Three Phase 3 Wins Lift Amgen's Growth Case Beyond MariTide, but Denosumab Steepens to a 45% Decline and the Stock Already Trades at 20.5x

Results vs. Consensus

Year-over-year comparison (Q3'26 vs. Q3'25)

Sequential comparison (Q3'26 vs. Q2'26): where the spending went

Quality of Beat/Miss

• Revenue: High quality, with one large hole. Seventeen products grew at double digits and seventeen annualize above $1B. The six growth drivers added $1,518M of sales year over year while Prolia and XGEVA lost $758M, and the rest of the portfolio was flat. Product sales outside denosumab grew 20%. The quality issues are below the brand level. Enbrel's 5% growth came from a 15-point favorable change in estimated sales deductions and an 8-point inventory build against a 14% lower net price. KRYSTEXXA's 31% was mostly price. Repatha cleared the Street's $1.04B estimate and passed $1B in a quarter for the first time.
• Margins: Steady, with a timing benefit. Non-GAAP operating margin of 47.3% was up 0.2 points year over year and down 1.1 points from Q2 as R&D rose 13.7% sequentially on the guided business-development upfront and early-pipeline spend. Cost of sales stayed elevated at 19.3% of product sales on profit share and mix, and management guided it higher again in Q4.
• EPS: Operational versus consensus, flattered versus last year. Non-GAAP pre-tax income grew 12.0%, helped slightly by interest expense of $653M versus $685M and higher non-GAAP other income ($164M versus $117M). The 2.3-point lower tax rate then lifted net income growth to 15.2%. GAAP EPS of $7.08 includes $1,465M of net gains on equity investments, mostly the BeOne stake, and is not a measure of the operating trend.

Segment Performance

Amgen reports one segment, so the useful cut is again management's six growth drivers against the loss-of-exclusivity base. The growth engine was steady to slightly faster: oncology accelerated, biosimilars cooled from a very strong Q2, and everything else held its pace. The change this quarter is on the other side of the ledger, where the denosumab decline broke from a 33% run-rate to 45%.

General Medicine: Repatha and EVENITY hold their pace while denosumab breaks lower

Repatha delivered $1,085M, up 37% for a second straight quarter, and the US business grew 51% to $666M. US average weekly new-to-brand prescriptions rose more than 60% year over year, up from more than 50% in Q2. Management cited urgency to treat, the VESALIUS-CV data and the ACC and AHA guidelines updated in March, which support PCSK9 inhibitors without prior ezetimibe, and it expects more than 3 million additional patients worldwide to start Repatha this year. EVENITY grew 39% to $751M, with international sales up 56%, and holds nearly 60% volume share of the Japanese bone-builder category.

"Repatha's momentum continues to accelerate with U.S. average weekly new-to-brand prescriptions increasing over 60% compared with the prior year." — Murdo Gordon, EVP, Global Commercial Operations

Denosumab is the change. Prolia fell 45% to $632M and XGEVA fell 47% to $288M, so the franchise lost $758M of sales year over year and $191M from Q2. In the US the declines were 49% and 52%. The release attributes Prolia's fall to 23% lower volume, with the remainder split equally between lower net price and changes to estimated sales deductions, and XGEVA's to 32% lower volume and 11% lower price. Unlike in Q2, management offered no characterization against its own expectations.

Assessment: Repatha is the strongest pillar in the company and still getting stronger, with the March guideline update supporting prescriptions as an oral competitor launches. Denosumab is the reverse of our August read. We argued then that a 32% to 33% run-rate tracking to plan deserved a lower risk premium. A move to 45% with a 17% sequential dollar drop is not a run-rate, it is a second leg down as volume loss and price concessions compound. The franchise is now 9.3% of product sales, so each further point of erosion costs less, but the 2026 drag is larger than we modeled.

Inflammation: TEZSPIRE compounds, Enbrel's growth is accounting, Otezla's decline is real

TEZSPIRE grew 43% to $538M on volume. US new-to-brand prescriptions grew 15% against 3% for the severe asthma market, and the drug holds a 34% new-to-brand share among US allergists. Enbrel grew 5% to $611M, but on 15% favorable changes to estimated sales deductions and 8% higher inventory, against a 14% lower net price from Medicare Part D price setting and lower volume. Otezla fell 18% to $481M on 13% lower volume, 9% unfavorable deduction changes and 4% lower price, and its international sales fell 53%.

"TEZSPIRE's growth continues to outperform the severe uncontrolled asthma market with U.S. new-to-brand prescriptions growing 15% year-over-year versus 3% market growth." — Murdo Gordon, EVP, Global Commercial Operations

Assessment: TEZSPIRE now looks like a durable 40%-plus grower with a third indication on the way, discussed below. The legacy half of the franchise is weaker than the reported numbers suggest. Strip out the deduction true-up and the inventory build and Enbrel declined. Otezla lost 13% of volume globally, its US sales fell 10%, European generics more than halved its international sales, and it enters Medicare price setting in 2027.

Rare Disease: UPLIZNA nearly triples, TEPEZZA turns negative in the US

Rare disease grew 21% to $1,611M, the same rate as Q2. UPLIZNA grew 191% to $451M, up 35% from Q2, with new and continuing gMG patients up 45% sequentially, and it is now launched in 25 markets outside the US. KRYSTEXXA grew 31% to $419M, mostly on net price. TEPEZZA fell 6% to $526M and 15% in the US, which the release attributes to lower inventory and the call to a build in the year-ago quarter. Ultra-rare products fell 45% to $110M. TAVNEOS fell 2% to $105M with international sales at zero.

"In the U.S., UPLIZNA is demonstrating significant growth in generalized myasthenia gravis, or gMG, with new and continuing patients increasing 45% quarter-over-quarter." — Murdo Gordon, EVP, Global Commercial Operations

Assessment: The concentration we flagged in August has deepened. UPLIZNA added $296M of the segment's $279M of growth, so the rest of rare disease shrank slightly in aggregate. That makes UPLIZNA's indication ladder, now including membranous nephropathy, the segment's growth case. TEPEZZA's US decline is partly a comparison effect, but a second quarter of deceleration after Q2's 14% means the international launches are now carrying the brand.

