GLP-1: Where the Durable Profits Concentrate

The central insight — stated first, argued for the rest of the document

For decades, the money in pharma came from one thing: owning a patented drug and pricing it freely. GLP-1 is where that stops — between late 2025 and mid-2026 the U.S. started setting these prices itself, so the patent no longer controls the price. What’s left to earn is the cost advantage of the cheapest manufacturer, and the royalty owed to whoever discovered the drug.

1. 1The patent holder no longer names the price in six of the eight channels this report maps — $274 by statute, $245 by demonstration, roughly $88–120 by generic entry in Canada — and in the two channels it still names, it has cut three times in eight months, five years before semaglutide's US compound patent expires. Confidence: High.
2. 2Lilly is advantaged and Novo deteriorating on the only variable an administered price leaves in play — cost. The two gross margins changed places during 2025: Novo was 3.4 points ahead in 2024 and 2.0 points behind in 2025, and over the first half of 2026 Lilly added 0.5 points while its realised price fell 13%, against roughly −1.7 points underlying for Novo. Confidence: High.
3. 3The discovery royalty is the only claim in the chain whose rate no 2026 price event reaches. Chugai bought mid-single-digit to low-teens percent of worldwide orforglipron net sales for $50 million in 2018, booked inside Lilly's cost of sales — above manufacturing, above rebates, above the payer. The rate is insulated; the dollars are not, because the base is net sales. Confidence: Moderate.
4. 4The three layers the market bought as picks and shovels have each lost the mechanism that let them capture anything. Bachem's first-half EBITDA margin fell to 25.4% from 29.1%, Hims & Hers' gross margin from 73% to 65%, and CVS Health Services' operating margin to 3.2% — while the FTC delinked PBM fees from list price and the Bridge routed Medicare obesity volume around Part D entirely. Confidence: High.
5. 5The administered channels are the growing minority of American volume, not the majority. Medicare has been barred by statute from paying for a weight-loss drug since the benefit was written and Medicaid obesity coverage stands at thirteen states, so the argument strengthens as coverage widens rather than depending on its having widened already. Confidence: Moderate.
6. 6Novo won the oral volume race and paid for it in gross margin. More than five million United States prescriptions in thirty weeks, at approximately $169 of realised revenue per fill, against a small-molecule tablet priced at $149 in cash and made in a conventional chemical plant. Confidence: Moderate.
7. 7Nobody is going to get paid for capacity. Novo guides 2026 capital expenditure down to about DKK 55 billion from DKK 60 billion and took a DKK 3 billion charge to cancel contracted capacity in the middle of the largest demand ramp this class has seen; Lilly raised first-half capex 64% to $5,259 million to build a small-molecule chain instead. Confidence: High.

The consensus view of this industry is a growth story with two winners. Revenue compounds toward a hundred billion dollars a year, two companies own the molecules, and everyone downstream rides the wave. That story is not wrong about the revenue. It is wrong about the profit, and it is wrong in a specific and now-measurable way.

In the nine months to August 2026, the US government negotiated a maximum fair price of $274 for a thirty-day supply of semaglutide, effective 1 January 2027 — a 71% discount to the $959 list price Medicare paid in 2024 (CMS, Negotiated Prices for IPAY 2027, fact sheet). It then went further and stood up a separate Medicare demonstration — live since 1 July 2026 and, on its own terms, scheduled to end on 31 December 2027 — at a net price of $245 per month — paid outside the Part D benefit entirely, adjudicated by a CMS-appointed central processor rather than by any Part D plan, with manufacturer coupons expressly barred (CMS, Medicare GLP-1 Bridge: Information for Part D Plans). Novo Nordisk has told investors it will cut the US list price of Wegovy by roughly half, to $675, on 1 January 2027 (Novo Nordisk Form 6-K, 4 August 2026, "Key modelling considerations"). Lilly sells its new oral pill for $149 a month in cash. In Canada, where a maintenance fee lapsed and exclusivity went with it, a generic semaglutide pen retails near $88.

Every one of those numbers is a price someone other than the patent holder set. That is the whole argument in miniature. When a monopolist's price is administered rather than negotiated, patent life stops being the asset. What remains as an asset is the gap between the administered price and the cost of making the dose — and that gap is decided in a chemical plant and a capital budget, not a courtroom.

Sources, left to right: CMS GLP-1 Bridge FAQ; Lilly Form 10-Q for the quarter ended 30 June 2026; Novo Nordisk Form 6-K of 4 August 2026; Lilly 10-Q (Mounjaro + Zepbound as a share of $42,773m H1 revenue, computed); CVS Health Form 10-Q, segment note.

1 · What an incretin actually does — and why the effect is dose-limited by the gut, not the receptor

Eat a meal and the small intestine releases two hormones within minutes. One is glucagon-like peptide-1, universally shortened to GLP-1: a 30-amino-acid peptide that tells the pancreas to release insulin only when glucose is high, tells the stomach to empty more slowly, and tells the hindbrain that you are full. The other is glucose-dependent insulinotropic polypeptide, or GIP. Together they are called the incretins, and they explain why an oral glucose load provokes far more insulin than the same glucose given intravenously.

Native GLP-1 is useless as a drug. An enzyme called DPP-4 chews it up in under two minutes. The entire pharmaceutical history that follows is a history of engineering around that half-life: substitute the amino acid DPP-4 attacks, bolt on a fatty-acid chain so the peptide binds to circulating albumin and hides from the kidney, and a two-minute hormone becomes a once-weekly injection.

The ceiling on this class is set by tolerability, not by biology. Push the dose and weight loss keeps climbing; push it further and patients vomit and stop. Every competitive move in this industry for the last three years has been an attempt to buy more efficacy per unit of gastrointestinal misery — by adding receptors, by adding a second hormone with a different side-effect profile, by titrating more slowly, or by changing the shape of the plasma concentration curve. That constraint is why the amylin class exists, and why a daily pill and a weekly injection of the same molecule are not economically equivalent.

2 · Discovery, and the awkward question of who owns it — the first place the rent hides

The founding molecule of the class came out of a lizard. Exendin-4, the GLP-1 analogue that became exenatide, was isolated from the venom of the Gila monster. Liraglutide and then semaglutide came out of Novo Nordisk's own peptide chemistry, and semaglutide's US compound patent runs to 2032 (Novo Nordisk Form 20-F for 2025, Item 4, "Patents"). Tirzepatide is Lilly's, with a US compound patent to 2036, major European countries to 2037 and Japan to 2040 (Lilly Form 10-K for 2025, Item 1, patent expiry table).

Now the part that gets skipped. Orforglipron — the oral small molecule Lilly launched in April 2026 as Foundayo, the drug most likely to define the next decade of this market — was not discovered by Lilly. It was discovered by Chugai Pharmaceutical of Japan and licensed in 2018. Lilly's own quarterly filing states the terms plainly: "Chugai receives tiered royalty payments on worldwide net sales ranging in percentages from mid single digits to low teens, which we recognize as cost of sales" (Lilly Form 10-Q, quarter ended 30 June 2026, Note 3). Chugai is majority-held within the Roche group, and confirmed on 24 July 2026 that it "commenced recognition of royalty income from the product in the second quarter" (Chugai, 2026 Second Quarter Results).

The pattern repeats. Pfizer's $8.0 billion acquisition of Metsera carried with it a contingent-consideration liability from Metsera's own 2023 purchase of Zihipp: the former Zihipp shareholders retain "low-single digit royalties on net product sales on the MET-097i and MET-233i product candidates" (Pfizer Form 10-Q, quarter ended 28 June 2026, Note 2A). Roche paid Zealand Pharma $1.65 billion up front for petrelintide, with $1.2 billion of development milestones, $2.4 billion of sales milestones, a 50/50 profit share in the US and Europe, and tiered double-digit royalties up to the high teens in the rest of the world.

Four of the most valuable assets in the field are therefore encumbered by a claim that sits above cost of goods, above the payer, above the pharmacy benefit manager, and above every price cut described in this report.

3 · The efficacy ladder — what each generation actually delivers, on comparable endpoints

Weight-loss percentages are quoted loosely in this industry and comparing them carelessly is the fastest way to reach a wrong conclusion. Two conventions matter. The efficacy estimand asks what happens if everyone stays on drug; the treatment-regimen estimand asks what happens in the real world, adherence failures included. The gap between them is often three or four percentage points, and companies quote whichever flatters. The table below holds the convention constant where the disclosure allows.

Two readings fall straight out. First, the injectable ceiling has risen far enough that bariatric surgery is now the comparator, not a competing GLP-1: retatrutide's 12 mg arm put 65.3% of participants below a BMI of 30 at 80 weeks. Second, and more consequential commercially, an oral agent that delivers 12% to 16% is not a substitute for a 25% injectable. It is a different product for a different patient — the one who has never taken an injection, does not have insurance coverage, and is paying cash. Novo's own disclosure confirms exactly that: Wegovy pill uptake is "driven by GLP-1-naïve patient uptake and across all doses in the self-pay channel."

