desks carry inventory that turns over on flow. What is left — the discretionary money that chose the position — is the only part that reflects a view, and it is the smallest of the three.

Between those two quarters 148 managers opened a position and 39 closed one.¹⁶³ The total nonetheless fell, because the desks left faster than anyone arrived — Jane Street went from 14.2m shares to 704,232 and Goldman from 12.3m to 3.2m, while D. E. Shaw came in at 12.6m. That is inventory rotating, not a view changing. The line that carries information is the discretionary one, and it grew 13%.

The June quarter separates the two more sharply. Carrying the sixty largest holders forward into their June 30, 2026 filings, which covers 91% of the March institutional total,¹⁶⁴ index holdings rose 19% while discretionary holdings fell 13%. BlackRock added 7.7m and State Street 3.7m, which is the June index reconstitution rather than an opinion. Woodline Partners, a healthcare fund and the largest discretionary holder in March, cut from 3,447,346 shares to 2,500,189. Two Sigma went from 4.4m to 11.1m and Millennium from 1.4m to 5.6m, which is again inventory. The coverage cuts one way — a manager opening a new position in the June quarter is not in the sixty — so the discretionary fall is an upper bound and the index rise a lower bound.

Short interest has been falling from its spring peak while the days needed to cover it have roughly tripled. Nasdaq's semi-monthly series puts short interest at 140,326,066 shares on April 15, 2026, 131,830,884 at June 30 and 128,997,049 at July 31.¹⁶⁵ Against shares outstanding that last figure is 12.2%; against the float it is close to a third. Days to cover moved from 4.7 at the end of March to 14.5 at the end of July, because average daily volume fell by more than the short position did. A crowded short in a thin float is the condition for the January move described above, and it has not gone away.

The counterweight#

Against all of it sits a track record, and it is one company rather than two. American Pharmaceutical Partners bought his privately held American BioScience in 2006 and took the Abraxis BioScience name. In November 2007 that company separated into two independently traded halves. The hospital injectables business became APP Pharmaceuticals and took on $1.0bn of term debt, roughly $975m of which was contributed in cash to the proprietary half that kept Abraxane and the Abraxis name.¹⁸⁴ He sold the injectables half to Fresenius ten months later for roughly $3.7bn before contingent payments, and the Abraxane half to Celgene twenty-three months after that for about $2.9bn plus milestones.⁷⁸ The sequence is the point: he separated the business that generated cash from the asset he believed in, sold the first to fund the second, and sold the second only when a buyer would pay for what it might become. He has been a net acquirer of ImmunityBio stock, converting affiliated debt into equity rather than selling. That behavior is also why this report carries no takeout value in any scenario. A controlling holder who funds the company by converting his own debt is not positioning it to be sold, and an acquirer would have to buy him rather than the float.

Concentrated ownership is not by itself a defect. It funds a company that would otherwise have faced worse terms, and it aligns the largest holder with the smallest. What it removes is the ordinary check: no external party can replace this board, and the going-concern conclusion rests on one person’s continuing willingness. The structure carries a price either way, and this model does not assume it away.

The report cites company guidance in several places, so how that guidance has held is part of the evidence. The commercial record is good. Saudi distribution was promised within sixty days in February 2026 and delivered in April.¹³⁰ QUILT-2.005 completed enrollment in February, on the schedule given. The Q4 2026 filing date for the BCG-naive submission has not moved since it was first stated. Revenue has grown sequentially for eight consecutive quarters.

The regulatory record is worse, and it is the one the catalysts depend on. The papillary application was refused filing after the company said the agency had encouraged it. A registrational path for lymphopenia was described as a BLA the company intended to submit after RMAT designation in February 2025;¹³¹ no such application has been filed, and the framing has since become a randomized trial in design. In March 2026 the FDA issued a warning letter finding that the founder's public statements about ANKTIVA were false or misleading,¹³² having sent two earlier untitled letters on the same subject. What emerges is a company that executes what it controls and describes the FDA more favorably than events support. Guidance about shipping product deserves more weight here than guidance about what a regulator is going to do.

Which raises the question the filings do not address. The founder is in his seventies. He is the controlling shareholder, the affiliated lender, the scientific architect and the stated reason the going-concern doubt is alleviated. No succession plan is disclosed, and no committed facility exists that would survive his departure. The structure concentrates an unusual number of dependencies in one person. This report can document every one of them and price none of them.

What a bid would have to look like#

He has sold the business twice over, in the two halves described above, and the record of both sales — what the buyers paid, and how the contingent pieces resolved — is the only evidence of the price at which he transacts. What follows reads that record as a calibration. It changes nothing in the stance of this report, which carries no takeout value in any scenario; it explains, in dollars, why that stance is the right one.

Fresenius paid $23.00 a share in cash for APP Pharmaceuticals in 2008 — $3.7bn for the equity, with $940m of debt assumed — plus a contingent value right, a tradeable instrument paying up to a further $6.00 a share if the business hit earnings targets over 2008–2010. Against 2007 revenue of $647m, the cash-and-debt price was roughly seven times trailing revenue. The targets were missed. Cumulative adjusted EBITDA reached $1.126bn against a $1.268bn threshold, the rights expired without paying anything, and the instrument was delisted in March 2011.¹⁷⁶

Celgene’s 2010 purchase of Abraxis BioScience ran the same shape at a higher pitch: $58.00 in cash plus 0.2617 Celgene shares plus one tradeable contingent value right per share — roughly $2.9bn upfront net of cash, about eight times Abraxis’s 2009 revenue of $359m, around $3.2bn in aggregate. The right carried a $250m milestone for a US lung-cancer approval with a progression-free-survival claim on the label, $300m for a US pancreatic approval with an overall-survival claim, $100m more if the pancreatic approval landed before April 1, 2013, and royalties on any year in which Abraxane sales passed $1bn.¹⁷⁷