Oncology: IMDELLTRA and BLINCYTO accelerate the segment ahead of the first-line data

Innovative oncology grew 21% to $2,141M, up from 18% in Q2. IMDELLTRA grew 128% to $406M and 41% from Q2 on volume. A September 14 label update cut post-dose monitoring for the first two doses from 22 to 24 hours to 6 to 8 hours, and management says it is already helping community clinics. BLINCYTO grew 27% to $496M, with international sales up 46%. Nplate grew 3%, or 11% excluding US government orders of $65M versus $90M a year ago. KYPROLIS fell 8% on volume.

"The FDA recently shortened the recommended monitoring period for the first 2 doses from 22 to 24 hours to just 6 to 8 hours." — Murdo Gordon, EVP, Global Commercial Operations

Assessment: In August we called the monitoring burden IMDELLTRA's real constraint and expected relief to arrive gradually. The label change arrived within a quarter, and management says it is already helping uptake. This revenue is all second-line or later. The first-line maintenance data discussed below are not in these numbers yet.

Biosimilars: a slower 26%, with the next wave filed

Biosimilars grew 26% to $977M, down from 29% in Q2. PAVBLU grew 41% to $300M on 71% volume growth against an 18% lower net price, and the biosimilars inside other products grew to $258M from $151M. The OPDIVO biosimilar application is under FDA review with action expected by year-end, and the KEYTRUDA biosimilar results have been submitted, with the company waiting for FDA to accept the filing.

Assessment: PAVBLU's 18% price decline suggests the first-mover premium is now being competed away, which is normal for a biosimilar. The OPDIVO and KEYTRUDA filings matter more. They are the largest biologics Amgen has targeted, and first-wave entry is the only position in biosimilars that earns an attractive return.

Grading Last Quarter's Commitments

We grade twelve commitments carried from the Q2 release, call and note. The scorecard:

Net: four delivered or beaten, one partly delivered, five not delivered or unresolved, two deteriorated. The split from August holds and has sharpened. Everything inside Amgen's own laboratories and sales force went well. Both readouts on the list that arrived, dazodalibep and TEZSPIRE, came in positive, and DeLLphi-305, which was not on the list, was the best result of all. Everything that sits with a regulator, a court or the obesity Phase 3 program is where it was or worse. The difference this quarter is that the controllable side now includes new, large growth vectors, not only execution on existing ones.

One event that was not on our list moved the stock more than most that were. A competitor's Lp(a) outcomes trial failed on September 4, and AMGN fell 10.1% on September 8 on the read-across to olpasiran, then reached a period-low close of $375.65 on September 15. We had flagged that risk in the Q2 Q&A section but had not carried it forward as a commitment to grade.

Key Topics & Management Commentary

Overall Management Tone: Management was the most assertive it has been in the four quarters we have covered, and on the oncology data the confidence was earned. The assertiveness was concentrated in the pipeline. The in-line base was discussed mostly through its winners, with no word on the denosumab acceleration or Otezla, and TEPEZZA's decline explained only as an inventory comparison. On MariTide the answers stayed in process language for a third straight quarter, and the CEO again asked analysts to hold to one question each.

1. The growth engine is now big enough to carry a steeper cliff

The six growth drivers added $1,518M of sales year over year, against a $758M loss on Prolia and XGEVA, while everything else in the portfolio was flat in aggregate. Excluding denosumab, product sales grew 20%. The drivers reached $7,103M, or 72% of product sales, up from roughly 70% in Q2. All six grew at double digits. Total product sales still grew only 8%, because the denosumab decline took about eight points off the total.

"Our 6 key growth drivers grew 27% year-over-year and represented over 70% of our third quarter product sales." — Robert Bradway, Chairman & CEO

Assessment: This is the arithmetic that matters for 2027. The denosumab base that can still erode is now $920M a quarter, while the growth base compounding at 27% is $7.1B a quarter. Even if denosumab keeps falling at 45%, its dollar drag shrinks each quarter while the drivers' dollar contribution grows. On these numbers the cliff's dollar drag should be smaller in 2027 than in 2026, even at a 45% rate of decline, and that is the strongest structural argument for the multiple the market now pays.

2. Denosumab: 33% becomes 45%, and the call did not address it

Combined Prolia and XGEVA sales of $920M were down 45%, against declines of 33% in Q2 and 32% in Q1, and down 17% from Q2 in dollars. The US fell faster than international markets: Prolia fell 49% in the US against 33% abroad, and XGEVA fell 52% against 35%. Price joined volume as a driver. For Prolia, lower net price and adverse changes to estimated sales deductions together accounted for about as much of the decline as the 23% volume loss.

"Prolia (denosumab) sales decreased 45% year-over-year to $632 million in the third quarter as multiple biosimilars have launched globally." — Amgen, Q3 2026 earnings release

The Street had expected a step-down, to roughly 40%. The surprise was the extra five points: Prolia's $632M was $54M to $65M below published estimates of $686M to $697M. In Q2, management described the erosion as "in line with our expectations." This quarter the prepared remarks did not mention either brand, and none of the twelve analysts asked.

Assessment: We were wrong to read Q2's stability as a run-rate. A second leg down, driven by both price and volume, is the normal pattern once several biosimilars compete for formulary position, and it means the 2026 drag is larger than the 32% to 33% we carried into this quarter. It does not break the thesis, because of the arithmetic in topic 1. It does move this bear point from "on plan" back to "worse than modeled." Management's silence on its largest declining franchise, in a quarter it went from 33% to 45%, is the most notable omission of the call.

3. The guidance raise beats the revenue beat again, but the margin math puts the spending in Q4

Amgen raised 2026 revenue to $38.9B to $39.7B from $38.2B to $39.4B, a $0.5B midpoint increase with the low end up $0.7B, and non-GAAP EPS to $22.90 to $23.70 from $22.30 to $23.50, a $0.40 midpoint increase. The revenue raise again exceeds the quarter's revenue beat of roughly $0.3B. The EPS raise does not match the EPS beat of roughly $0.66. The operating-margin range of roughly 45% to 46% was reaffirmed, and the non-GAAP other income and expense guide improved by $0.1B to $2.0B to $2.1B of expense.