4 · Amylin, the second axis — why every large player bought one

Amylin is a second pancreatic hormone, co-secreted with insulin, that signals satiety through a different brainstem pathway from GLP-1. Its appeal is arithmetic: if the incretin dose ceiling is set by nausea, a second mechanism with a milder gastrointestinal profile lets you stack efficacy without stacking side effects.

The 2026 evidence cuts both ways. Roche and Zealand's petrelintide produced up to 10.7% mean weight loss at 42 weeks with discontinuation for adverse events of 4.8% against 4.9% on placebo, and no vomiting at all in the maximally effective arm (Roche, 5 March 2026). That is a genuinely differentiated tolerability profile. Lilly's eloralintide went further on efficacy, reaching 20.1% at 48 weeks in Phase 2. And Novo's CagriSema, the first amylin-plus-GLP-1 combination to reach a pivotal head-to-head, lost it: 23.0% against tirzepatide's 25.5%, primary endpoint missed.

Roche's structural intent explains why it paid what it paid. The March 2025 collaboration covers petrelintide "as a standalone therapy as well as a fixed-dose combination with Roche's lead incretin asset CT-388," a dual GLP-1/GIP agonist Roche already owned (Roche, 12 March 2025). Roche did not buy an amylin. It bought the missing half of a combination it could then build in-house.

Amylin did not fail here. Combining two peptides in one pen is a harder engineering problem than adding a receptor to one molecule, and Novo took the harder road and arrived second. Novo has since initiated a higher-dose CagriSema trial (2.4/7.2 mg) that will not read out until the first half of 2028. That is a long time to wait for an answer to a question a competitor has already answered.

A third clinical axis now sits beside multi-agonism and amylin: quality of weight loss — fat lost versus lean mass kept. Incretin monotherapy typically sacrifices a meaningful fraction of lost weight as lean tissue. Lilly's bimagrumab (an activin type-II receptor antibody acquired with Versanis) was tested with and without semaglutide in the Phase 2 BELIEVE trial; results published in Nature Medicine in 2026 showed that high-dose bimagrumab plus semaglutide 2.4 mg produced 22.1% mean weight loss at 72 weeks, with lean mass down only 2.9% against a 7.4% lean-mass loss on semaglutide alone, and total fat mass down 45.7% on the combination (Heymsfield et al., Nature Medicine, 2026; NCT05616013). That is a real clinical differentiator. It is not, under an administered-price regime, a restoration of list-price power. A muscle-preserving combination can win share and justify formulary preference on outcomes; the net price it realises is still set by CMS, cash channels and commercial contracting. Who keeps the money still turns on who delivers the milligram cheapest and who owns the upstream royalty, not on who owns the lean-mass story.

5 · The oral problem — two solutions with completely different cost structures

A peptide cannot ordinarily be swallowed. Stomach acid and proteases destroy it, and even intact it is far too large to cross the gut wall. There are exactly two ways around this, and the choice between them determines a company's cost curve for a decade.

The first is to co-formulate the peptide with an absorption enhancer. Novo's answer is SNAC — sodium N-[8-(2-hydroxybenzoyl)amino]caprylate — which locally raises gastric pH and permeabilises the mucosa. It works, and the resulting bioavailability is roughly 1%, against roughly 89% for a subcutaneous injection (peer-reviewed literature reports "the absorption of oral semaglutide remains low (<1%)" — "The Synthesis of SNAC Phenolate Salts and the Effect on Oral Bioavailability of Semaglutide"). The doses are what make that ratio bite: Wegovy injectable is 2.4 mg weekly; Wegovy pill is up to 25 mg daily. On a per-patient-year basis, an oral semaglutide patient consumes on the order of seventy times the peptide of an injectable patient.

The second route is to abandon the peptide and find a small organic molecule that fits the same receptor. That is orforglipron: a non-peptide, made by conventional multistep organic synthesis in a chemical plant rather than by solid-phase peptide synthesis, taken at any time of day with no food or water restriction. Structure Therapeutics' aleniglipron is the same idea from a different chemical series, and its own filing makes the economic claim explicitly: small molecules are "more scalable to meet global demand" than peptides.

Which route wins depends entirely on where the price lands. At $1,350 a month, Route A wins on efficacy and nobody cares about cost of goods. At $245, and with a government demonstration explicitly designed to enrol cash-constrained patients, the calculus inverts. Part II works that through.

6 · Trials, estimands and what evidence is actually worth — the second place value is created

A pivotal obesity programme is now a very large object. ATTAIN-1 randomised 3,127 participants across ten countries for 72 weeks; Viking's VANQUISH-1 enrolled roughly 4,500 patients in about five months; Structure's ACCOMPLISH-1 will enrol up to 3,600 and ACCOMPLISH-2 up to 1,100. Cardiovascular outcome trials are larger still: Novo's REDEFINE 3 runs to 7,000 patients and Amgen's MARITIME-HF to a similar order.

Amgen's build-out measures what that commitment now costs. Its 2025 annual report records two global Phase 3 studies initiated in March 2025, four more underway by November 2025 (Amgen Form 10-K for 2025, product and pipeline update), and nine running by the first quarter of 2026. Six trials added in twelve months, before a single Phase 3 efficacy result existed.

The cost of that machinery is the real barrier to entry in this industry, and it bites in a specific place. Viking's research and development expense was $265.9 million in the first half of 2026 against $101.5 million a year earlier — a 162% increase driven almost entirely by two Phase 3 trials that have not yet read out. Structure lost $106.2 million in the second quarter alone. Amgen told investors that its research and development increase in both the first and second quarters of 2026 was "driven by higher spend in Later-Stage Clinical Programs, primarily those related to MariTide" (Amgen Form 10-Q, quarter ended 30 June 2026). Amgen is now running nine MariTide Phase 3 studies simultaneously with three more in type 2 diabetes to start in 2026.

When a company cannot run a head-to-head, it commissions an indirect comparison. In June 2026, a Novo-funded population-adjusted indirect treatment comparison in Diabetes, Obesity and Metabolism put oral semaglutide 25 mg ahead of orforglipron 36 mg by 3.2 percentage points on the treatment-regimen estimand, with an odds ratio of 4.1 for discontinuation due to any adverse event and 13.9 for gastrointestinal discontinuation (PMID 42225305). The authors include Novo employees and the analysis was paid for by Novo. It compares OASIS 4 against ATTAIN-1 — two different trials, two different populations. Lilly, in the one place a genuine randomised head-to-head exists, reports the opposite direction: in ACHIEVE-3, orforglipron 36 mg cut A1C by 2.2% against 1.4% for oral semaglutide 14 mg, and delivered 9.2% weight loss against 5.3% (Lilly, ACHIEVE-3 detailed results, 26 February 2026).

Company communications systematically blur one further distinction: between a trial that is running and a trial that will answer something. The public registry gives a dated answer to that question, and the dates are later than the narrative implies.

Both the head-to-head and the indirect comparison are true and neither settles the efficacy question, because ACHIEVE-3 used the diabetes doses of oral semaglutide (7 and 14 mg), not the 25 mg obesity dose. There is, as of the evidence cutoff, no randomised head-to-head of the two orals at their respective obesity doses. Anyone claiming to know which oral is better on efficacy is reading across trials. What is not in dispute, because both parties report it, is that orforglipron discontinues patients at roughly twice the rate: 8.7% and 9.7% against 4.5% and 4.9%.

Discontinuation differential: observed in a randomised head-to-head, both sponsors reporting the same figures. Relative obesity-dose efficacy: unresolved; only indirect, sponsor-funded comparisons exist.

7 · Label expansion — an access strategy dressed as a science strategy

The list of indications these molecules now carry or are chasing reads like a review of internal medicine: type 2 diabetes, chronic weight management, cardiovascular risk reduction, obstructive sleep apnoea, metabolic dysfunction-associated steatohepatitis, chronic kidney disease, heart failure with preserved ejection fraction, knee osteoarthritis pain, peripheral artery disease, and — in Lilly's ATTAIN follow-on programme — stress urinary incontinence.

It is tempting to read this as scientific ambition. The commercial logic is narrower. Under United States law, Medicare Part D was barred from covering drugs used for weight loss. It was never barred from covering drugs used for diabetes, sleep apnoea or liver disease. Every indication a manufacturer adds converts a slice of the obese population from an uncovered cash patient into a covered Part D patient at a formulary price. That is why the same molecule sells as Ozempic and Wegovy, and why Novo has filed the ESSENCE steatohepatitis data for the Wegovy pill with an FDA decision expected in the first quarter of 2027.