The vision came true and the rights still paid almost nothing. The FDA approved Abraxane for lung cancer in the US in October 2012 without the progression-free-survival claim, so the $250m lapsed. The pancreatic approval came on September 6, 2013, on overall survival, five months past the bonus date: $300m paid, $100m forfeited. Sales crossed $1bn exactly once during the royalty term, in 2018, generating a payment of roughly $1.6m, and by then Celgene carried the whole instrument at $19m against a $650m milestone headline. Bristol Myers Squibb’s final notice, in February 2026, reported $368m of Abraxane sales for 2025 and nothing due.¹⁷⁸ Celgene’s stated prize at signing was about $1bn of revenue by 2015; Abraxane peaked at $1.06bn, three years late, so the aggregate price came to roughly three times the realized annual peak. That is the seller’s signature: seven to eight times trailing revenue upfront, roughly three times the credible annual peak in aggregate, and the dream itself taken in contingent form — which resolved to roughly nothing, twice, while the dream itself, in Abraxane’s case, materialized on schedule. A founder who has twice watched contingent value evaporate as the underlying vision paid off is not a seller who takes the vision in contingent form a third time.

Applied to ImmunityBio, the revenue multiple lands below the market price. The model’s next four quarters sum to $296.5m; at seven to eight times forward revenue — the multiple his own companies fetched — the enterprise is worth roughly $2.1–2.4bn, and after net debt the equity comes to roughly a dollar a share against a market price near $7.79 and a convertible that converts at $5.427. The market is not pricing the bladder business; it is pricing the platform. A bid built on the business the drug has today would be a discount rather than a premium, and there is no version of this company’s owner who sells at a discount to his own conversion price.

The calibration that could clear the market price is the one Celgene actually used: roughly three times the credible annual peak of what the buyer believes the asset becomes. Stack the franchises at the sizes this report defends — the bladder business, a second-line NSCLC peak in the low single-digit billions, the persistent-deficit lymphopenia column of $2.0–4.5bn from Table 12, and a sepsis line its own precedent caps — and the credible aggregate annual peak runs $4–8bn, three times which, after debt, is roughly $10 to $21 a share. The sepsis line carries its own cap. Xigris, the only drug ever approved in the US specifically for sepsis, sold roughly $100–200m a year worldwide at its best, and was withdrawn in October 2011 when its confirmatory trial failed. A $35,800 dose inside a DRG bundle faces the same wall. So that line is worth $0.5–1bn a year at most in this stack, rather than what the 48.9m annual cases imply.¹⁷⁹

What that stack still leaves out is the column a checkpoint-inhibitor owner would price. Roughly 250,000–400,000 US patients start a checkpoint inhibitor each year — an estimate, because no authoritative count exists — and somewhere between a quarter and two-fifths of them carry or develop lymphopenia along the way. Low counts travel with checkpoint failure across the largest cohorts: overall survival of 9.8 against 18.3 months for patients lymphopenic at three months in one series, an independent hazard ratio of 1.68 in kidney cancer, and in stage III lung cancer a measurable attenuation of the checkpoint consolidation’s own survival benefit when radiation depletes lymphocytes first.¹⁸⁰ At $100,000 a course, the slice of that population not already counted in Table 12 is worth perhaps $3–8bn a year. That holds only if correcting the count corrects the outcome. No prospective trial has shown it, and it is the proposition bempegaldesleukin carried into 783 randomized patients and lost.⁵⁴ Weighted the way this report weights everything else, at 15–35%, the column adds a dollar to eight dollars of floor. Taken at face value — which is how its owner takes it — it carries the ask into the thirties and low forties a share, just under this report’s own bull case of $46.39. The founder’s reservation price is not a premium to the market. It is his bull case, and his conduct — converting his own debt to equity at $5.427 rather than selling a share — is what believing it looks like.

The path between today’s price and that ask has a known first step, and it is the step that would move the stock most. A partnership or equity investment from a large checkpoint franchise re-rates a company of this size before any bid exists. Strategic-stake announcements have moved small-cap partners by double digits in a day, and Summit Therapeutics tripled in a week — though on head-to-head data against Keytruda rather than on a signature.¹⁸¹ A 20–50% announcement re-rate followed by the 50–65% median control premium builds an executable path to roughly $14–19 a share. The gap between that executable number and the reservation price is the negotiating range, which is another way of saying the negotiation fails — and the deeper problem is the date. That arithmetic prices today’s information against a $50.7m quarter, and nobody bids on this company at a $50.7m quarter. Process does not close it. Under Delaware law a transaction the controlling holder leads faces entire-fairness review unless conditioned from the outset on both an independent special committee and a majority-of-the-minority vote — the MFW framework, reaffirmed in the Match Group decision of April 2024.¹⁸² This company has already run that process once. The 2021 merger that created it was negotiated for the public side by a special committee with its own counsel and conditioned on approval by a majority of NantKwest’s unaffiliated shareholders, which it received.¹⁸⁵ So the mechanics of a controller-led deal here are not hypothetical, and they are also not the constraint: those protections govern how the minority is treated, not what his number is.

Merck is the buyer this frame keeps producing, and the frame should carry its weaknesses. Keytruda and its subcutaneous successor Keytruda Qlex earned $31.7bn in 2025, roughly half of Merck’s revenue, with US price negotiation beginning in January 2028 and the compound patent expiring that December; management’s stated answer is a claimed $70bn of pipeline opportunity, not an acquisition.¹⁸³ Its bladder position needs nothing from ImmunityBio — it is the sole US supplier of TICE BCG, it runs KEYNOTE-676, the pembrolizumab-plus-BCG Phase 3 in this disease, and Keytruda Qlex was approved in the US in September 2025 across pembrolizumab’s indications, the second-line bladder setting included. A Merck bid would be a bet on the platform: on IL-15 restoring the patients its own franchise loses. No Merck statement on IL-15, NK cells or immune restoration exists in the public record. And one vertical fact would follow the deal into any review: the acquirer of ANKTIVA would also own the rationed input its US label requires.