"As we look to the fourth quarter, we expect a meaningful sequential increase in operating expenses, consistent with our established pattern of higher spending in the final quarter of the year." — Thomas Dittrich, EVP & CFO

The full-year range leaves Q4 revenue of $9.86B to $10.66B, flat to up 8.1% against Q4'25's $9,866M, and Q4 non-GAAP EPS of $5.02 to $5.82 against $5.29. At the revenue midpoint, a 45% to 46% full-year margin implies a Q4 margin of about 39% to 43%, against 47.3% in Q3 and 42.8% in Q4'25. The CFO named the sources: a higher cost-of-sales ratio from mix and profit share, and higher SG&A as the Q3 timing reverses.

Assessment: The revenue guide was lifted by more than the beat for a second quarter, which says management genuinely raised its view of the year's sales. The EPS guide was not lifted by the full beat, which says part of Q3's earnings upside was borrowed from Q4. That is consistent with the SG&A timing and with Amgen's history: Q4'25's margin was 42.8%. Anyone annualizing Q3's 47.3% is again extrapolating the wrong quarter.

4. The new CFO's first forward signal: no 2027 margin step-down expected

On the first call in the role, the new CFO raised the 2026 guide and added a statement about next year: the company does not anticipate a step-down in the 2027 operating margin from the 2026 guidance. With the 2026 range at roughly 45% to 46% of product sales, that points to a similar margin next year while R&D spending keeps rising and MariTide launch preparation continues. It is an expectation, not formal guidance.

"Looking ahead to 2027, we do not anticipate a step-down in operating margin compared with our 2026 guidance." — Thomas Dittrich, EVP & CFO

Assessment: This is the most useful financial disclosure of the quarter. If it holds, 2027 EPS should grow roughly in line with product sales, plus whatever comes from lower interest expense, which the debt paydown supports. It also means absorbing the Otezla price-setting cut, continued denosumab erosion and late-stage trial spending inside a stable margin. A new CFO who puts that expectation on the record on the first call is either confident in the plan or setting a bar that will have to be defended. We think the first, because the growth mix supports it, but no analyst tested it.

5. IMDELLTRA: DeLLphi-305 changes the size of the franchise

At ESMO in October, the Phase 3 DeLLphi-305 study showed that adding IMDELLTRA to IMFINZI as first-line maintenance in extensive-stage small cell lung cancer extended median overall survival to 21.1 months from 12.7 months. That is a 58% reduction in the risk of death, with a 66% reduction in the risk of progression or death, at a median follow-up of 15.9 months. The results were published in the New England Journal of Medicine. The study permitted one to two hours of monitoring, and grade 3 cytokine release syndrome occurred in one patient. A Phase 1b first-line maintenance study showed a median overall survival of 37.9 months at a median follow-up of 30.8 months.

"If approved in first-line maintenance, IMDELLTRA could address the needs of 23,000 patients in the U.S. alone, importantly used earlier in the treatment journey and for a longer duration of therapy." — James Bradner, EVP, R&D

Management is preparing regulatory filings worldwide and pursuing NCCN guideline inclusion. It declined to size the opportunity in dollars but described the US small cell lung cancer population addressable in earlier lines as 35,000 patients.

Assessment: This is the single most valuable clinical result Amgen has reported in the four quarters we have covered, and unlike MariTide it now rests on Phase 3 survival data. Second-line use is capped by how few patients reach second line. First-line maintenance roughly multiplies both the number of patients and the duration of therapy. A drug already annualizing above $1.6B on the smaller population, with a lighter monitoring label, now has a credible path to a much larger franchise. It is still a filing, not an approval, and the survival data are immature.

6. Dazodalibep: a positive Phase 3 in a disease with no approved disease-modifying therapy, with the harder study still to read

OASIZ-301 met its primary endpoint in systemic Sjögren's disease. The ESSDAI improved by 5.6 points with monthly dosing and 5.2 points with every-twelve-week dosing, or 1.8 and 1.4 points better than placebo, with benefit visible by week 4 and sustained to week 48. Monthly dosing also met the patient-reported dryness endpoints. Full data come at ACR on November 11. The second Phase 3, OASIZ-303 in 434 patients with high symptom burden and low systemic activity, is expected to complete in Q4.

"It is hard to read through, though, from 301 into 303, the systemic to the symptomatic. These are different patient populations." — James Bradner, EVP, R&D

Assessment: A positive Phase 3 in a disease with no approved disease-modifying therapy is a real asset, and the quarterly arm hitting its primary endpoint supports a convenient dosing story. The placebo-adjusted effect is modest in absolute terms, because the placebo arm improved by about 3.8 points, and management itself declined to read across to the symptomatic study. We treat dazodalibep as a probability-weighted new launch, with OASIZ-303 and the competitive field deciding its size.

7. TEZSPIRE: a 43% grower adds a third indication

The Phase 3 CROSSING study in eosinophilic esophagitis met both co-primary endpoints. Histologic remission at week 24 reached 64.8% and 68.9% at the two doses against 3.5% on placebo, with benefit sustained through week 52, which management called a first for a biologic in a placebo-controlled EoE trial. Swallowing symptoms improved by 18.8 and 22.0 points against 14.7 on placebo. Two Phase 3 studies in COPD are enrolling.

"This is the first time sustained efficacy through 52 weeks has been shown for a biologic in a placebo-controlled EoE trial." — James Bradner, EVP, R&D

Assessment: The histology result is large. The symptom result is smaller, at 4 to 7 points better than placebo, and symptoms are what patients and gastroenterologists will weigh. EoE affects more than 470,000 people in the US, and every-four-week dosing is a practical advantage. This adds a third indication to a brand already growing 43% in asthma, and it raises the value of the AstraZeneca collaboration even with the profit-share drag on gross margin.

8. Repatha: guideline support and a mortality analysis as the oral competitor launches

Repatha grew 37% to $1,085M and 51% in the US. The ACC and AHA guidelines updated in March support earlier, more intensive LDL lowering and the use of PCSK9 inhibitors without prior ezetimibe. At ESC in August, a pre-specified VESALIUS-CV analysis showed nominal reductions of 20% in all-cause mortality and 21% in cardiovascular death, and a second pre-specified analysis showed a 27% reduction in three-point MACE.