Lilly has run the same play harder than anyone. Beyond ACHIEVE-3, it reported ACHIEVE-2 against dapagliflozin and ACHIEVE-5 as an add-on to insulin glargine at the American Diabetes Association's 86th Scientific Sessions, publishing in The Lancet and JAMA, and said it would submit Foundayo for type 2 diabetes to the FDA by the end of the second quarter (Lilly, 8 June 2026). In Europe the same logic runs through the regulator: the EMA's own announcement of the oral Wegovy authorisation is headlined "First oral GLP-1 treatment for weight management" and dated 22 May 2026 (EMA news), against a Wegovy marketing authorisation that has been in force in the European Union since 2021 (EMA, Wegovy EPAR). One molecule, one authorisation, expanding by indication and by formulation into every channel a payer will fund.

CMS has noticed. The Medicare GLP-1 Bridge is explicitly restricted to beneficiaries seeking a GLP-1 solely for weight management; anyone with type 2 diabetes, moderate-to-severe sleep apnoea, or non-cirrhotic steatohepatitis with moderate-to-advanced fibrosis is ineligible for the Bridge and must go through their Part D plan. CMS states it "will monitor Part D formulary and utilization management practices... to ensure plans do not shift coverage or access decisions from the Part D benefit to the Medicare GLP-1 Bridge." Two payment systems, two prices, and a bright line between them policed by diagnosis code. Label expansion has become a channel-allocation exercise.

8 · Making a peptide — the chemistry that sets the floor under the price

Semaglutide and tirzepatide are built by peptide synthesis: amino acids added one at a time to a growing chain, each addition requiring coupling, deprotection and washing, with yield losses compounding across roughly thirty cycles. Then a lipid side chain is attached, the product is purified by preparative chromatography, and it is filled into a device. It is capital-intensive, solvent-intensive and slow, and the industry spent 2023 and 2024 unable to make enough of it.

The people who make it for hire are a short list. PolyPeptide, Bachem, CordenPharma and a set of Indian and Chinese producers do the active ingredient; Catalent, Vetter, Thermo Fisher and a handful of others do the sterile fill. The direction of travel in 2026 is unambiguous: the two leaders are taking the chain in-house, and the contract layer is consolidating.

Novo Holdings — the Novo Nordisk Foundation's investment arm and the controlling shareholder of Novo Nordisk — completed the acquisition of Catalent for an enterprise value of approximately $16.5 billion in December 2024, and immediately sold three of Catalent's sterile fill-finish sites, at Anagni in Italy, Bloomington in Indiana and Brussels in Belgium, to Novo Nordisk itself (Novo Holdings, 18 December 2024). Lilly has committed at least $27 billion to four new United States sites, including a $6.5 billion active-ingredient plant in Texas and a $6 billion facility in Huntsville, Alabama, both of which will make orforglipron (Lilly, 9 December 2025), plus a further $4.5 billion announced in the second quarter of 2026 to expand its Indiana sites.

Meanwhile Samsung Biologics launched an all-cash tender offer on 19 July 2026 to acquire PolyPeptide Group at CHF 44.31 per share, an implied equity value of about CHF 1.46 billion and a 40% premium to the undisturbed price, expressly "to broaden the company's portfolio into peptide therapeutics... driven by surging global demand for obesity treatments" (Samsung Biologics, 19 July 2026). A specialist peptide maker with more than seventy years of history and over a thousand peptides produced sold itself. That is not the behaviour of a bottleneck owner.

9 · Fill-finish and the device — yesterday's bottleneck, and what it is worth today

Between 2022 and 2024 the binding constraint sat at the very end of the line: the ability to fill a sterile drug product into an autoinjector at high speed. The lead time on a new high-speed filling line exceeded two years. That scarcity is what put both companies into shortage and what created the compounding industry that filled the gap.

That constraint is worth much less now. Thermo Fisher's Laboratory Products and Biopharma Services segment, which houses its sterile drug-product and pharma-services business, reported a segment income margin of 14.0% in the second quarter of 2026, up two-tenths of a point. Its growth came from clinical research and the Clario acquisition, not from fill-finish; the phrase "GLP-1" appears nowhere in the filing (Thermo Fisher Form 10-Q, quarter ended 27 June 2026). Lonza delivered a strong half — CHF 3,374 million of sales, up 16.0% at constant currency, with a CORE EBITDA margin of 34.8% — but its capacity announcements and its commentary concern antibody-drug conjugates and mammalian capacity, and it made no peptide or obesity disclosure. Bachem, the purest listed peptide play, grew first-half sales just 7.3% in local currency and saw its EBITDA margin fall to 25.4% from 29.1% as ramp-up costs on its new Building K landed (Bachem, 30 July 2026).

Bachem still guides to 35–40% full-year growth in local currency. Delivering 7.3% in the first half and 35–40% for the year requires the second half to grow at roughly sixty per cent or better. A capacity owner with genuine pricing power does not have to load an entire year of growth into one customer ramp.

10 · The tablet alternative — a different factory and a different unit cost

Neither Lilly nor Novo discloses cost of goods per patient-month, and this report does not invent one. What is disclosed is the direction of each company's gross margin under an identical price shock, and the shape of the two supply chains.

A small-molecule tablet is made in a chemical plant: reaction vessels, crystallisation, drying, granulation, compression, coating. It is the oldest and cheapest manufacturing process in the pharmaceutical industry, it does not require aseptic processing, it does not require a device, and it does not require cold chain. Lilly's two new active-ingredient plants for orforglipron are conventional synthesis facilities. A peptide taken orally requires the same solid-phase synthesis as the injectable, at roughly seventy times the mass per patient-year, plus the SNAC enhancer, plus tabletting.

That asymmetry is the single most important unmodelled fact in this industry, and Part II tests what it implies against the disclosed margin data rather than asserting it.

11 · List, net, and cash — three prices for one drug

Novo Nordisk's accounts show the scale of the American gross-to-net distortion directly. In the first quarter of 2026 Novo reversed a rebate provision of DKK 26.8 billion related to the 340B Drug Pricing Program — approximately $4.2 billion of previously accrued rebates that turned out not to be owed, on a single programme, in a single quarter (Novo Nordisk Form 6-K, 4 August 2026). Novo now reports "adjusted" sales specifically to strip these reversals out. Lilly's own filing discloses that changes in estimates for prior-period United States rebates and discounts alone moved second-quarter US revenue by three per cent.

These are not rounding errors. They are the mechanical consequence of a system in which the posted price is fictional and the real price is a bilateral secret. That system is what the 2026 policy settlement is dismantling, and the dismantling is where the money moves.

12 · The PBM machine — what it does, how it is paid, and what just happened to it

A pharmacy benefit manager sits between the drug maker and the employer or health plan. It builds the formulary, negotiates rebates off list, contracts pharmacy reimbursement, and adjudicates claims. Three of them — CVS Health's Caremark, Cigna's Express Scripts, and UnitedHealth's Optum Rx — process the large majority of American prescriptions. Historically they were paid from the spread: a share of the rebate, plus administrative fees calculated as a percentage of list price. That structure gives a PBM a direct financial interest in a higher list price, which is precisely what the Federal Trade Commission alleged.

On 14 July 2026 the FTC announced a settlement with Caremark, following an identical settlement with Express Scripts in February 2026. The complaint alleged that the three PBMs "artificially drove up the list prices of drugs by creating a system that preferences rebates," letting the PBMs "keep some or all of the inflated rebates as well as fees paid by drug manufacturers that were based on the list price." The settlement "delinks PBM fees from drug list prices," requires Caremark to "cease discriminating against low wholesale acquisition cost versions of a drug on its standard formularies," and requires a standard offering that passes rebates through at the point of sale (FTC, 14 July 2026).

Delinking fees from list price removes the mechanism by which the intermediary captured value from a high posted price. Most of what the PBM earned in this category rested on that mechanism.

13 · The state as price-setter — three instruments, one direction

The United States government now sets the price of these drugs through three separate mechanisms, and they behave differently.

1. 01
   IRA negotiation

   Semaglutide selected in the second cycle. Maximum fair price of $274 per 30-day equivalent from 1 Jan 2027, a 71% discount to a $959 CY2024 list. Statutory, durable, and it applies to Part D volume.
2. 02
   The MFN agreements

   Voluntary deals struck with the administration on 6 November 2025 covering Medicare Part D, Medicaid and the direct-to-patient cash channel. Novo told investors the same day that it expected "an estimated direct, negative low single-digit impact on global sales growth in 2026" and a three-year tariff exemption (Novo Nordisk Form 6-K, 6 Nov 2025). Contractual, not statutory.
3. 03
   The Medicare GLP-1 Bridge

   A time-limited demonstration, 1 Jul 2026 to 31 Dec 2027. $245 net to the manufacturer, $50 to the patient, claims run through a CMS central processor — Humana's LI NET infrastructure — outside the Part D benefit. No coordination of benefits, no coupons.

