Eliquis Rewrites the 2026 Guide While Underlying EPS Sits Flat and Two Flagship Readouts Slip to 2027

Results vs. Consensus

Q2 2026 Scorecard

Consensus figures are the pre-print sell-side average for revenue and adjusted diluted EPS. Gross margin, operating margin and the cash-flow lines are not consistently consensus-tracked for this name, so no beat or miss is claimed on them. Guidance rows compare the July revision against the guide issued with the Q1 print in April.

Year-Over-Year Comparison

Operating margin follows the company's own definition: gross profit less SG&A and R&D, as a percentage of revenues, excluding specified items. Percentage changes above are computed from the dollar columns shown; the filing's own percentages are calculated from unrounded amounts and differ by up to one point in a few places.

Quarter-Over-Quarter Comparison

Q1 2026 figures are reconciled out of the six-month columns of the Q2 filing, so every comparative traces to a filed document. Each one cross-checks against the Q1 8-K exactly, with the single exception of Eliquis, where the six-month minus three-month arithmetic gives $4,136M against the $4,137M the Q1 release printed. The filing states that certain totals may not sum due to rounding, and the Q1 printed figure is used above.

Quality of Beat

Revenue assessment. A 10.4% revenue beat at a company this size is not an execution surprise, it is a modelling gap, and the gap belonged to management before it belonged to the Street. The consensus figure of $11.75B implied a 4.2% year-over-year decline, which was the arithmetic of BMS's own April guide of $46.0B to $47.5B against FY2025's $48,194M. Both the Street and the company were understating Eliquis by roughly 1,000bps of growth. The right way to read this print is not that Bristol Myers outperformed by 10%, but that the company reset a guide it had set too low three months earlier on a product whose 2026 pricing dynamics it should have understood better than anyone. That reflects worse on the April guide than it reflects well on the June quarter.

Margin assessment. Two consecutive quarters above the gross-margin range, with the range left unchanged, is now a pattern rather than a rounding difference. The full-year math is stark: at the 69.5% midpoint on $49.5B of revenue, full-year gross profit is about $34,403M; the first half already delivered $17,344M; the implied second half is roughly 68.1%, which is 280bps below the run-rate just achieved and 90bps below even the bottom of the guided range. Either the mix compression management has been describing arrives all at once in the back half, or the guide is carrying a cushion that will show up as another raise. Nobody asked about it on the call, and management volunteered a single sentence: gross margin reflected product mix.

EPS assessment. The flat underlying EPS is the most important number in this release and it appears nowhere in the narrative. Management's prepared remarks characterized the quarter as "another strong quarter" and the EPS line got one sentence: diluted earnings per share were $2.04. The structural pressures we flagged in April are all still operating. The diabetes royalty stream that contributed $286M in Q2 2025 is gone and contributed nothing this quarter. Gross margin compressed 120bps. Operating expenses grew 4.8% against revenue growth of 5.7%, which is fine, but core operating profit grew only 3.2% and the below-the-line hole swallowed it. What changed is that Eliquis got large enough to make the revenue line look like a growth company while the profit line did not move.

Segment Performance

Bristol Myers reports as a single operating segment and discloses performance through the Growth Portfolio and Legacy Portfolio split. The tables below present the worldwide product detail as filed.

Portfolio Mix

Growth Portfolio by Product (worldwide)

Change and Ex-FX columns are the percentages as filed, which the company calculates from unrounded amounts. Recomputing from the rounded dollar columns shown gives the same figure in every row except Sotyktu and Cobenfy, where the difference is a single point of rounding. Other Growth Products comprises Abecma, Augtyro, Onureg, Inrebic, Nulojix, Empliciti and royalty revenues, including royalties received from Merck on Winrevair.

Legacy Portfolio by Product (worldwide)

Eliquis: the whole quarter, and a borrowed one

Eliquis worldwide revenue of $4,481M grew 22% reported and 21% Ex-FX, with the U.S. line up 27% to $3,357M. That single product is now 34.5% of company revenue, up from 30.0% a year ago, and its $801M of year-over-year growth exceeded the entire company's $704M increase. Management raised the full-year Eliquis growth guide from 10% to 15% up to 20% to 25%, a 1,000bps revision on the largest product in the portfolio one quarter after setting it.

"Eliquis revenue was approximately $4.5 billion, up 21%, driven by strong demand through continued market share gains. As a result of the price decrease implemented at the start of the year, U.S. second half sales are expected to benefit from the elimination of the accumulated CPI penalty in certain government channels." — David Elkins, EVP and Chief Financial Officer

Two things are happening at once and they should not be conflated. Demand is genuinely strong: new-to-brand share in the U.S. is approaching 80% per the Chief Commercialization Officer. But the second driver is a pricing mechanic. The January list-price reduction eliminated an accumulated inflation-rebate penalty in government channels, which lifts net realized price on the same volume. That is a one-time reset of the gross-to-net bridge, not a durable revenue line.