"The new guidelines, supported by Repatha's differentiated outcomes data, strengthen clinical conviction that Repatha should be the first choice when additional LDL-cholesterol lowering is needed beyond statins." — Murdo Gordon, EVP, Global Commercial Operations

Assessment: A guideline that removes the ezetimibe step widens the funnel at the point of prescribing, and the mortality analysis, though nominal, is the kind of data that helps with payers. The oral PCSK9 competitor did not come up once on the call, from either side. At 51% US growth that silence is affordable this year, and the competitor's own launch commentary points the same way: Merck reported more than 11,000 prescriptions in its first quarter on the market, mostly from patients not switching off injectables, and expects broad commercial coverage only in 2027 and Medicare coverage in 2028. The 2027 access negotiations, where an oral option gives formularies leverage on rebates, remain the thing to watch.

9. Olpasiran: the Lp(a) class failed its first outcomes test, and the call did not mention it

On September 4, Novartis reported that pelacarsen's Phase 3 Lp(a)HORIZON trial in 8,323 patients with established cardiovascular disease missed its primary endpoint of reducing cardiovascular events, even though it lowered Lp(a). Amgen's shares fell 10.1% on September 8, the first session after the holiday, as the market re-rated olpasiran, Amgen's own Lp(a) drug, now in the Phase 3 OCEAN(a)-Outcomes study. At a September investor conference, management argued that olpasiran differs on potency, at about 95% Lp(a) reduction against 70% to 80% for pelacarsen, and on trial design, with a higher Lp(a) entry threshold and a three-point MACE endpoint. The full pelacarsen data had not been presented by the time of the call.

"Olpasiran is a potentially best-in-class small interfering ribonucleic acid (siRNA) molecule that reduces lipoprotein(a) (Lp(a)) synthesis in the liver." — Amgen, Q3 2026 earnings release

The release lists all three OCEAN(a) studies as ongoing or enrolling. On the call, olpasiran was not mentioned by management or by any of the twelve analysts.

Assessment: In August we wrote that a competitor failure would raise questions about the target itself, and that management had framed the readout so that only a competitor success was discussed. The failure came, and it cost the stock 10% in a session. Management's points on potency and design are reasonable, but they are arguments of degree about a target whose first outcomes trial just failed. Olpasiran was never one of our pillars, so the thesis impact is small. The stock recovered the September loss on other news within weeks, which says the market has set the question aside. The call did nothing to answer it.

10. UPLIZNA's indication ladder adds a rung

UPLIZNA grew 191% to $451M, annualizing above $1.8B. Amgen started the Phase 3 MOMENT study in primary membranous nephropathy and the Phase 2/3 MERCURY study in autoimmune hepatitis, with a CIDP Phase 3 planned for the second half of 2026 into the first half of 2027. One-year open-label data from the MINT study in gMG showed sustained numerical improvement, and the MITIGATE extension in IgG4-related disease showed all continuing patients flare-free.

"UPLIZNA offers a highly differentiated clinical profile, combining outstanding efficacy that resonates with physicians and convenient twice-yearly dosing that benefits patients." — Murdo Gordon, EVP, Global Commercial Operations

Assessment: We called UPLIZNA the most underrated growth vector in the company in August, and a 35% sequential gain since then supports that. Membranous nephropathy is an antibody-mediated kidney disease, so the ladder now spans neurology, rheumatology, hepatology and nephrology. It is also the whole of rare disease's growth this quarter, which makes the brand's execution a concentration risk as well as an upside.

11. MariTide: enrollment milestones, one T2D start, and a third quarter without a tolerability read

MariTide's Phase 3 program completed enrollment in the cardiovascular outcomes study and both sleep-apnea studies, started the first of three Phase 3 studies in type 2 diabetes, MARITIME-T2D-3 against semaglutide, and completed enrollment in a Phase 2b liver-fat study. Two more T2D studies start in Q4, so one of the three promised 2026 starts landed in Q3. Management again described 2026 as a year of data generation, which puts the first pivotal readout no earlier than 2027.

"We'll have more to share around the efficacy and tolerability of MariTide at the time that these studies read out." — James Bradner, EVP, R&D

Asked whether it has access to blinded Phase 3 tolerability data, management did not say. Separately, management said early-phase symptom diaries show side effects that were mostly mild, sometimes moderate, and lasted a few days despite the drug's long half-life.

Assessment: Nothing in this quarter changes the probability distribution on the readout that matters. The symptom-duration point is reassuring on mechanism and comes from early-phase data. Sponsors commonly decline to say whether they see blinded aggregate safety data, so we do not read the non-answer as negative. But a third quarter without any characterization means the pillar remains entirely unproven, while the rest of the pipeline now has hard data. MariTide matters less to the stock than it did in August, because the stock now has other reasons to exist.

12. Depth behind the headlines: daxdilimab, xaluritamig and two large biosimilar filings

Daxdilimab, a plasmacytoid dendritic cell depleter, met its Phase 2 primary endpoint in discoid lupus with 6.0 and 5.7 point placebo-adjusted CLASI-A improvements, and Amgen is advancing it to Phase 3. Xaluritamig, the STEAP1 T-cell engager, has two Phase 3 studies in prostate cancer and showed feasibility in a fully outpatient cohort. The OPDIVO biosimilar has an FDA action expected by year-end, and the KEYTRUDA biosimilar data were submitted. TEPEZZA met both primary endpoints in a Japanese chronic thyroid eye disease study, and the Nplate PROCLAIM study in chemotherapy-induced thrombocytopenia is complete.

"But it's not lost on us that we've stumbled upon a very exciting and hopeful mechanism targeting these plasmacytoid dendritic cells in this disease." — James Bradner, EVP, R&D

Assessment: One quarter produced three major positive Phase 3 readouts, a positive Japanese TEPEZZA Phase 3 and a Phase 2 that goes to Phase 3. The immunology bench behind UPLIZNA and TEZSPIRE now includes dazodalibep and daxdilimab, which makes inflammation a second growth franchise rather than a legacy one. Nplate's PROCLAIM result was called complete rather than positive, and its full results were deferred to a congress, so we treat it as unknown.