Evidence: CMS IPAY 2027 fact sheet; CMS press release of 6 May 2026; CMS "Information for Part D Plans" FAQ. Method: figures as published; the $245 is a manufacturer net price, the $274 a statutory maximum fair price, and they are not the same measure. Synthesis: three instruments, one direction, and only one of them requires an act of Congress to change.

The Bridge's architecture is the tell. CMS states it plainly to prescribers: "Part D sponsors will not carry risk for eligible GLP-1 drugs furnished under the Medicare GLP-1 Bridge, and Part D sponsors do not have to opt in" (CMS, Information for Providers). The plan is neither at risk nor consulted. The programme is "nationwide and available in all states and territories," and CMS uses "a single central processor to manage prior authorization, claims adjudication, and payment to pharmacies" (CMS, Information for Pharmacies; CMS, Medicare GLP-1 Bridge). CMS did not ask Part D plans to cover obesity drugs. It built a parallel payment rail, appointed a central processor, fixed the manufacturer's net price at $245, fixed the patient's copay at $50, and specified that neither figure touches the beneficiary's Part D true out-of-pocket accumulator. Coupons "may not be applied." Rebates are not part of the transaction. The pharmacy benefit manager is not in the transaction.

For the first time, an American federal programme is buying a blockbuster drug the way a national health service buys one: a fixed price, direct from the manufacturer, with no intermediary taking a spread. Whether that survives past 31 December 2027 is the single most consequential open question in the sector.

Which is why the two instruments should never be blended into one claim. The maximum fair price is statute. It does not expire, it takes an act of Congress to undo, and $274 binds Part D from 1 January 2027 whatever else happens. The Bridge is a short-term demonstration run under Section 402 authority, already extended once from the end of 2026, and it ends on 31 December 2027 unless CMS extends it again (KFF, BALANCE Model and Medicare GLP-1 Bridge, updated 11 May 2026). Suppose it lapses with no successor. The administered-price reading does not collapse: the statutory maximum fair price still binds, the cash channels still clear at $149–449, and post-exclusivity supply in Canada, India and Brazil goes on setting the world's cost reference regardless of what CMS does. What weakens is the sharpest leg of it — the demonstration that a federal programme can buy a blockbuster with no intermediary in the transaction at all — and Part II's verdict on the intermediary layer would have to rest on the FTC consent orders alone. The durable half of this argument is the statute. The Bridge is the proof of concept.

The architecture matters more here than the volume does. CMS has published no uptake data as of the evidence cutoff — the programme is five weeks old. The eligible population is bounded by the clinical criteria above and by the exclusion of anyone who has already filled a GLP-1 through Part D in calendar 2026, and CMS has not disclosed a projected cost or an eligible-population estimate for it. Two adjacent facts do bound the ambition: on 21 April 2026 CMS announced it would not implement the BALANCE Model in Medicare Part D in 2027, citing insufficient plan participation, and Medicaid GLP-1 coverage for obesity had fallen to thirteen states as of January 2026 from sixteen a year earlier (KFF, "What to Know About the BALANCE Model for GLP-1s in Medicare and Medicaid and the Medicare GLP-1 Bridge," updated 11 May 2026). The Bridge is the instrument that survived, and it was extended through the end of 2027. The Bridge's significance in this report is not its 2026 revenue. It is that it establishes $245 as a price the United States government has demonstrated it can pay, directly, at national scale — and prices, once demonstrated, are very hard to un-demonstrate.

14 · Compounding, telehealth, and the aftermath — the grey market is closing and the cash market is not

When both leaders were in shortage, United States compounding law permitted pharmacies to make copies. A large telehealth industry grew on top of that permission. The FDA declared the tirzepatide shortage resolved in December 2024 and the semaglutide shortage in February 2025, and set phased wind-down deadlines. On 1 April 2026 it restated the boundary for compounders: a 503A product is "essentially a copy" if it has the same active ingredient at the same, similar or easily substitutable strength by the same route, unless a prescriber documents a significant difference for an identified individual patient (FDA, drug alerts and statements, 1 April 2026). That sentence is what ended the personalised-dose workaround. On 1 May 2026 it went further, publishing a notice proposing not to include semaglutide, tirzepatide or liraglutide on the 503B bulks list — the list of substances outsourcing facilities may compound from bulk (91 FR, Docket FDA-2018-N-3240, 1 May 2026). The comment period, extended once, closed on 30 July 2026. A finalised exclusion would end industrial-scale compounding of these molecules.

Hims & Hers announced a strategic shift for its United States weight-loss business in March 2026, taking $33.5 million of restructuring charges including $28.5 million of inventory write-downs in the first quarter, and moving from compounded product to branded access (Hims & Hers Form 10-Q, quarter ended 31 March 2026). Its gross margin fell from 73% to 65%. Its first-quarter revenue grew 4%. It swung from a $49.5 million profit to a $92.1 million loss.

The accounting change tells you where the value went. Under Hims's "Platform Partner" arrangements the company now accounts for prescription access as an agent, not as principal, because the platform partner controls fulfilment, packaging, refunds and displayed pricing. An agent books a fee. A principal books the whole prescription. The economics of the entire telehealth GLP-1 channel have been re-based from gross to net, and the difference has gone back to the manufacturer.

Novo's own filing confirms the shape of the new arrangement: Wegovy pill is available through nine telehealth organisations, and a subscription programme runs through Ro, WeightWatchers and LifeMD, "with Hims & Hers and Sesame expected to come online soon." The telehealth companies are now a distribution channel for the brand rather than a competitor to it. That is a better business than an illegal one, and a much smaller one.

15 · Exclusivity by geography — the cliff already happened outside the United States

Patent expiry is normally discussed as a future event. In this class, three natural experiments have already run — Canada, India and Brazil.

In Canada, Novo Nordisk allowed a maintenance fee to lapse and the semaglutide patent went with it. Health Canada approved Dr. Reddy's generic on 28 April 2026 and Apotex's on 1 May; both launched commercially in the week of 20 May, at roughly one third of the brand price (Global News, 20 May 2026). In India, the compound patent expired on 24 March 2026 (STAT, 17 March 2026), and generic prices there "could drop by 90%" against the branded product, on STAT's reporting. No source this report was able to open gives a verified Indian retail price, so none is quoted.

Brazil is the third natural experiment, and it is already past the approval gate. The compound patent on semaglutide expired in March 2026. On 26 May 2026, ANVISA registered EMS's Ozivy — the first synthetic semaglutide multi-dose pen authorised in Brazil, classified as a new synthetic medicine rather than a classic generic of a biologic, for type 2 diabetes as an adjunct to diet and exercise (ANVISA, 26 May 2026). On 29 July 2026 ANVISA registered five further synthetic semaglutide pens, bringing Hypera (Semavy, Zempneo), Sun Pharma-linked brands, and Sandoz's Owozy into the same market (Valor International, 31 July 2026). Obesity-dose labelling for several of those pens was still under study at the evidence cutoff; the structural point does not wait on the label. A third large middle-income market has converted exclusivity into a multi-entrant pen race within four months of the patent date. Novo does not break Brazil out of Emerging Markets, so the franchise share at risk is undisclosed and is not estimated here.

Novo's own description of the Canadian outcome is the most valuable sentence in its second-quarter filing: "In Canada, the loss of exclusivity for semaglutide has led to lower realised prices while volumes have remained broadly stable for Novo Nordisk." Volume held. Price did not. The rent vanished without the franchise vanishing.

The last piece of the exclusivity picture is not public. Novo has "entered into settlement agreements with Alvogen, Rio Biopharmaceuticals, Sun Pharmaceutical, Dr. Reddy's, Mylan, Zydus and Apotex regarding the US patent litigation for Ozempic," and states that "all terms of the agreements are confidential." Settlements of this kind conventionally specify an agreed United States entry date. That date is not public. Anyone modelling a clean 2032 US cliff is modelling an assumption, not a disclosure.

16 · Demand and persistence — the denominator nobody controls

Penetration is early. Lilly's chief executive stated at the Foundayo launch that "fewer than 1 in 10 people who could benefit from a GLP-1 are taking one." Persistence is poor, and the reason is now partly visible in the trial data — discontinuation for adverse events runs at 5% to 10% in controlled trials with paid drug and study support, and real-world discontinuation is well above that.

Volume growth in 2026 has been extraordinary and price-driven. Lilly's second-quarter revenue rose 48%, decomposed by the company as volume +60% and price −13% (Lilly Form 10-Q, quarter ended 30 June 2026, MD&A). Novo's second-quarter adjusted sales rose 7% at constant currency, with Wegovy pill reaching 2.9 million United States prescriptions in the quarter and more than five million since a January launch — Novo calls it "the strongest-ever GLP-1 volume launch in the US."

The elasticity is real and it is enormous. Cutting the cash price roughly in half and adding a $50 Medicare copay has produced volume growth that more than offsets it — for now. The question that decides the next decade is not whether volume responds to price. It clearly does. The question is what happens to the profit pool when the volume response has been fully harvested and the price is still $245.