Neither prior company was sold before it had revenue, which makes the precedents apply more closely than they might. APP was sold profitably, on $647m of revenue and $253m of EBITDA. Abraxis was sold on $359m of 2009 revenue and a $104m net loss for that year — revenue-generating and unprofitable, which is the structural position ImmunityBio occupies today. What Celgene bought was a marketed drug with one approved indication and a thesis about three more.¹⁸⁶

The wait is the instructive part. Revenue arrived on day one, because Abraxane was already selling; the thesis took considerably longer. Celgene told investors the deal would be modestly dilutive in 2011, accretive from 2012, and would add about $1bn of revenue in 2015. The lung approval came in October 2012, two years after closing, and pancreatic in September 2013, just under three. Sales moved as the indications landed — $427m in 2012, $649m in 2013, $848m in 2014 — and the $1bn promised for 2015 arrived in 2018, three years late and eight years after the close.¹⁸⁶ An acquirer of this platform would face the same shape on a similar clock: the bladder business bills from day one, ResQ201A reads out around 2028 with approval in 2029 or 2030, and a lymphopenia indication sits behind that. Three to five years to the first thesis revenue, longer to any peak.

For Merck that clock is the argument rather than an objection to it. US price negotiation on Keytruda begins in January 2028 and the compound patent expires that December, so an asset bought now would deliver its first thesis revenue into precisely the years the franchise starts losing it. Size it against what is being defended. Against that $31.7bn franchise, the immune-restoration lines this report sizes — the persistent-deficit column at $2.0–4.5bn and the checkpoint-adjunct slice beyond it — would together be worth something in the high single-digit billions a year if they work, against a purchase price in the low twenties of billions at his reservation. That is a meaningful fraction of the erosion and nowhere near all of it. Weighted at the 15 to 35% this report applies elsewhere, the expected contribution falls to perhaps a fifth of that, which is the honest way to state it: buying ImmunityBio would be a large bet on one mechanism, not a replacement for Keytruda. Merck’s own answer to 2028 is a claimed $70bn of pipeline opportunity by the mid-2030s, and a company that believes it has that does not need this.

So model the offer at the date it would actually arrive, which is after the revenue. Take the bladder franchise at roughly a quarter of the eligible US NMIBC pools — a share the model’s own penetration tracks place between the base and bull year-three paths, so around 2030 or 2031. A quarter of the 31,000 eligible BCG-naive patients is roughly 7,800 a year, worth about $1.2bn at the $150,000 the model recognizes per launch-year patient; a quarter of the twelve-to-sixteen-thousand BCG-unresponsive pool adds roughly half a billion at current per-patient revenue; papillary adds its line. Call it a US bladder franchise around $1.7bn a year. At the seven to eight times forward revenue his own companies fetched, the bladder business alone supports an enterprise value near $12–14bn — against the $2.1–2.4bn the same multiple produces today.

What changes more than the base is the optionality, because the clock that produces the revenue also produces the evidence.

Said as one sentence: the spread between the executable price and the reservation price exists because buyer and founder weight the same franchises at fifteen to thirty-five percent against roughly a hundred, and evidence is the only thing that closes it — so revenue scale is the precondition for a deal not because the multiple needs a bigger base, but because the proof that makes a buyer pay conviction prices arrives on the same clock as the revenue. For Merck the two clocks meet neatly: the 25%-share world and the post-readout world both land as Keytruda’s US exclusivity ends.

None of this adds a line to the sum-of-the-parts, and it is not meant to. It establishes three prices for one company: what the business supports, which is about a dollar; what a negotiated process could execute, which is the teens; and what the owner’s own arithmetic demands, which is his bull case. The distance between the second and the third is why the takeout is carried at zero. What would collapse that distance is not a bid but evidence: the sepsis readout, a lymphopenia endpoint the FDA accepts, ResQ201A. The catalysts that would move the stock and the ones that would move his reservation price are the same list. That is the deepest sense in which this remains a founder-dependency story.

15Catalysts and timing

Two of the entries below decide whether the base case happens at all: the BCG-naive filing at the end of this year and the papillary decision six weeks later. The rest set the pace or the ceiling. Dates of three kinds appear together, and they fail differently — company guidance slips, regulatory action dates hold but can be extended, and registry completion dates move as events accrue.

One row above carries a history the date alone does not convey. The papillary sBLA reaching a January 2027 decision is the same application the FDA refused to file on May 2, 2025.¹²⁷ One regulator has since accepted that package: the UAE authorization of July 2026 covers papillary-only disease without CIS.¹⁶⁶ No other approval anywhere does, and it is the same indication the January decision would grant in the US. That is a data point rather than a precedent. Gulf regulators lean on FDA and EMA review, so an approval granted where the FDA has not yet ruled carries less weight than an independent one would. ImmunityBio had submitted it that March, and said afterwards that leadership from CBER, CDER and the Oncology Center of Excellence had unanimously encouraged the submission at a meeting in January.¹²⁷ That account is the company's own and the FDA has never commented on it, which is the agency's practice with individual sponsor meetings. What the agency wanted, per the company's account after the meeting that followed, was a randomized trial against chemotherapy, and in August 2025 ImmunityBio committed to running one.¹²⁸ No such trial was started. The application was resubmitted in March 2026 on extrapolation from the carcinoma in situ cohort instead, and accepted in May.¹²⁹

Two readings survive that sequence and this report does not choose between them. The agency may have moved, which would make the resubmission a fair reading of a shifted position. Or the company may have described a meeting more favorably than the record supported, submitted on that basis, and later found a filing route around the trial it had been asked to run. Either way the file has been refused once on evidence the agency called insufficient, and the acceptance letter reserved the extrapolation question rather than settling it. A PDUFA date is a deadline for a decision and not a signal about which way it goes.