13. The balance sheet keeps de-levering, and the CEO questions the M&A premise

Amgen repaid $3.05B of debt in the quarter, the $1.8B term loan and $1.25B of 2.60% notes. Debt fell to $54.0B from $57.3B at Q2. Cash was $13.2B, so net debt fell to about $40.8B from $43.3B in the quarter and $45.5B at year-end. Equity rose to $14.4B from $8.7B at year-end, as nine-month net income of $8.1B exceeded $4.1B of dividends, no stock was repurchased and other equity items added the rest. The BeOne stake, still about 17%, was valued at $6.9B. The $6.8B repurchase authorization was again untouched. Capex was $587M, including capacity for a potential MariTide launch.

"Obviously, a meaningful portion of our revenues come from molecules that we license or acquired, and I think that will continue to be true through time." — Robert Bradway, Chairman & CEO

Assessment: Using free cash flow to retire debt rather than buy back stock at 20 times earnings is the right choice with an unquantified tax exposure outstanding. It also supports 2027 EPS through lower interest expense. On M&A, the CEO did not accept that Amgen has been sitting out the cycle and gave a defense of discipline, and with several new late-stage programs to fund, internal opportunities now compete for the same capital. We read it as continuity: bolt-ons yes, transformational deals no.

14. The overhangs widened in the filings, again

The Q3 10-Q adds two developments. In September the IRS issued a draft notice asserting penalties for 2016 to 2018, "substantially similar" to those asserted for 2013 to 2015, which were about $2.0B. The company says the 2016 to 2018 adjustments and penalties, if sustained in full, "could have a material impact," and the Tax Court decision on 2010 to 2015 is still expected no earlier than late 2026 or early 2027. On TAVNEOS, the European Commission revoked the EU authorization in August, CDER has not responded to the July submission, and the US intangible still carries $2.2B on the balance sheet. Separately, a July 31 Form 8-K disclosed a material cybersecurity incident in which data, including patient protected health information, was exfiltrated from third-party cloud environments.

"If sustained in full, these proposed adjustments and penalties could have a material impact on our financial statements." — Amgen, Q3 2026 Form 10-Q

Assessment: None of the three came up on the call. The tax exposure has now grown in each of the last three quarters: a new 2016 to 2018 notice at Q1, a new audit window at Q2 and penalties at Q3. A TAVNEOS withdrawal would mean an impairment of up to $2.2B, which is non-cash and excluded from non-GAAP results but real for book equity. The cyber incident is the one item our own August note did not cover, although the 8-K preceded that call. The company said at the time it was not reasonably likely to have a material financial impact, and no amendment has followed, but the scope of the patient-data exposure has not been described in any filing or on either call since.

Guidance & Outlook

The pattern from August repeats on both lines, and the earnings half is weaker. The revenue midpoint rose $0.5B against a beat of roughly $0.3B, and the low end rose $0.7B, so management raised its view of fourth-quarter sales rather than flowing the quarter through. The EPS midpoint rose $0.40 against a beat of roughly $0.66. In August it rose $0.50 against a beat of roughly $0.68, so the share of the beat carried into the guide fell from about 74% to about 61%. Less flowed through this time because the operating-margin range was left at roughly 45% to 46% after a 47.1% nine months. The tax range was untouched, so none of the raise is a tax artifact.

Implied Q4: Nine-month revenue of $29,037M leaves $9.86B to $10.66B for Q4, flat to up 8.1% on Q4'25's $9,866M, with a midpoint of 4.0%. Nine-month non-GAAP EPS of $17.88 leaves $5.02 to $5.82, against $5.29 a year ago. At the revenue midpoint, the margin range implies a Q4 non-GAAP operating margin of about 39% to 43%. Full-year revenue growth at the new range is 5.8% to 8.0%, against about 5.6% at the August midpoint.

Street at: The Street entered the print at about $22.96 to $23.02 for 2026 EPS and about $38.9B for revenue. The new revenue midpoint of $39.3B sits about $0.4B above that. The new EPS midpoint of $23.30 sits about $0.30 above the old consensus, but Q3 alone beat by $0.66, so a Street that kept its fourth-quarter estimate would land near the top of the new range. One published fourth-quarter EPS consensus is $5.79, against the $5.42 the guide implies at its midpoint. Fourteen firms raised price targets the next day and none changed ratings, though the published average target remains well below the share price.

Guidance style: Classic Amgen, and the new CFO kept it that way. The company raised early, left the margin range alone and told investors in advance that Q4 spending will be heavier. The one new element is a forward margin statement for 2027, given before the formal 2027 guide with Q4 results. That shortens the period in which the Street has to guess at next year's cost base.

Analyst Q&A Highlights

Whether management can see blinded MariTide tolerability data

A two-part question first asked whether Amgen might shrink its biosimilar business to free manufacturing capacity for a MariTide launch, then asked directly whether R&D has access to blinded tolerability data from the ongoing Phase 3 studies. Management dismissed the capacity idea, saying the expansion is sized for much larger protein volumes. On the second part, it answered with a statement about trial conduct rather than a yes or no.

Q: "And then the kind of related question is, can you tell us, Jay, if you guys have access to the blinded tolerability data from the ongoing MariTide Phase III trials?"
— Terence Flynn, Morgan Stanley

A: "We are, in 2026, just noses down with disciplined data generation on MariTide, and the MariTide studies are being carried out with a total adherence to good clinical practices. We'll have more to share around the efficacy and tolerability of MariTide at the time that these studies read out."
— James Bradner, EVP, R&D

Assessment: The question was built to get either a characterization or an admission that none is possible, and management gave neither. The reference to good clinical practice reads as a signal that the company will not comment on blinded data at all, which is defensible. It also signals there will be no tolerability information before the readouts. This is the third consecutive quarter in which a direct tolerability question has gone unanswered.

Why MariTide's side effects may not last as long as the drug does

A question followed up on a September small-group discussion, in which management had suggested that MariTide's side effects last about as long as those of weekly incretins despite monthly dosing. It asked how that is measured and how MariTide fits in a fast-moving obesity market. The answer was the most substantive MariTide content of the call.