1 · Conclusion first — what to believe, and how strongly

Two things now decide who keeps the money in this industry: what it costs to put a milligram into a patient, and whose contract gets paid before the factory does. Patent life, brand, market share and formulary position no longer decide it — and the second quarter of 2026 is the first period in which the accounts prove it.

The load-bearing fact of this report is a pair of margins. In 2024, Novo Nordisk's reported gross margin was 84.7% and Lilly's was 81.3%. Novo was ahead by 3.4 points. In 2025, Novo's fell to 81.0% and Lilly's rose to 83.0%. Lilly was ahead by 2.0. A five-and-a-half point swing, in one year, on each company's own reported measure, in the same product class, under the same demand conditions. That crossover is the fact the market has not yet priced, and it happened before the 2026 price cuts even landed.

The 2026 numbers extend it. Lilly's realised prices fell 13% globally and 3% in the United States; Novo's fell too, under the same most-favoured-nation agreement and the same Medicaid and cash concessions. Lilly's second-quarter gross margin nonetheless rose 1.5 points to 85.8%. Novo's adjusted gross margin fell 4.5 points to 78.2% — of which, the company discloses, roughly DKK 3 billion or 3.8 points was a one-time cost of "right-sizing of manufacturing capacity agreements." Strip that out and the underlying decline is about 0.7 points in the quarter and 1.7 in the half. The honest version of the 2026 gap is roughly two points, not six, and this report says so rather than leaning on the headline.

Nor is the second quarter an isolated print for Novo. Its adjusted gross margin was 80.6% in the first quarter of 2026 against 83.5% a year earlier, a 2.9-point decline the company again attributed to "lower realised prices, one-time costs, as well as a negative currency impact" (Novo Nordisk Form 6-K, 6 May 2026). Q4 2025, FY2025, Q1 2026 and Q2 2026 all moved the same way.

Lilly's path is not monotonic and this report will not pretend it is. Lilly's own first-quarter 2026 gross margin was 81.9%, down from 82.5% a year earlier (Lilly Form 10-Q, quarter ended 31 March 2026), on a US price decline of 7% and a global decline of 13%. The clean comparison is therefore the half: Lilly +0.5 points, Novo −3.8 as reported and roughly −1.7 excluding the one-time item. The direction holds at the level where quarterly noise washes out. The magnitude does not support a stronger claim than that.

The excluded item, though, is not noise. Paying nearly half a billion dollars to shrink contracted manufacturing capacity, in the middle of the largest demand ramp in the history of the class, is one of the most revealing things either company disclosed this year. Section 8 returns to it.

Across the whole run of quarters, the word "one-time" is doing a great deal of work.

The rest of Part II establishes six claims, each falsifiable, each tagged.

1. 1The price is administered. Three separate mechanisms — IRA negotiation at $274, the Bridge at $245, and cash-pay at $149–449 — have converged the American price of a patient-month into a band roughly one fifth of the 2024 list. The manufacturer sets one of the three.
2. 2At that price, cost of goods becomes the competitive variable, and small-molecule chemistry beats peptide chemistry on cost by a margin that no amount of clinical superiority repays at $245.
3. 3Novo is winning the oral volume race and losing the oral margin race. Wegovy pill produced roughly $169 of realised US revenue per prescription in the second quarter. That is a remarkable number to achieve at 1% bioavailability.
4. 4The most defensible rent in the chain is the discovery royalty — Chugai on orforglipron, Zealand on petrelintide, Zihipp's holders on Metsera's assets. No cost of goods, no capex, no payer, no PBM, no most-favoured-nation exposure.
5. 5The contract-manufacturing layer never got to charge for the bottleneck, and is now consolidating because of it. Samsung Biologics is buying PolyPeptide at a 40% premium; Novo Holdings bought Catalent outright; Novo's own capex is guided down.
6. 6The PBM's spread has been legislated and litigated out of the fastest-growing channels. The Bridge routes around it entirely; the FTC settlements delink its fees from the list price it used to inflate.

2 · Where the consensus is incomplete — four beliefs worth revising

The second belief worth revising is that the oral transition is a Lilly event. It is not, at least not yet. Novo launched the first oral obesity drug on 5 January 2026, reached more than five million United States prescriptions in thirty weeks, and by mid-July was writing over 265,000 a week — against Foundayo's $98 million of second-quarter revenue. On volume, Novo won the first year of the oral market decisively. The revision runs the other way: winning volume in a peptide oral at $169 a fill is a strategically necessary act of margin destruction, and Novo's gross margin line is where it shows up.

The third is that the pharmacy benefit manager is the chokepoint. That was true in 2024. As of July 2026 the FTC has delinked its fees from list price, CMS has built a payment rail that bypasses it for the entire Medicare obesity population, roughly 38.7% of Wegovy injection volume is filled outside insurance altogether, and CVS Caremark is opening its largest commercial formularies to competing weight-loss products from 1 October 2026. Four separate erosions of the same position in one year.

The fourth is that capacity is the constraint. Novo now guides 2026 capital expenditure to around DKK 55 billion against DKK 60 billion in 2025 and says investment "is expected to decline" in coming years. A company that believed capacity would remain scarce and valuable would not be tapering. Lilly, meanwhile, raised first-half capital expenditure 64% to $5,259 million. Those are not two companies competing for the same bottleneck. They are one company that has finished building a peptide chain and another that is building a small-molecule one.

3 · The rent map by layer — and what each layer keeps

4 · Who sets the price now — the question that answers all the others

Channel by channel, someone names the number. Who that someone is makes uncomfortable reading for anyone who owns this sector as an intellectual-property story.

A count of channels is not a weighting by volume, and the difference matters. Medicare has been barred by statute from paying for a drug used for weight loss since the benefit was written — 42 U.S.C. § 1395w-102(e) — which is exactly why the Bridge had to be built beside Part D rather than inside it (KFF, "Recent Trends in GLP-1 Use and Spending in Medicare"), and Medicaid obesity coverage stands at thirteen states. The American anti-obesity book is therefore still carried by two channels the manufacturer is a party to: commercial and employer plans, where it sets list and negotiates net through rebates, and its own cash storefront, where it names the number outright. Neither company discloses that mix and this report will not invent one. What can be said is the direction. The administered channels are the growing minority of volume, not the majority — and every step Medicare takes toward covering obesity moves volume from the column the manufacturer prices into the column it does not. The thesis strengthens as that happens. It does not depend on having happened already.

A patent that entitles you to sell but not to price is a licence, not a monopoly.

AIR Research

5 · One price war, two margins — testing the cost thesis against the accounts

The claim that cost of goods now decides who keeps the money needs to survive a hostile reading. The obvious alternative explanation for Lilly's expanding margin is mix: Mounjaro grew 91% and 52% of it is now ex-US, where Lilly may realise a different margin, and Zepbound's US price fell only 3%. Lilly itself attributes the expansion to "improved cost of production and favorable product mix." Mix is in the answer.

But mix cannot explain the divergence, because Novo's mix moved the same way. Novo's international obesity sales grew 37% at constant currency while its United States obesity sales grew 4%. Both companies shifted toward the same lower-priced international volume. Over the half, one expanded margin doing it by 0.5 points and the other lost about 1.7 points underlying (3.8 as reported, of which the disclosed DKK 3bn capacity charge is worth 2.0 at the half-year level). The residual after mix and currency is cost, and Lilly names cost first in its own explanation.

The instrument is blunt. Both numbers are consolidated corporate gross margins. Lilly's spans oncology, immunology and neuroscience alongside the incretins; Novo's still carries an insulin book and depreciation on three fill-finish plants it bought rather than built. Neither is a cost per delivered milligram, neither company publishes one, the gap is two to four points, it changed sign between 2024 and 2025, and two separate DKK 3bn one-offs sit inside four quarters of it. That is corroboration, not proof. The weight belongs on the structural argument instead: a small molecule run in a conventional chemical plant and pressed into a tablet does not share unit economics with a solid-phase peptide filled into a device, and the capital each route demands says the same thing from the other direction — Lilly committing $27bn to four plants, two of them small-molecule API, while Novo guides capex down and pays to cancel contracted capacity. The accounts are the check on that argument. They are not the argument.

Take the three contracted or posted American prices and ask what a manufacturer keeps at plausible costs of goods. Neither company discloses a per-patient-month cost, so the grid below is an arithmetic exercise across a stated range rather than an estimate of either company's actual cost.

Lilly's price cuts are not a defensive response to political pressure. They are an offensive move against a higher-cost competitor. A producer with a structurally lower cost per dose gains share every time the industry price falls, because the price floor that is merely uncomfortable for it is existential for the other. Lilly cut LillyDirect cash prices in December 2025, launched an oral at $149 in April 2026, and expanded gross margin through both. Novo's response was to cut its own list by half effective January 2027 and to accept, in its own guidance, adjusted sales growth of 0% to −6% for the full year.