What would break this. The two entries this section calls decisive are only decisive if the naive setting stays open. A positive SunRISe-3 or a durvalumab launch that takes the first-line slot before ANKTIVA files would make the January decision a second-line event rather than a gateway, and the dates below would matter less than the order they arrive in. The reverse also holds: an approval in January on extrapolation alone would establish a filing route this report treats as unavailable, and several later rows would move forward with it.

16Valuation — a sum-of-the-parts

A single revenue multiple cannot value this company, and forcing one produces a systematically low answer. The reason is visible in a comparable: CG Oncology carries roughly $4.5–5bn of enterprise value on essentially no revenue, entirely for cretostimogene's pre-commercial NMIBC optionality. If the market pays that for one unapproved asset in this indication, then running all of ImmunityBio's value through a multiple on approved-indication revenue assigns near-zero worth to BCG-naive, NSCLC, lymphopenia and sepsis. That is not a conservative assumption; it is an incorrect one.

What follows separates the two. The commercial franchise — approved indications, generating revenue — takes a revenue multiple. Unapproved programs are valued separately against the CGON benchmark and folded into the commercial line only when approved.

Where the commercial multiple belongs#

The historical record is unambiguous about where multiples sit during the scaling phase, and it is far above where a mechanical discount would put them.

At $200m of revenue the range is 40–72x. At $500m it is 12–30x. At $1bn, 10–23x. Mature franchises with durable economics and high gross margins settle at 9–15x, not the 4.6x of the single weakest observation. Vertex is the better mature analog for a 99%-gross-margin franchise with patent protection to 2036.

The commercial multiples used below sit in the middle of that historical range: 20x in 2026 falling to 9x at maturity in the base case. IBRX-specific discounts — going-concern language, 62–66% insider control³⁹¹⁷, the Oberland revenue interest, the FDA warning letter — are reflected in that positioning, applied once rather than compounded.

Pipeline optionality, benchmarked to CGON#

The cross-check behind that caption is a bottom-up build on the BCG-naive line, run in the spreadsheet with its inputs exposed: the ~31,000-patient pool, a 15–25% mature share, course economics of roughly $150,000 per patient-year, an EV-to-peak-sales multiple of 2.5–3.5x, and a 35–45% probability of approval on the full QUILT-2.005 readout. That machinery brackets the bear at roughly $600m and reaches about $1,800m at the top of every input — below the $2,200m the CGON benchmark supplies. The base therefore implies a 30–40% mature BCG-naive share on these economics, and the input carrying the difference is share, not price or probability. The benchmark drives the result above roughly the midpoint, which is worth knowing before relying on it.

The balance sheet, and the financing it forces#

Cash and marketable securities were $357.4m at 30 June against a derived quarter-standalone operating burn of $66.5m³⁵ — roughly five quarters standalone. On the base case, cumulative burn to breakeven is about $224m, and cash troughs near $113m in Q4 2027 with the Dunkirk capital commitment included. Tight, but not forced. Non-operating capital is ample: approximately $132m of in-the-money warrants (13.5m at $3.101, 27.96m at $3.240) and $349m of remaining ATM capacity.

The complications are structural rather than immediate. Total liabilities are $1.67bn against a stockholders' deficit of $1.05bn. The going-concern footnote persists, mitigated explicitly by the Founder's stated intent to fund operations — a dependency with no analog among peers. The Oberland revenue interest¹⁶ skims 5.625–12.50% of ex-China net sales, but the drag is bounded, and the bound changes which scenario it punishes. Payments stop permanently once Oberland has received 195% of what it funded — roughly $731m against $375m drawn. If cumulative payments by the December 31, 2029 test date have covered the $375m, the rate steps down to a flat 2.8125%; if they fall short, the rate increases and a lump-sum true-up is owed. Fast growth therefore retires the obligation early at the capped amount, and slow growth is what the structure punishes. On the model’s revenue paths the test date arrives short in every scenario, so the step-up is the expected outcome — the difference is that the base and bull then hit the $731m ceiling by 2031–2032 and stop paying, while the bear pays every would-be dollar and owes the true-up besides. And the $480m related-party convertible matures December 31, 2027, one quarter before the cash trough. That is the largest dilution event in the forecast window. The balance sheet carries the note at fair value — $774.4m at June 30 against $480m of principal — because the company elected fair-value treatment and the conversion right is deep in the money.¹ The gap between the two is the market value of the shares Nant would receive on conversion, and swings in it flow through net loss as a non-cash charge — one more reason headline net loss overstates cash burn. What that founder has previously built ahead of demand, and how those bets resolved, is set out in Section 10.

The convertible deserves more attention than its balance-sheet line suggests, because its behavior inverts with the share price. It converts at $5.427, and only at the holder's option. Above that price it behaves as equity: Nant converts, roughly 88m shares are issued, and $480m of obligation simply disappears from the balance sheet. Below it, Nant has no reason to convert — and the company cannot repay $480m out of roughly $113m of cash. The outcome is then a negotiated restructuring with an affiliated holder who owns three-fifths of the equity, almost certainly at a conversion price struck near the prevailing market. At $3.50 that is 137m shares rather than 88m. A lower stock price does not remove the dilution; it returns larger — and it arrives in the same quarter the warrants, struck at $3.101 and $3.240, would expire worthless. That interaction is modeled explicitly in Section 16.

Those constraints set the denominator.

Two mechanics govern the denominator, and both are frequently mishandled. The first is that dilution and its proceeds are a single transaction. The 41.5m in-the-money warrants appear in every fully diluted count, but they also deliver roughly $132.5m of cash on exercise. The $480m convertible produces 88.4m shares, but extinguishes $480m of debt in the same motion. Counting one side without the other understates equity value twice over.

The second is that ATM issuance is a function of need and price, not a fixed schedule. A company generating cash does not draw it; a company burning cash draws it at whatever price the market offers — which is lowest precisely when the need is greatest.