Q: "I think, Jay, when we met in the small group meeting back in September, you suggested that although MariTide is administered monthly or even less frequently, the side effect duration is comparable in length to what you get with weekly incretins. I would love it if you expand more on how you can measure that in kind of a Phase 1 of any of your trials?"
— Evan Seigerman, BMO Capital Markets

A: "Here, we've done very exhaustive diaries in early phase clinical investigation to understand the onset, frequency, severity and duration of symptoms. We've shared some of these early plots that we call Manhattan plots that reveal the duration of symptoms should they arise. To date, MariTide has proven very well tolerated its target doses and like all GLP-1s benefits from dose escalation."
— James Bradner, EVP, R&D

Assessment: The dissociation between a long half-life and short-lived symptoms is the right mechanism to emphasize, because the fear with a monthly incretin is a month of nausea after a bad dose. The evidence cited is early-phase diary data, not Phase 3. It is a coherent argument, and it is still an argument rather than a result.

How large IMDELLTRA can become in first-line maintenance

A question pressed for a patient and dollar frame after the ESMO data, offering a specific scenario of more than 20,000 US patients and more than $5B in first-line maintenance alone. Management declined to give product guidance but supplied the population figure that bounds the opportunity and emphasized duration of therapy.

Q: "Is it possible or is it realistic for us to assume something approaching 20,000-plus patients in U.S. being on tarlatamab? And do you envision the maintenance opportunity in first-line alone being $5 billion plus with induction trial next year adding additional upside?"
— Umer Raffat, Evercore ISI

A: "And so, as you move into the earlier lines, the small cell lung cancer addressable population in the U.S. is 35,000. So, it's a decent size patient population. The question is, how big is it in each setting and frontline maintenance." [...] "And so the duration of treatment here might be quite long as well."
— Murdo Gordon, EVP, Global Commercial Operations

Assessment: Management did not push back on the scenario, and "very rapid adoption" is strong language from a company that does not give product guidance. The bounding fact is useful: 35,000 addressable US patients in earlier lines, of whom R&D sized first-line maintenance at 23,000. Duration is the variable that converts patients into revenue, and the immature survival curve is what will set it.

The speed of the first-line ramp and the role of monitoring

A question asked how quickly IMDELLTRA can ramp after the ESMO data, and how much the monitoring requirement had held back second-line uptake. Management described the label change as already helping community oncology clinics and said it hopes the first-line label will require even less monitoring.

Q: "As part of that answer, can you just talk about the reduced monitoring requirements for the drug as well? I'm trying to get a sense of how much of a hurdle monitoring had been for uptake in the later line settings, that obviously seems to be addressed in this earlier line setting as well."
— Christopher Schott, JPMorgan

A: "And I would expect that with things like updated NCCN guidelines with additional awareness of these data, the behavior change in the clinic will be rapid. Obviously, we're moving as quick as we can to ensure regulatory filings and approvals and all the rest of it, but we are ready to support the rate of change that clinicians will effect."
— Murdo Gordon, EVP, Global Commercial Operations

Assessment: The sequence management describes is guideline first, label second, and that order matters, because NCCN inclusion can drive use before an approval. The one-to-two-hour monitoring in DeLLphi-305 is what makes first-line maintenance a community-practice drug rather than an academic-center drug. This exchange and the previous one are the clearest evidence that management sees IMDELLTRA as a next-year growth driver, not a 2028 one.

What the systemic Sjögren's result says about the symptomatic study

The call opened with a question asking how to handicap the second dazodalibep Phase 3 and what the regulatory path looks like. Management was careful to separate the two studies and gave nothing on the filing strategy.

Q: "On the back of the positive Phase III results in systemic Sjogren's disease, could you help us think about the confidence or kind of handicap how we should think about the readout here in the symptomatic patients and then speak to the regulatory path for this program overall?"
— Salveen Richter, Goldman Sachs

A: "It is hard to read through, though, from 301 into 303, the systemic to the symptomatic. These are different patient populations. They are being studied separately. And this is important to tease out the possibility of impact on these 434 patients in the 303 study. So we'll have more to say shortly after we see data in Q4 of this year."
— James Bradner, EVP, R&D

Assessment: Declining to read across is scientifically correct, and it also lowers the bar before a binary readout due within weeks. Low-systemic-activity, high-symptom Sjögren's is the harder population to show a benefit in, so OASIZ-303 decides whether dazodalibep is a niche systemic drug or a broad one. We would not assume the second study repeats the first.

Speed of onset as a differentiator for dazodalibep

A question noted that the onset of effect in the systemic study appeared faster than in Phase 2 and asked how that reads across. Management tied the week-4 onset to the mechanism and contrasted it with the delayed onset that B-cell depletion can show.

Q: "Can you talk about what we are seeing on onset of effect with the systemic trial? And how that reads across to the symptomatic study that's coming up?"
— Akash Tewari, Jefferies

A: "So to see effect at week 4 is highly consistent with our mechanistic thesis, that as an inhibitor of CD40 ligand, CD40 signaling, you have this chance for an immediate inhibition on B-cell-driven inflammation."
— James Bradner, EVP, R&D

Assessment: Fast onset is a commercial argument in a crowded immunology field, and management framed it against B-cell depletion, a class in which Amgen itself sells UPLIZNA. The same answer called the placebo-adjusted effect "in line with our expectations," which is a more measured description than the word "surprised" in the question.

Whether Amgen is sitting out a record M&A cycle

A question asked why Amgen has been quiet during a period of record industry deal-making and whether investors should still expect bolt-ons. The CEO declined to accept the premise and restated the criteria.

Q: "Bob, Amgen has been quieter, I guess, on the M&A front in a period of record M&A deal making. Could you just comment about your perspective about what you think is going on?"
— Michael Yee, UBS

A: "As we've said for some time, we're looking for opportunities that align well with our 4 therapeutic areas of focus. We're focused on delivering consistent long-term growth, and we feel confident we're on track to do that through time." [...] "And I'd like to think when -- particularly when we go externally that we see a path for our shareholders to earn a return and not just for the selling shareholders to earn a return."
— Robert Bradway, Chairman & CEO

Assessment: The line about selling shareholders is a pointed comment on industry deal pricing, and it tells investors Amgen is not going to pay up for assets in this market. With net debt still about $40.8B and a pipeline that just produced three Phase 3 wins, internal investment is the better use of capital anyway. The answer removes a source of upside surprise and a source of downside risk at the same time.