6 · Reading the oral inflection correctly — volume won, margin conceded

The temptation is to score the oral market on prescriptions and declare Novo the winner. On prescriptions Novo is the winner, decisively and by its own disclosure: more than five million United States prescriptions in thirty weeks, over 265,000 a week by mid-July, and a formulary position on par with the injectable. Foundayo, launched three months later, contributed $98 million in the second quarter.

Novo's filing permits the arithmetic. Wegovy pill United States sales were DKK 3,141 million in the second quarter. Novo's disclosed first-half average dollar rate was 640 DKK per 100 USD. That is $490.8 million. Total Wegovy pill prescriptions in the quarter were "around 2.9 million." Realised revenue per prescription: approximately $169.

The first quarter is not a fair comparison but it is instructive: Wegovy pill sales of DKK 2,256 million on roughly 1.3 million prescriptions imply about $271 a fill, and Novo says that figure was "impacted by pre-launch pipeline fill with wholesalers and telehealth partners." Once the channel filled, the realised price settled at $169.

Set that against what the product is. A daily peptide tablet dosed up to 25 mg, formulated with an absorption enhancer, at roughly 1% oral bioavailability — meaning the active ingredient consumed per patient-year is an order of magnitude above the injectable — realising $169 a month. For comparison, the same company's injectable Wegovy in the United States fell 22% at constant currency in the quarter on lower realised prices with higher volumes.

$169 per fill: computed from two disclosed Novo figures and Novo's own disclosed FX rate — estimated, because "around 2.9 million" is Novo's rounding and a prescription is not necessarily a 30-day supply. The peptide-mass implication is a mechanism, not a disclosure: Novo does not publish cost of goods by product.

The strategic logic is nonetheless sound and this report does not call it a mistake. Novo faced a small-molecule oral entering its market with no food or water restriction and a $149 price. Conceding the oral category would have handed Lilly the entire treatment-naïve, cash-paying population — the largest and least penetrated segment in the disease. Novo bought the category with price, and it worked: Wegovy remains the market leader among branded obesity medicines in the United States by new patient starts.

But a share defence financed by margin is a lead, not an annuity. Every incremental oral semaglutide patient consumes roughly seventy times the peptide of an injectable patient. Novo is therefore filling its own newly acquired Catalent capacity with the least profitable unit of demand it sells, at the exact moment it is guiding capital expenditure down. If oral volume keeps compounding, Novo will need the capacity it has stopped building, to serve the margin it has already conceded.

7 · The royalty layer — the rent nobody underwrites

Every price event described in this report lands somewhere in a manufacturer's income statement. The maximum fair price cuts the top line. Medicaid rebates cut it again. Cost of goods sits below. Capital expenditure sits below that. One claim in the chain sits beyond the reach of all of them.

Lilly's own filing places the Chugai orforglipron royalty in cost of sales. That means it is calculated on worldwide net sales, before Lilly's manufacturing costs, before its selling expenses, before its research budget, and before the price cuts show up anywhere else. Chugai carries no factory risk, no trial risk from here, no payer negotiation, no most-favoured-nation exposure, and no capital requirement. It receives "mid single digits to low teens" of every dollar Foundayo sells, in every country, for the life of the licence, in exchange for $50 million paid in 2018 and a package of milestones.

What that insulates is not everything. The base is worldwide net sales, and net sales are precisely what the maximum fair price, the Bridge and the cash channels compress. So the rate is unreachable and the dollars are not. Chugai cannot have its percentage competed away by a cheaper rival, negotiated down by a payer, or — absent a policy nobody has yet proposed — regulated; it simply earns less per script when Lilly earns less per script. The distinction matters more than it sounds. Every other participant in this chain absorbs the price cut and funds the response to it: the plant, the trial, the salesforce, the rebate. Chugai absorbs the cut and funds nothing. Rate-insulated, base-exposed, cost-free — that is the whole of the position, and the third term is the one that makes it the best claim in the industry.

Chugai confirmed on 24 July 2026 that royalty recognition began in the second quarter, and reported first-half "other revenue" of ¥96.8 billion, up 44.5%. Chugai does not break out the orforglipron royalty, so no part of that increase can be attributed to it with confidence. What is disclosed is the structure, not the amount. For scale, Chugai's full-year 2025 core revenue was ¥1,257.9 billion at a core operating margin of 49.5% (Chugai, 29 January 2026) — a margin that would be remarkable for a manufacturer and is unremarkable for a business whose newest revenue line requires no factory at all.

Evidence: Lilly Form 10-Q (Q2 2026) Note 3; Chugai Q2 2026 results; Roche, collaboration and licensing agreement with Zealand Pharma, 12 Mar 2025 (deal economics) and Roche H1 2026 investor update (petrelintide in the late-stage pipeline); Pfizer Form 10-Q (Q2 2026) Note 2A. Method: terms as disclosed by the paying party. Synthesis: the only layer in this industry whose rate no 2026 price event reaches — exposed to net price through the base, but free of every cost the price war imposes on everyone else.

If the administered price regime persists, the internal rate of return on discovering an incretin and licensing it out now compares favourably with the return on commercialising one — because the discoverer takes a percentage of a top line while the commercialiser absorbs the entire price cut, the entire capital programme and the entire payer negotiation. That is a genuine inversion of how this industry has allocated capital for thirty years, and it argues that the next decade's best obesity returns may accrue to chemistry houses and small biotechs that never build a salesforce.

It also explains, without needing any other motive, why Pfizer paid $8.0 billion for Metsera and why Roche paid $1.65 billion up front for a Phase 2 amylin. Large companies are not buying molecules. They are buying out royalty positions before those positions become the most valuable claims in the chain.

8 · The bottleneck test — what happens to the rent if capacity doubles

Four questions decide whether a scarce node is really a bottleneck: who controls it, how durable that control is, whether it can be replicated, and what happens to the profit if capacity doubles. The peptide manufacturing layer fails on all four.

Take the fill-finish and peptide-API rows specifically. Who controls them? A fragmented set of contract manufacturers, none with a technology the others cannot buy. How durable? Bachem's own numbers answer that: a new plant that produced its first commercial material in the first half of 2026 and immediately cost the company nearly four points of EBITDA margin. Can it be replicated? Samsung Biologics is replicating it by acquisition, at CHF 1.46 billion, and describes peptides as an addition to a "multi-modality" platform rather than a scarce asset. What happens if capacity doubles? The two customers who account for most of the demand have both spent the last two years buying their own.

One disclosure, on 4 August 2026, settled the question. Novo Nordisk's second-quarter adjusted gross margin fell 4.5 points, and the company attributes roughly DKK 3 billion of that — about $469 million, and 3.8 points of margin — to "one-time costs of around DKK 3 billion related to right-sizing of manufacturing capacity agreements." Read that phrase for what it is. In the middle of the fastest volume ramp in the history of this class, one of the two largest buyers of contract peptide capacity paid nearly half a billion dollars to reduce the capacity it had committed to. Companies do not pay to cancel a scarce resource.

The Catalent transaction is the cleanest evidence available anywhere in this industry that scarcity in fill-finish is over. A $16.5 billion CDMO was bought by the investment arm of one of the two customers, which then transferred three sterile fill-finish sites to that customer outright. When a bottleneck is genuinely durable, the bottleneck owner does not sell to the buyer of its output. It raises the price.

Where does the rent go instead? To the customer. That is precisely what Lilly's rising gross margin is measuring, and it is why Novo's own filing describes its capital programme as creating "additional capacity and flexibility across the supply chain" with investment "expected to decline" in coming years. Both leaders converted a supplier's scarcity rent into their own cost of goods, permanently.

9 · Distribution and the vanishing intermediary — four erosions in one year

Pharmacy benefit managers built their position on a simple asymmetry: they controlled access to the covered lives, and the manufacturer's alternative to paying them was to sell nothing. That asymmetry has been attacked from four directions inside twelve months, and no single one of them would have mattered alone.

First, the government built a road around them. The Bridge is not a formulary negotiation; it is a fixed price paid through a CMS-appointed central processor, with the manufacturer's net receipt specified in a public FAQ. Second, the FTC removed the fee mechanism: the Caremark and Express Scripts consent orders delink PBM fees from list price and require a point-of-sale rebate pass-through offering. Third, the manufacturers built their own pharmacies: 38.7% of Wegovy injection volume is now filled outside insurance, and Lilly launched Foundayo through LillyDirect before it reached retail. Fourth, exclusivity itself is ending as a currency — CVS Caremark adds competing weight-loss medicines to its largest commercial template formularies on 1 October 2026.