The composition tells you more than the total does. In the bear case the warrants deliver no shares and no cash, while the convertible delivers 48.7m additional shares because it must be restructured at a distressed price. Dilution does not disappear when the stock falls — it changes form and gets worse.

The asymmetry is the point. Base and bull differ by only 11m shares, because once operating cash flow turns positive the ATM is a buffer rather than a lifeline — and at $9 a share, $100m costs almost nothing in dilution. The bear case issues 183m shares for $550m because it must raise at $3, and must raise more besides because the warrants deliver nothing. Between bear and bull the share count differs by 375m, or 29%, and essentially all of that gap is created by whether the company needs the money. The unhelpful part is that needing it is also when the cheapest sources of it stop working.

Share growth beyond the ATM is equity compensation at roughly 1.5% a year on the outstanding base — 18m shares per two-year step. No buyback is assumed in any scenario.

The sum-of-the-parts#

What it means for the price today#

On base assumptions the 2026 sum-of-the-parts is $6.74 against a market price of $7.32 at publication — so the market is paying roughly a 9% premium to a mid-range valuation of today's business plus a risk-adjusted pipeline. That is a premium, not a bubble, and it is smaller than the dispersion in the inputs. Push the commercial multiple to 23x — still inside the historical range at this revenue level — and the 2026 figure meets the market price exactly.

The forward picture is the one to weigh. On base assumptions the model reaches $21.30 by 2032, about 2.9 times the $7.32 publication price, or roughly 19% a year compounded. The bull case is $46.39 and the bear is $3.07 — a spread of fifteen to one, where the bear case loses well over half the money. So the middle outcome earns a high-teens annual return, with a much larger payoff if the platform works and a severe loss if it does not. Valuing the company on a single multiple of today’s business misses all of that.

17Risks

• Medicare Advantage friction. More than half of Medicare beneficiaries are now in Medicare Advantage, and the Congressional Budget Office projects 64% by 2034.¹¹⁴ These plans generally honor average sales price plus 6%, but they add prior authorization that fee-for-service does not. Roughly half of this drug’s Medicare volume is therefore exposed to plan-level review, and that share grows every year. The friction is procedural rather than structural, but it slows starts and it is not in the model.
• Positioning, in both directions. Short interest has run between roughly 14% of shares outstanding and, on some float-based measures, above a third — well above the peer group.⁸⁵ That is a statement about volatility rather than about value. It amplifies bad news and it reverses violently, as it did in January 2026. Position sizing should assume the price can move a long way on news that changes the model very little.
• Regulatory conduct. The FDA issued a warning letter on March 13, 2026⁴ — its first full warning letter of the year from OPDP, the FDA office that polices drug promotion — following two untitled letters to subsidiary Altor. It cited a television advertisement that never aired and a podcast appearance. The company implemented new promotional controls in April. The distinction the company would draw is fair: FDA adjudicates what is provable for promotion, not whether a claim might eventually prove true, and aspirational commentary from a founder is not the same as a marketing claim. But one cited item was a route-of-administration error, and with two applications under review the compliance bar is now materially higher.
• Litigation. Douglas v. ImmunityBio²¹ (C.D. Cal.), class period 19 January to March 24, 2026. Lead plaintiff appointed July 15, 2026. Consolidated derivative suits stayed pending the motion-to-dismiss phase.
• Competitive erosion. INLEXZO is ramping faster than ANKTIVA did from an equivalent launch point, aided by BCG independence and J&J's installed infrastructure. Cretostimogene's profile — 75.5% CR, median DOR ≥27.9 months⁷, zero grade ≥3 treatment-related events, BCG-free — is the most credible medium-term threat in CIS.
• The platform thesis fails in a randomized trial. Roughly a third of the pipeline value in this model sits in NSCLC and lymphopenia, and both rest on the proposition that an IL-15 superagonist restores checkpoint-inhibitor benefit. Nektar's bempegaldesleukin carried the same logic on the IL-2 pathway and missed every endpoint in a 783-patient randomized trial against nivolumab alone.⁵⁴ Bempegaldesleukin never reached approval anywhere; ANKTIVA is approved and selling. What the two share is that ANKTIVA’s systemic evidence is also single-arm, and ResQ201A is the first randomized test of it. A miss there does more than trim these lines. It ends the argument that this is a platform rather than a bladder-cancer company, and takes the commercial multiple with it.
• Externally imposed price cuts. The model holds US net price constant, and Section 02 sets out why a voluntary cut would be self-defeating. A cut imposed from outside is a different matter. Executive Order 14297 directs HHS toward most-favored-nation pricing benchmarked to the lowest price in a qualifying OECD country, and CMS has proposed the GLOBE model for Part B with an effective date of October 1, 2026.⁶⁰ Neither is finalized. But a drug priced at $35,800 a dose in the US, sold into Gulf tender markets that themselves reference-price against an OECD basket, sits directly in the path of that mechanism. A 20% net-price reduction takes roughly $530m off the 2032 base commercial line before any volume response.
• Governance. Roughly 62–66% insider control, a related-party convertible held by an affiliate, CVRs partly held by the Founder, and a going-concern mitigation resting on his continued support. Directors have sold consistently under 10b5-1 plans and no open-market insider buying has been identified since the April 2024 approval, though a purchase would have been legally fraught from March 2026 once the securities class action was filed. The founder's own transactions are share withholding on vesting rather than sales.
• Payer risk in the naive setting. Layering a ~$380k net regimen onto near-free generic BCG across tens of thousands of first-line patients invites step therapy and net-price concession. Bull models hold per-patient net price flat at scale; that is unlikely to survive a pharmacy benefit committee.

18Conclusion

ANKTIVA is approved in the United States, selling, and supported by three-year durability data, in a niche where it has reached only 10–15% penetration. The operating loss is closing, cash flow turns positive in Q1 2028 on the base case, and the financing the company still does is opportunistic rather than forced. Behind the commercial franchise sits an unusual configuration: a recognized serious condition with no approved therapy, a biomarker measured on every routine blood count, and an FDA designation that accepts the premise.