Timing for the OPDIVO and KEYTRUDA biosimilars

A question asked when the two filed oncology biosimilars could launch and how they compete as the reference products move to subcutaneous formulations. Management gave no timing and pointed to its first-wave track record.

Q: "With both the OPDIVO and KEYTRUDA BLA filed, just curious what your assumptions are around being able to -- when you can commercially launch those? And how do you think about penetrating the market that is kind of being moved a subcutaneous formulation?"
— Matthew Phipps, William Blair

A: "Well, as we always say about our focus on biosimilars will be a first or first wave with those launches, which is really important to establish a decent competitive position, and we believe that we're there."
— Murdo Gordon, EVP, Global Commercial Operations

Assessment: The subcutaneous-conversion half of the question went unanswered, and it is the real risk to these biosimilars: if the originators move most patients to new formulations before entry, the addressable intravenous market shrinks. First-wave positioning protects the price, not the volume. We give these launches option value, not base-case revenue.

What They're NOT Saying

1. Anything about denosumab. The largest declining franchise went from a 33% decline to 45% and came in below the Street's estimates, and neither the prepared remarks nor any of the twelve questions touched it. In Q2 management at least said the erosion was in line with its expectations. This quarter it gave no comparison with plan, no exit rate and no 2027 view.
2. Olpasiran, after its class setback. A competitor's Lp(a) outcomes failure took 10.1% off the stock in one session in September, and Amgen's own Lp(a) drug was not mentioned once on the call. Management gave no update on OCEAN(a)-Outcomes event accrual or timing, and no analyst asked.
3. The size of the SG&A timing shift. SG&A fell 1% "mainly driven by the timing of investments," and the CFO said the spend comes back in Q4. The amount that moved was not given. At Q2's growth rate it would have been about $93M, close to a point of margin.
4. The fourth-quarter margin the guide implies. Reaffirming roughly 45% to 46% after a 47.1% nine months implies a Q4 margin near 39% to 43%. Management described the direction but did not give the number, and no analyst asked how a 47% run-rate becomes a low-40s quarter.
5. The new IRS penalty notice. The September draft notice asserting 2016 to 2018 penalties appears only in the Q3 10-Q, not in the release. It was not mentioned on the call and not asked about, for a second straight quarter of silence on the tax dispute after the CFO addressed it in Q1.
6. TAVNEOS outside the US. The release keeps the US withdrawal language. The 10-Q adds that the European Commission revoked the EU authorization in August and that other regulators have taken or announced actions. International TAVNEOS sales were zero in the quarter. None of this was discussed.
7. The July cybersecurity incident. The July 31 Form 8-K reported exfiltration of data that included patient protected health information and called the incident material. A text search of the Q3 10-Q finds no reference to it, no amendment to the 8-K had been filed as of November 4, and neither call has addressed it. Our own August note also missed it.
8. The dryness result for quarterly dazodalibep. Both doses met the primary endpoint, but the release and the call report the dryness benefit only for monthly dosing. When asked whether both regimens can be filed, management declined to discuss the regulatory strategy between readouts.
9. The oral PCSK9 competitor. In August it drew a direct two-part question. This quarter it was not mentioned by management or by any analyst, and there was still no Repatha net-price disclosure.
10. Why Enbrel's deductions moved. A 15-point favorable change in estimated sales deductions turned a price-driven decline into 5% growth. The release gives the size but not the reason, or whether it recurs.
11. Who the business-development counterparties are. Q3 R&D included business-development activity, and contingent consideration more than doubled to $367M from $161M at year-end on "other business development activity in 2025 and 2026." The assets and partners were not named on the call.

Market Reaction

• Pre-print setup: AMGN closed at $461.26 on 2026-11-03 before the after-market release, up 40.9% year to date against the S&P 500's 10.6%, up 55.7% over the trailing twelve months and up 14.4% over the trailing thirty days. The 52-week closing range was $296.30 to $466.01, so the stock entered the print just below its high. The options market implied a move of roughly 4.3% to 4.5%.
• After-hours move: Shares fell about 1.4% to 2.4% after the release, trading around $450 to $455, as the first read focused on the Prolia and XGEVA declines and on a guidance raise seen as roughly in line with expectations.
• Next-day session: On 2026-11-04 the stock opened at $479.61, a 4.0% gap up, traded between $464.13 and $482.48, and closed at $478.48, up 3.7% or $17.22, on 4.5M shares against a 2.8M thirty-day average, or 1.6 times normal volume. The close set a new 52-week closing high.
• Relative: The S&P 500 rose 1.6% in a broad rally the day after the US midterm elections, with healthcare and biotech participating. AMGN outperformed the index by about 2.1 percentage points.

The overnight reversal is the informative part. The after-hours read was the headline read: denosumab worse than expected, and a raise smaller than the beat. By the next morning the call had been digested, and the call was about IMDELLTRA, dazodalibep and an expected stable 2027 margin, none of which is in a headline. A flat-to-down first reaction that turns into a gain after the call is the same two-stage pattern as August, and again it says the market is paying for what it learns from management, not from the release.

The size matters as much as the direction. The 3.7% gain was smaller than the move the options market implied, it came on a day when the index rose 1.6% for reasons unrelated to Amgen, and it compares with a 4.6% gain and 4.8 points of outperformance in August. The stock also entered the print up 13.1% from our August reference, and the path matters. It fell 10.1% on September 8 on the Lp(a) read-across, closed at a period low of $375.65 on September 15, then rose 22.8% to the pre-print close as the dazodalibep topline arrived and as large-cap healthcare reached record highs in an October rotation out of technology stocks. The full IMDELLTRA data at ESMO added little on the day. The market is still buying each beat-and-raise, but in smaller increments, because more of the good news is in the price before the release.