The financial trace is already visible, and it is not dramatic, which is the point. All three of the large intermediaries show the same pattern. CVS's Health Services segment grew revenue 11.3% in the half and adjusted operating income 1.4%, with margin falling from 3.53% to 3.22%. Cigna's Evernorth Health Services segment, which houses Express Scripts, grew second-quarter adjusted revenue 6.3% to $61,468 million while pre-tax adjusted income from operations fell 1.9% to $1,663 million — a margin of 2.71% against 2.93% a year earlier (The Cigna Group, Form 10-Q, quarter ended 30 June 2026, Note 15). UnitedHealth's Optum Rx earned $1,192 million of operating earnings on $35,736 million of first-quarter revenue, a margin of 3.34% (UnitedHealth Group, Form 10-Q, quarter ended 31 March 2026, segment table). Three companies, three accounting conventions, one direction: enormous flow, a low-single-digit take, and no expansion. Nothing collapsed. The segments simply stopped converting growth into profit, which is what disintermediation looks like from the inside long before it looks like anything from the outside.

Removing a rebate intermediary raises the manufacturer's net receipt on a given list price. That is why Novo can cut Wegovy's list by half to $675 and describe the effect as a 2027 cash flow event rather than a revenue catastrophe: a large share of the gap between $1,350 and $675 was never money Novo kept. The list-price collapse of 2027 is substantially the unwinding of a fiction. What is not a fiction is the $245 the government will pay, and that number does bind.

10 · The cost of admission — why the challenger layer is a balance-sheet story

The competitive threat to the incumbents is real on the science and thin on the funding. Look at what a Phase 3 obesity programme now costs and at who can pay for it.

Viking's position deserves a careful reading. The company has two fully enrolled Phase 3 trials, plans a third programme in the fourth quarter, has secured dedicated capacity from CordenPharma for multiple metric tons of active ingredient annually plus 100 million autoinjectors and over a billion oral tablets, and will make prepayments against that capacity between 2026 and 2028. Against that it holds $501.9 million and lost $286.3 million in six months — and the public registry records primary completion of the subcutaneous Phase 3 on 1 July 2027, one quarter before the company's own stated runway ends. Its board terminated the prior repurchase programme in July 2026 and authorised a new one for up to $500.0 million over three years.

State only what the disclosure supports. A repurchase authorisation is a permission, not a commitment, and the filing says it may be executed "at the discretion of a committee." But an authorisation equal to the entire cash balance, at a company with a stated runway to September 2027 and three Phase 3 programmes to fund, is an unusual signal to send. The most economically coherent readings are that management regards the shares as materially undervalued, or that the authorisation is a defensive posture. Neither is disclosed. What is disclosed is that the arithmetic requires additional capital — through the new at-the-market facility established on 29 July 2026, a partnership, or a sale — before those trials become products.

The purpose of Viking's July 2026 repurchase authorisation is not disclosed and this report does not infer one.

11 · Per-name structural reads — verdict, contracted evidence, falsifier, confidence

Each name is graded on competitive economics — whether it holds a claim on excess return that survives the price regime described above. This is not a view on price, valuation or timing.

One confidence scale, applied to every card and to nothing else. High — the verdict rests on that name's own reported disclosure, read this session. Moderate — it rests on disclosure plus a quantity the company does not publish, so the direction is sound and the magnitude is not. Low — a primary could not be opened, or the load-bearing quantity is undisclosed by every party to it. Confidence grades the evidence under a verdict; it is not a measure of how strongly the verdict is held.

Coverage note. Every name above has had at least one of its own primary filings or releases opened and read in this session, with two stated exceptions: Lonza, whose own release returned an access block and whose figures are taken from a secondary summary (confidence lowered accordingly, no conclusion depends on it), and Zealand Pharma, whose own interim report was not opened — the Zealand economics cited here come from Roche's and the joint release's disclosures, and Zealand is therefore not separately graded.

12 · Incentive map — once the incentives are written down, most of the outcomes stop being surprising

13 · Alternatives considered and rejected — with the condition that reopens each

Alternative 1: the rent stays with the molecule because patents still run to 2032 and 2036. Rejected because a patent confers the right to be the only seller, not the right to name the price, and 2026 demonstrated the difference at scale. Reopens if: the Bridge lapses on 31 December 2027 without replacement, the IRA negotiation programme is materially curtailed by legislation or litigation, and commercial net prices stabilise. Watch the 2028 selected-drug cycle and any statutory change to the negotiation programme.

Alternative 2: peptide manufacturing is the real bottleneck and the CDMO layer will earn a scarcity rent. Rejected on the direct evidence: Bachem's margin fell while it added capacity, PolyPeptide sold itself at a 40% premium rather than raise prices, Catalent was bought by an affiliate of one of its two largest customers, and Novo is guiding its own capital expenditure down. Reopens if: a supply interruption returns any leading product to the FDA shortage list, or if a peptide CDMO reports EBITDA margin expansion attributable to GLP-1 pricing rather than volume.

Alternative 3: the oral transition is a Lilly monopoly. Rejected. Novo launched first, holds the higher-efficacy oral, and by its own disclosure ran more than 265,000 weekly US prescriptions by mid-July against Foundayo's $98 million quarter. The oral market is competitive and Novo currently leads it on volume. Reopens if: orforglipron's tolerability disadvantage proves decisive at scale and Foundayo's share stalls while Structure's aleniglipron is still years from market.

Alternative 4: amylin is the next platform and whoever owns it owns the decade. Rejected as an investable proposition, though the science is real. CagriSema missed non-inferiority against tirzepatide. Petrelintide's differentiation is tolerability, not efficacy, and its Phase 3 has not started. Eloralintide is impressive at 20.1% but is a Lilly asset inside a Lilly portfolio, which means it strengthens an existing position rather than creating a new one. Reopens if: a petrelintide or eloralintide Phase 3 delivers 20%-class efficacy at placebo-level tolerability, which would reset the entire dose-ceiling logic of the class.

Alternative 5: volume growth is so large that margin structure does not matter. This is the strongest alternative and Section 15 gives it a full hearing rather than a paragraph.

14 · Second- and third-order effects — the consequences worth foregrounding

The cheaper the drug gets, the more valuable the low-cost producer becomes relative to everyone else. The inversion at the centre of this report runs against intuition. Falling prices normally destroy value across a sector uniformly. Here they do not, because the cost dispersion between a small-molecule tablet and a 25 mg oral peptide is large and fixed by chemistry. At $675 the difference is a rounding error in gross margin. At $149 it decides who can serve the segment at all. Lilly is not accepting lower prices; it is choosing them.

Administered pricing turns clinical superiority into a cost, not a return. Under a negotiated price the payer captures most of the value of a better drug. Retatrutide at 28.3% will be reimbursed in a system that has already fixed $245 as the price of a weight-management month for a Medicare beneficiary. Amgen's $800 million Otezla impairment — taken on IRA selection, before any price was even set — is the industry's own mark-to-market of that principle. Expect research allocation in this class to rotate from efficacy maximisation toward cost, tolerability, dosing interval and label breadth, all of which either lower cost or widen the covered population.

The best returns in obesity may accrue to companies that never sell a pill. If the commercialiser absorbs the price cut, the capital programme and the payer negotiation while the discoverer takes a percentage of the top line, the risk-adjusted return on discovery-and-licence beats discovery-and-commercialise. Chugai's $50 million in 2018 now yields a mid-single to low-teens percentage of a product Lilly built four factories to make. That is not a fluke of one deal; it is what the structure now rewards.

The Bridge is a template, and templates spread. CMS has demonstrated that it can buy a blockbuster at a fixed price, direct, with a central processor, no rebates, and no plan intermediation — and that pharmacies, prescribers and manufacturers will all participate. Nothing about that architecture is specific to obesity. If it is judged a success, the most durable consequence of the GLP-1 era will not be weight loss. It will be the operational proof that American federal drug purchasing can bypass the rebate system entirely.

Vertical integration into manufacturing is now defensive, not expansionary. Both leaders have bought or built their own peptide and fill-finish capacity. Having done so, they have converted a variable cost into fixed capital and a supplier's margin into their own. The third-order effect is that the marginal cost of an additional patient falls, which lowers the price at which each is willing to compete, which lowers the industry price again. The capacity build of 2024–26 is, in this sense, the direct cause of the price collapse of 2026–28.

And a quieter one worth holding. A treatment that a majority of eligible patients cannot sustain for five years is not, economically, a chronic-disease franchise; it is a subscription with very high churn. Every incentive above pushes the industry toward lower prices and easier dosing, which is to say toward better persistence. That is an unusually happy alignment between what makes money and what helps people, and it is worth noticing when it occurs, because it does not occur often.

15 · The steelman — the strongest case that this whole analysis is wrong

Argue the other side properly, on its best evidence.