Section 16 prices today's business plus a risk-adjusted pipeline at $6.74 a share against the $7.32 the market paid at publication — a gap that one turn of the commercial multiple closes, and smaller than the noise in its own inputs. From there the base case compounds to $21.30 by 2032, a high-teens annual return for the middle outcome.

That is the case for ownership. The case for care is the shape of the distribution rather than its center. The bull path reaches $46.39 and the bear $3.07 — a fifteen-fold spread across six years, driven by a small number of binary events. QUILT-2.005's full readout rests on a 43-patient interim in which the six-month comparison did not clear significance. The convertible matures one quarter before the cash trough. Three-fifths of the equity sits with one holder whose promotional conduct has already drawn a warning letter.

None of that argues against owning it. It argues that position size should be set against the bear case rather than the base — because a $3.07 outcome is not a drawdown to be waited out, and the events that would produce it are knowable in advance. The catalysts in Section 15 are, in effect, the schedule on which this thesis will be settled.

On the platform question: the ambition is large and the science is not fanciful, but it is unproven where it counts. Sepsis in August 2027 and ResQ201A around 2029 are the two readouts that convert optionality into franchise. Until then ImmunityBio is a bladder-cancer company with credible options attached, and on this construction it is priced at roughly the value of that business with the options only partly counted.

19Glossary

This report uses the vocabulary of three separate fields — urologic oncology, drug regulation and pharmacy reimbursement — and assumes fluency in all of them. That assumption is unreasonable. Terms are grouped by where they appear rather than alphabetically, and each entry says why it matters here, not merely what it stands for.

Disease and treatment#

NMIBC

Non-muscle-invasive bladder cancer. The tumor sits in the bladder lining and has not grown into the muscle wall. Usually survivable, frequently recurrent, and managed for years — which is why durability of response matters more here than in advanced cancer.

CIS

Carcinoma in situ. Flat, high-grade disease spread across the bladder lining rather than a discrete lump. Harder to remove surgically, more likely to progress, and the population ANKTIVA is currently approved to treat.

papillary (Ta / T1)

Tumors that grow outward on a stalk rather than lying flat. Ta is confined to the lining, T1 has grown into the tissue just beneath it. Both can be high-grade. ANKTIVA's pending January 2027 decision covers papillary-only disease without CIS.

High-Grade (HG)

Cells that look aggressive under a microscope. High-grade disease drives the treatment decisions in this report; low-grade disease is generally watched rather than treated aggressively.

TURBT

Transurethral resection of bladder tumor. Scraping the tumor out through the urethra. The starting point for nearly every patient here, after which drug therapy aims to stop recurrence.

radical cystectomy

Removing the entire bladder. The fallback when drug therapy fails. It carries roughly 5–11% mortality within 90 days and permanently changes how a patient urinates, which is why avoiding it is the outcome patients and urologists care about most.

intravesical

Delivered directly into the bladder through a catheter. The drug acts locally and barely enters the bloodstream. ANKTIVA is intravesical in bladder cancer — central to this report, because it means a urologist administers it in their own office.

subcutaneous

Injected under the skin, so the drug circulates through the whole body. ANKTIVA's lung-cancer and lymphopenia programs use this route, which is a different drug behavior from the bladder use despite being the same molecule.

Drugs and mechanisms#

BCG

Bacillus Calmette-Guérin. A live, weakened tuberculosis bacterium, originally a TB vaccine, dripped into the bladder where it provokes an immune response that attacks tumor cells. Standard care for decades, cheap, and made by a single US supplier — which is why its supply governs so much of this analysis.

TICE BCG

The specific BCG strain Merck manufactures, and the only one approved in the United States. Europe has six strains available; the US has one. That difference explains why the shortage is a US problem.

N-803

ANKTIVA under its development code. The same molecule also appears as ALT-803, from Altor BioScience where it was invented, and as nogapendekin alfa inbakicept-pmln, its generic name. Trial registries and papers written before the April 2024 approval use N-803 or ALT-803 throughout, which is why both names recur in the sources.

rBCG

Recombinant BCG. ImmunityBio's genetically modified version, supplied free or at cost under an expanded-access program while the shortage persists. Investigational, so it carries no billing code — a point the report treats as decisive.

BCG-naive

A patient who has not yet received BCG. The larger, first-line population — roughly 31,000 US patients a year — and the setting whose trial readout the entire bull case depends on.

BCG-unresponsive

A patient whose cancer returned despite adequate BCG. A formal FDA-defined status, not a judgement call, and the smaller second-line population ANKTIVA is approved in today.

IL-15 / superagonist

Interleukin-15 is a natural signaling protein that tells immune cells to multiply. A superagonist is an engineered version that does this far more strongly and lasts far longer in the body. ANKTIVA is one.

NK cell

Natural killer cell. An immune cell that kills abnormal cells without needing to be trained on a specific target first. IL-15 expands them, which is the mechanistic basis of the platform argument.

checkpoint inhibitor

A drug that releases a brake tumors use to switch off immune cells — Keytruda and IMFINZI are examples. They only work when functional immune cells are present, which is the basis of the complementarity argument in Section 07.

Treg

Regulatory T cell. An immune cell that suppresses immune responses. Expanding them is counterproductive in cancer, and IL-15 does not. Whether that difference is protective is examined in Section 07.

lymphopenia

An abnormally low count of lymphocytes, the immune cells that fight cancer. Common after chemotherapy and radiation, and it predicts poor response to checkpoint inhibitors. Reversing it is a potential ANKTIVA indication in its own right.

ALC

Absolute lymphocyte count. The blood measurement of the above. ImmunityBio's lung-cancer case rests substantially on raising it — but no regulator has yet accepted that raising it proves clinical benefit.⁹⁹

ADC — antibody-drug conjugate

A tumor-targeting antibody with a chemotherapy payload attached. Much of the second-line NSCLC field is ADCs, and several have missed on overall survival.