Street Perspective

Debate: Has the pipeline earned a growth multiple, or is the stock pricing approvals it does not have yet?

Bull view: The bull case on the Street is that Amgen now has three new late-stage growth vectors with Phase 3 data, seventeen billion-dollar products, and no margin step-down expected next year. A diversified biopharma growing at a mid-to-high single-digit rate with that pipeline deserves a premium to its history, and the multiple expansion is a correction rather than an overshoot.

Bear view: The bear camp argues that the forward multiple now depends on pipeline delivery. IMDELLTRA's first-line use is not yet approved, dazodalibep's second study has not read out, and the MariTide data that could move the stock most are a 2027 binary. The published average price target still sits below the share price.

Our take: Both are partly right, and the bull side has the better of it on fundamentals for the first time in our coverage. DeLLphi-305 is a survival result, not a promise, and it moves IMDELLTRA from a niche to a franchise. But 20.5 times this year's earnings already credits most of that, and the obesity binary is still unpriced in both directions. We raise our frame and stay below the price.

Debate: How clean was the beat?

Bull view: The optimistic read is that the beat was broad: Repatha, EVENITY, UPLIZNA and IMDELLTRA all came in ahead, product sales grew on volume, and the revenue raise exceeded the beat. Some desks see fourth-quarter EPS upside against a conservative guide.

Bear view: The skeptical read is that the quality was mixed. A meaningful share of the revenue upside came from Enbrel's deduction changes and inventory, part of the EPS upside came from spending timing, and the EPS raise was smaller than the beat. Prolia and XGEVA undershot.

Our take: The skeptics have the details right and the bulls have the direction right. Strip out Enbrel's true-up and the SG&A timing and Q3 still beat, by less. The fourth-quarter guide is conservative by Amgen's habit, and a Street that keeps its prior Q4 estimate lands near the top of the range, so modest upside to the EPS midpoint is the likely outcome.

Debate: Is the denosumab acceleration a step or a new slope?

Bull view: The bull case treats 45% as a one-time step as more biosimilars reached formularies, and stresses that the franchise is now under a tenth of product sales. The dollar drag shrinks from here even at the same rate, and the growth drivers add far more each quarter.

Bear view: The bear case notes that price has now joined volume as a driver, which is how biosimilar erosion compounds, and that the rest of the legacy base is also under pressure: Enbrel's net price fell 14%, Otezla fell 18% and enters Medicare price setting in 2027, and TEPEZZA fell in the US.

Our take: The bull side is right about 2027 and the bear side is right about this year. The arithmetic of a shrinking base is real, and it is why this does not break the thesis. But we had denosumab tracking to plan in August, and it did not. Until management gives an exit rate, we model the 45% pace into 2027 and treat any stabilization as upside.

Debate: Does the tax dispute matter before the Tax Court rules?

Bull view: The dismissive view is that the dispute is years old, contested, reserved for in management's judgment and not likely to be resolved within the next year, so adding one more notice changes nothing that the stock can act on.

Bear view: The concerned view is that the exposure has grown every quarter this year. The litigated years carry roughly $3.6B and $5.1B of asserted tax plus $2.0B of penalties, partly offset by repatriation tax paid, and the IRS has now proposed penalties on 2016 to 2018 in draft form, with 2019 to 2022 under audit, where the company believes the IRS may pursue similar issues.

Our take: The bear side is right that the direction of travel is one-way, and right that the market has stopped asking. The company expects a decision no earlier than late 2026 or early 2027. An adverse ruling would reprice every later year at once, and at 20.5 times earnings there is less room to absorb it than in May. We treat it as the main downside catalyst on the calendar.

Model Update Needed

Valuation impact: At the reaction-day close of $478.48, AMGN trades at 20.5 times the raised 2026 non-GAAP EPS midpoint of $23.30, against 17.8 times the August midpoint at our Q2 reference price of $407.83. On the same 2026 basis the stock is up 17.3% while the midpoint rose 1.7%, so the move is again almost entirely multiple expansion. Unlike in August, part of it is earned: three Phase 3 programs read out positive and the CFO said no 2027 margin step-down is anticipated. We now set the frame on 2027, which starts in two months. Our working 2027 EPS is about $25, from roughly 7% growth at the flat margin the CFO said is expected. At 17 to 18 times that, the frame is $425 to $450. On the 2026 midpoint that is 18.2 to 19.3 times, against 17 to 17.7 times in August. The top of the new frame is about 10% above the August reaction close, a level we judged at or above fair value at the time. The step-up reflects the new Phase 3 data and the margin outlook, net of a steeper denosumab decline and a larger tax exposure. The stock sits 6% to 13% above the frame and at about 19.1 times our 2027 estimate.

Thesis Scorecard Post-Earnings

Scoring this quarter against the standing thesis carried since our initiation. Status tags reflect the direction of travel this quarter.

Overall: Thesis strengthened on the growth side and weakened on the loss-of-exclusivity and risk side, and the share price moved further than either. Bull #3 had its best quarter of our coverage, with a survival result that changes the size of IMDELLTRA and two more positive Phase 3 studies. Bear #1 reversed our August read, Bear #3 widened for the third quarter running, and the Lp(a) class failure put a question mark on olpasiran that management did not address. The net is a better business than we could argue in August, at a price that already reflects it.

Action: Maintain Hold, with conviction lowered to 5 from 6 because the stock now sits further above our frame than at any point in our coverage. Our August triggers stay on the record, and none fired. MariTide pivotal efficacy has not arrived, and the IRS and TAVNEOS overhangs both widened rather than resolved. The valuation trigger came closest but did not fire: the lowest close was $375.65 on September 15, 2.9% above the mid-$360s. That pullback also followed the Lp(a) class failure and reversed within weeks. We add two triggers, both outside our $425 to $450 frame. A pullback to $400 or below with no fundamental deterioration would restore enough cushion to reconsider an upgrade on valuation alone. A move above $525, about 21 times our 2027 estimate, without MariTide data, or an adverse Tax Court decision on 2010 to 2015, would move us to Underperform. If positive MariTide efficacy arrives in the same window as either downgrade trigger, we would treat the two as offsetting and reassess on the data.