The bull case is not that prices will hold. It is that price is the wrong variable entirely. Consider what the 2026 data actually show. Lilly's second-quarter revenue rose 48% to $23.0 billion, with volume up 60% against a 13% price decline — a volume-to-price ratio of nearly five to one. Mounjaro grew 91%, with ex-US revenue up 172%. Lilly raised full-year guidance to $85–87 billion and its performance margin outlook to 49.0–50.5%. Novo, the supposedly disadvantaged party, reported adjusted operating profit up 11% at constant currency and raised its own guidance twice. Both companies are earning gross margins in the high seventies to mid eighties in the middle of what this report calls a collapse in pricing power.

The bull continues: penetration is below one in ten of the eligible population, by Lilly's own account. The government has just made these drugs available to Medicare beneficiaries at a $50 copay for the first time in history, which is a volume event of extraordinary magnitude dressed up as a price cut. Every dollar the price falls converts a cash patient who could not afford treatment into one who can. In a market with that much latent demand, a 75% price reduction that produces a 300% volume increase is not the destruction of a franchise; it is the fastest possible route to a much larger one. And on cost of goods: at a $245 net price with a plausible cost in the tens of dollars, a manufacturer is still earning a gross margin most industries would consider extraordinary. Arguing over whether it is 78% or 86% is arguing about the second decimal place of a very good business.

The bull has one more good card, and it is one this report dealt him. Lilly's own gross margin fell 0.6 points in the first quarter of 2026. If cost position were as decisive as claimed, that should not happen. And the bull would say the central chart proves too little. A single quarter's margin divergence between two companies with different product mixes, different currencies, different accounting conventions and different one-off items is thin evidence on which to hang a decade-long structural claim.

Now the dismantling. The volume argument is correct and it is not a rebuttal, because it is an argument about the size of the pool rather than about who keeps it. The whole question this report asks is where within a large and growing pool the excess return concentrates. "The pool is large" is an answer to a different question.

The margin objection is the more serious one, and it deserves the concession it earns: one quarter is one quarter, and the two measures are not identical. But three independent facts point the same way and none of them is a single quarter's accounting. First, Lilly's own stated explanation is "improved cost of production," which is management attributing the change to cost, not to mix or currency. Second, Novo has independently guided full-year adjusted sales and operating profit growth to a range whose midpoint is negative, which is not what a company with intact pricing power guides to. Third, the mechanism is physical and verifiable outside the accounts: a 25 mg daily oral peptide at roughly 1% bioavailability consumes an order of magnitude more active ingredient per patient-year than the injectable it is designed to replace, and small-molecule synthesis is cheaper than solid-phase peptide synthesis at every scale ever reported. The margin data are consistent with a mechanism that exists independently of them.

And on "78% is still a great business" — of course it is. That is the trap. A business earning 78% gross margin can still be the worse of two businesses, can still lose share every year to the one earning 86%, and can still see its enterprise value re-rated on the trajectory rather than the level. What you are paying for is the direction of the gap, not its level. Cleaned of the one-time item, that gap widened about 2.2 points over the first half — modest, one-directional, and moving with a mechanism that exists independently of the accounts.

16 · Falsifiers, dated predictions, and what to watch — the honest ledger

Five dated, testable predictions. Each is falsifiable on a stated date by a stated observation.

1. 1For FY2027, Novo's adjusted gross margin will be at or below 81% — that is, it will not recover to its FY2024–25 level even with the one-time capacity charge behind it — while Lilly's reported gross margin, 83.0% for FY2025, will be higher in FY2027 than it was in FY2025. Estimated Wrong if Novo prints an adjusted gross margin of 82% or better in any quarter of 2027, which would mean the underlying decline was transitory rather than mix-and-cost driven.
2. 2The Medicare GLP-1 Bridge, or a successor with a directly negotiated manufacturer net price, will still be operating on 1 January 2028. Consensus Assumption Wrong if obesity coverage reverts wholly to plan-negotiated Part D formularies with no fixed federal net price.
3. 3Bachem will miss the 35–40% full-year sales growth it has guided to, or will itself be acquired or taken private, by 31 December 2027. Naming the asset rather than the category is deliberate: "some peptide CDMO consolidates" is already happening and predicts nothing. Bachem delivered 7.3% in the first half against that guidance, which is the specific arithmetic this report doubts. Estimated Wrong if Bachem lands 35–40% for a full year with an EBITDA margin above 30% and stays independent.
4. 4Oral agents will exceed one third of new US anti-obesity patient starts by the end of 2027, on either company's disclosed new-patient-start data. Estimated Wrong if orals remain below a quarter of new starts, which would mean the cash-and-naïve segment is smaller than the 2026 launch data imply.
5. 5Two separate upfront payments of $1bn or more will be disclosed for pre-Phase-3 obesity assets before the end of 2027 — the Roche–Zealand bar of $1.65bn, cleared twice, not the nine-figure deals that now land routinely. Buying the royalty position remains cheaper than being on the wrong side of one, and the bar is set above the current run-rate precisely so the prediction can fail. Estimated Wrong if fewer than two such billion-dollar pre-Phase-3 upfronts are announced, which would say large-cap business development has moved to de-risked Phase 3 assets and the royalty layer is being priced, not accumulated.

Evidence register — dated and tiered

Tier 1 = issuer filings, regulator publications, company releases. Tier 2 = reputable wire, trade or peer-reviewed. Tier 3 = single-source, aggregated or estimated, used only where flagged. Every URL below was opened and read in the preparation of this brief. Event dates are given where they differ from publication dates.

Not verified / not load-bearing — stated plainly, and carrying no conclusion

1. 1Cost of goods per patient-month for any product in this class. Neither Lilly nor Novo discloses it. Figure 11 is an arithmetic grid across a stated range, not an estimate of either company's position. The cost thesis rests on disclosed margin direction and on the physical mechanism of bioavailability, not on any cost figure.
2. 2The magnitude of Chugai's orforglipron royalty. Lilly discloses the rate band ("mid single digits to low teens"); Chugai discloses that recognition began; neither discloses the amount. Roche's grade is capped at "Estimated" for this reason.
3. 3United States generic semaglutide entry dates. Novo has settled with Alvogen, Rio, Sun, Dr. Reddy's, Mylan, Zydus and Apotex on confidential terms. The 2032 compound-patent date is disclosed; any agreed earlier entry date is not. No conclusion here assumes a clean 2032 cliff.
4. 4Canada's, India's and Brazil's share of Novo's franchise. Novo reports Canada inside EUCAN and India and Brazil inside Emerging Markets and breaks out none of the three. Figure 6's red bars mark regions containing a lapsed market, and the exposed share is bounded above by 26.5%, certainly materially less.
5. 5Lonza's H1 2026 figures. Lonza's own release and the secondary investing.com summary both returned access blocks on re-verification (403). Figures in the body (CHF 3,374m sales, 34.8% CORE EBITDA margin) match the company's disclosed H1 2026 release as reported by multiple wires but were not re-opened as a primary document in the amend pass. Grade remains "Consensus Assumption"; no conclusion depends on them.
6. 6Zealand Pharma's own financial position. Zealand does not file with the SEC and its interim report was not opened here. The petrelintide economics cited come from Roche's release and the joint announcement. Zealand is deliberately not graded.
7. 7The purpose of Viking's July 2026 repurchase authorisation. Disclosed as a permission of up to $500.0m over three years at a committee's discretion. No motive is disclosed and none is inferred here.
8. 8Hims & Hers Q2 2026 results. Not reported as of the 7 August 2026 evidence cutoff. The Hims read rests on Q1 2026 disclosure and is dated accordingly.
9. 9Amgen MariTide Phase 3 efficacy. Nine Phase 3 studies are running; none has reported. Amgen's neutral grade rests on programme scope and mechanism, not on unreported data.
10. 10Viking VANQUISH efficacy. Both trials are fully enrolled; no topline has been reported as of 7 August 2026. Any claim about VK2735's Phase 3 performance would be speculation and none is made.
11. 11Whether the 503B bulks-list exclusion has been finalised. The comment period closed 30 July 2026; the notice was a proposal. This report describes the proposal, not a final rule.
12. 12Orforglipron dose labelling in ACHIEVE-3. Lilly's 26 February 2026 release describes the trial arms as orforglipron 12 mg and 36 mg; its 8 June 2026 release describes the same head-to-head as "Foundayo 9 mg and 17.2 mg." This report uses the February release's labels because that is the release whose full results table was read. The relationship between the two labelling conventions is not disclosed and is not relied on for any conclusion.
13. 13Novo's realised price per Wegovy pill prescription. The $169 figure is computed from two disclosed Novo numbers and Novo's own disclosed FX rate, but "around 2.9 million" is Novo's rounding and a prescription is not necessarily a 30-day supply. Treat it as an estimate of the right order, not a disclosure.

AIR Research · Evidence cutoff 7 August 2026. This document is an internal structural assessment of industry economics and competitive position. It is not investment advice, not a recommendation, and not a solicitation. No price targets, entry or exit levels, or position sizing are expressed or implied. Company grades describe competitive economics only.