Trial endpoints#

CR — Complete Response

No detectable cancer after treatment. The endpoint used in CIS, where there is a visible lesion to clear. ANKTIVA's 62% CR rate is measured this way, against 77 evaluable patients.

DOR — Duration of Response

How long a complete response lasts before cancer returns. More important than the response rate itself: a high CR that lapses after six months is worth much less than a lower one lasting four years.

DFS / EFS

Disease-free and event-free survival. Time until the cancer returns or another defined setback occurs. Used for papillary disease, where there is no lesion to clear and therefore no complete response to measure.

OS — Overall Survival

Time until death from any cause. The endpoint regulators trust most and the one hardest to demonstrate. NCCN declined to recommend IMFINZI in BCG-naive disease specifically because no overall survival benefit was shown.

PFS — Progression-Free Survival

Time from randomization until the tumor grows or the patient dies. The second-line NSCLC comparisons are largely fought on this measure, which can improve without anyone living longer.

TRAE — treatment-related adverse event

A side effect the investigator attributes to the drug, graded 1–5; grade ≥3 means severe enough to interfere with daily life or require intervention. The competitive tables report it because tolerability decides what an office will administer.

DSS — Disease-Specific Survival

The share of patients who have not died of this cancer, ignoring deaths from other causes. In an elderly population it runs well above overall survival, which is why it appears alongside durability rather than in place of it.

ORR — Objective Response Rate

The share of patients whose tumors shrink by a predefined amount.

Hazard Ratio (HR)

A comparison of event rates between two trial arms. Below 1.0 favors the treatment: an HR of 0.68 means roughly a 32% reduction in the rate of the measured event.

single-arm trial

Everyone receives the drug; there is no control group. Faster and smaller, but it cannot prove the drug outperformed an alternative — a caveat that applies to most evidence in this field, ANKTIVA's included.

The trials#

QUILT-2.005

The randomized trial of ANKTIVA plus BCG against BCG alone in BCG-naive carcinoma in situ, 366 patients, fully enrolled February 2026. It is the single most consequential readout in this report: the BCG-naive line is the largest in the model, and this trial decides whether it arrives.

QUILT-3.032

The single-arm trial that won ANKTIVA its April 2024 approval in BCG-unresponsive carcinoma in situ. Cohort A produced the 62% complete response in the label; Cohort B, in papillary-only disease, is the basis of the supplemental application now before the FDA.

QUILT-3.055

The trial of ANKTIVA plus a checkpoint inhibitor in patients who have already progressed on one. Single-arm, so it can show what happened without showing the drug caused it, which is why the randomized ResQ201A matters more.

QUILT-2.023

The one randomized trial of systemic ANKTIVA that has reported. It was terminated at 102 patients against roughly 478 planned, leaving it underpowered for progression-free survival, and its headline result was a blood marker rather than survival. The report treats it as the weakest evidence in the systemic case.

QUILT-3.076

The trial pairing ANKTIVA with M-ceNK, the memory cytokine-enriched natural killer cells that are the company's main cell-therapy asset. Phase 1/2, so it establishes feasibility rather than benefit.

ResQ201A

The randomized Phase 3 of ANKTIVA plus tislelizumab plus docetaxel against docetaxel alone in second-line lung cancer, 507 patients, overall survival as the primary endpoint. Began October 2025. This is where the systemic thesis is settled or is not.

SunRISe-3

Johnson & Johnson's randomized Phase 3 of TAR-200 against BCG in the BCG-naive setting, 1,135 patients. A positive result would put a BCG-free regimen into the population this model depends on.

Regulation#

BLA / sBLA

Biologics License Application, and the supplemental version. The application to sell a biologic drug; the supplement asks to add a new use to an existing approval. ANKTIVA has two supplements pending.

PDUFA date

The deadline by which the FDA must decide on an application. ANKTIVA's papillary decision carries a target date of January 6, 2027.

CRL — Complete Response Letter

The FDA’s formal refusal to approve an application in its current form, with the deficiencies listed. ANKTIVA received one in 2023 over third-party manufacturing inspections before the 2024 approval.

EAP — expanded access program

FDA-sanctioned use of an unapproved product outside a trial for patients without alternatives, with consent and reporting obligations attached. The rBCG program runs through one, which is why prescribing it carries friction an approved product would not.

RMAT

Regenerative Medicine Advanced Therapy designation. An FDA status granting closer agency engagement and potential faster review. ANKTIVA holds one for lymphopenia — a designation, not an approval.

NCCN

National Comprehensive Cancer Network. Publishes the treatment guidelines US oncologists follow and insurers use to decide coverage. A listing here often matters more commercially than the FDA label itself.

AUA / SUO

American Urological Association and Society of Urologic Oncology. The professional bodies whose guidelines and conferences shape urology practice.

Composition of matter

A patent on the molecule itself, the strongest kind: it stops anyone making or selling that substance for any use. A method-of-use patent is narrower, covering only a stated way of using it, and is easier to work around by using the drug differently.

Patent term extension

Extra patent life granted to compensate for time lost while a drug sat in regulatory review. One patent per approval, up to five years, and it cannot run past fourteen years from approval. ImmunityBio has applied; nothing has been granted.

March-in rights

The government’s power to require licensing of an invention it helped fund, if the owner fails to bring it into use or public health demands it. Rarely exercised, and it has never been used to force a price reduction.

CHMP / MAH

The European Medicines Agency committee that recommends approvals, and the Marketing Authorisation Holder — the company legally responsible for a drug in Europe. ImmunityBio Ireland Limited holds ANKTIVA's.

NICE

The body that decides which drugs England's NHS will fund. Its appraisal of ANKTIVA was terminated in June 2026 after no submission, leaving the UK private-pay only.

SFDA / EDE

The Saudi and Emirati drug regulators. Both have approved ANKTIVA, including for lung cancer in Saudi Arabia.

Payment and access#

buy-and-bill

The practice purchases the drug itself, administers it, then bills the insurer. It keeps the margin between what it paid and what it is reimbursed. This is why an office-administered drug can be commercially preferable to one requiring referral elsewhere, and it recurs throughout this report.

J-code

The billing code identifying a specific drug on an insurance claim. Without a permanent one, claims are filed under a miscellaneous code and are frequently delayed or denied. ANKTIVA's (J9028) took effect January 2025 and its revenue roughly doubled in the following quarter.

ASP + 6%

How Medicare pays for office-administered drugs: average selling price plus a 6% margin, reduced to about 4.3% by budget sequestration. Because the margin is a percentage, the practice earns more in absolute dollars on a costlier drug. What a price cut would do to adoption is less obvious, and Section 02 sets out why.

WAC

Wholesale acquisition cost. The list price before discounts. ANKTIVA's is $35,800 per dose.

gross-to-net

The gap between list price and what the manufacturer actually keeps after rebates and discounts — roughly 15–20% here. Reported revenue is already net of it.

Medicaid best price / 340B / AMP

US programs that peg discounts to a manufacturer's lowest price. Cutting a list price resets these floors permanently, which is why manufacturers prefer rebates to headline price cuts.

prior authorization / step therapy

Insurer requirements to approve a drug in advance, or to fail a cheaper option first. The main practical brake on uptake of an expensive add-on.

DRG / NTAP

Hospitals are paid a fixed sum per admission (the DRG) regardless of drugs used, so an expensive drug erodes their margin. NTAP is a temporary top-up for genuinely new technology. This is why a hospital indication such as sepsis could not support the oncology price.

MFN

Most-favored-nation pricing. A US policy proposal to peg domestic prices to the lowest paid in comparable wealthy countries. Relevant because a low price offered abroad could be imported back into the US price.

AMNOG

The German pricing law. A drug launches at whatever price the company sets, then after six months a national body assesses whether it beats existing treatment and a reimbursed price is negotiated from that verdict. The free window is the only period of unconstrained German revenue.

G-BA

The German committee that runs the AMNOG benefit assessment and decides what a drug is worth relative to standard care. Its verdict sets the ceiling for the negotiated price. No assessment of ANKTIVA has started.

accès précoce

France's early-access route. A company may supply and be paid at its own price before the formal appraisal concludes, in exchange for collecting real-world evidence. It is the one European mechanism that produces revenue ahead of a negotiated price.

NUPCO

Saudi Arabia's national procurement company. It runs the tenders through which government hospitals buy drugs, so a NUPCO award is the practical gateway to public-sector volume in the Kingdom.

CCHI

The Saudi insurance regulator. A listing with CCHI or the Ministry of Health is what makes a drug reimbursable rather than merely purchasable.

MOHAP

The UAE health ministry, which sets drug prices nationally. Individual emirates then decide separately whether to add a drug to their hospital formularies.

Finance and valuation#

Enterprise Value (EV)

Market capitalization plus debt, less cash — what it would cost to buy the whole business. Used instead of market cap because it accounts for the balance sheet.

EV/Sales multiple

Enterprise value divided by annual revenue. The shorthand for how richly the market prices a company's sales. This report applies 9x to 2032 base-case revenue.

Sum-of-the-Parts (SOTP)

Valuing each part of a business separately and adding them up, rather than applying one multiple to everything. Used here because a single multiple assigns almost no value to drugs not yet approved.

risk-adjusted value

A program's potential value multiplied by its probability of success. A drug worth $2bn if approved, with a 30% chance, carries $600m.

convertible note

A loan the lender may convert into shares instead of demanding repayment. ImmunityBio's $480m note converts at $5.427 and matures December 2027. Above that price it becomes equity; below it, it becomes a problem — see Section 01.

warrant

A right to buy shares at a fixed price. Worthless if the share price stays below the strike, which is why this model does not credit warrant proceeds in the bear case.

ATM offering

At-the-market. Selling new shares gradually into the open market to raise cash. Flexible, but dilutive — and the worse a company's position, the lower the price at which it must sell.

dilution

Existing shareholders owning a smaller share of the company after new shares are issued. Central to the bear case here, where a falling price forces more shares to raise the same money.

revenue interest (Oberland)

A financing arrangement where an investor receives a share of future sales rather than interest. ImmunityBio pays Oberland a percentage of revenue up to a capped total.

CVR

Contingent value right. A conditional payout triggered by a future milestone, often used in acquisitions.

10-Q / 10-K

Quarterly and annual reports filed with the SEC. The primary source for most financial figures in this report.

20Sources

Every identifier below was checked against SEC EDGAR, FDA.gov, EMA.europa.eu, NICE.org.uk, ClinicalTrials.gov and DOI resolvers on August 5, 2026. Where a source carries a residual limitation it is flagged in amber at the point of citation. Nothing in the analysis rests on an unverified identifier. Footnote numbers mark the first point in the text at which each claim is made.

Residual limitations. Four items remain short of full independent confirmation and are flagged in place. The UAE registration numbers are company-reported via SEC exhibit; no public Emirates Drug Establishment register entry was retrievable. The SWOG S1602 efficacy figures are confirmed through named ASCO Daily News, UroToday and Urology Times coverage of the presentation, because the JCO abstract sits behind ASCO's subscriber wall — the abstract number, author, DOI and conclusions are nonetheless verified. The QUILT-2.023 subgroup figures are taken from trade-press coverage of the ASCO 2026 abstract rather than the abstract itself, and no peer-reviewed randomized efficacy publication exists for that trial. The checkpoint-complementarity mechanism rests on preclinical models; no human immune-monitoring dataset from an ANKTIVA trial confirms it. Pre-2021 comparable-company multiples are interpolated from quarterly market-capitalization snapshots and carry approximately ±10–15% error. Every other identifier below was retrieved from the originating document.