Assessment: The quarter's revenue upside is high quality in cash terms and low quality in duration. Management itself reiterated the $1.5B to $2.0B step-down expected in 2027 and flagged the European patent expiry in mid-Q4 of this year, before U.S. generic entry arrives in April 2028 under previously executed settlements. The 2026 guide raise is real money that lands in this year's cash flow. It is not a change in the shape of the cliff, only in when the descent begins.

Growth Portfolio: the milestone cleared, the coverage ratio slipped

The Growth Portfolio reached $7,560M, or 58.3% of company revenue, up from 54.2% in Q1 and 53.8% a year ago. Ten products delivered double-digit growth. On the metric that defines the equity story, this was the best quarter of the transition so far.

"Growth portfolio revenue rose 14% to $7.6 billion, now representing nearly 60% of total revenue. As Chris referenced, multiple products contributed to this growth as this portfolio continues to emerge as a broadening foundation for durable long-term growth, further supported by the broad pipeline." — David Elkins, EVP and Chief Financial Officer

The number that did not improve is the coverage ratio. In Q1 the Growth Portfolio added $664M against $932M of Legacy erosion outside Eliquis, a ratio of 0.71. In Q2 it added $964M against $1,052M of that erosion, a ratio of 0.92. The Growth Portfolio is growing faster in absolute dollars than it was three months ago, and faster than the ex-Eliquis Legacy book is shrinking faster still.

Assessment: The mix milestone matters for the 2029-and-beyond story and it was cleared decisively. It does not yet solve the 2027-and-2028 arithmetic, because the pace of Legacy erosion accelerated more than the Growth Portfolio did. Watch the coverage ratio rather than the mix percentage: the percentage is held down by Eliquis sitting inside the Legacy denominator and growing.

Opdivo and Qvantig: the franchise is flat once the inventory build is removed

Opdivo fell 3% to $2,485M while Opdivo Qvantig, the subcutaneous formulation, reached $261M from $30M. Taken together the franchise was $2,746M against $2,590M a year ago, up 6.0%. The company disclosed in its slide materials that Opdivo global sales reflect roughly $200M of sequential inventory build.

"For Opdivo, revenue decreased 4% to approximately $2.5 billion, primarily driven by continued conversion to Qvantig in the U.S. In the quarter, we saw some inventory build at the wholesale level, reversing some of the drawdown we referenced in the first quarter." — David Elkins, EVP and Chief Financial Officer

Strip the disclosed build and the franchise is roughly $2,546M, or about 1.7% below the year-ago quarter. The sequential recovery from Q1's $2,309M shrinks from 18.9% to roughly 10.3%. Qvantig is now 9.5% of franchise revenue and, per the CFO, holds 15% share against a 30% to 40% target, and is annualizing at over $1 billion.

Assessment: This is the cleanest instance of management doing what we asked for in April. The Q1 wholesaler drawdown was flagged then with weak language about monitoring whether levels normalize; this quarter the reversal was disclosed, quantified in the materials and named on the call. Credit where due. The underlying read is less encouraging: on a demand basis the second-largest franchise in the company is not growing, and the conversion to Qvantig is protecting the franchise rather than expanding it.

Reblozyl, Breyanzi and Camzyos: the three that carry the arc

Reblozyl grew 29% to $735M, Breyanzi 41% to $484M and Camzyos 60% to $416M. Together they added $463M of the Growth Portfolio's $964M increase, or roughly 48% of it, from a combined base of $1,172M.

"Camzyos revenues increased 59% to $416 million, reflecting continued promotional efforts, new patient prescriber additions and deeper penetration into the community setting." — David Elkins, EVP and Chief Financial Officer

Management flagged one near-term qualifier on Breyanzi, noting that as in prior years, typical summer demand patterns are anticipated to moderate growth in the third quarter. That is a helpful disclosure ahead of the print rather than after it. Camzyos has a supplemental application in adolescents with a September 30 PDUFA date and a Phase 3 study in non-obstructive hypertrophic cardiomyopathy expected to start by year end.

Assessment: These three are doing exactly what the thesis needs and are the reason the Growth Portfolio pillar stays intact. The concentration is worth naming: roughly half the Growth Portfolio's dollar increase comes from three products representing 21.6% of it. That is normal for a portfolio in transition and it is also a dependency.

Cobenfy: still small, still the dosing problem

Cobenfy grew 81% to $63M. On a base that small the growth rate is not informative; what matters is that the drug remains under $70M a quarter roughly two years into launch, in a schizophrenia market where the company has framed it as a category-defining asset.

"As far as dosing, we have been educating physicians on the importance of rapidly titrating Cobenfy to the effective dose of 125 milligrams as we see far too many patients that remain on either the 50 or 100-milligram dose." — Adam Lenkowsky, EVP and Chief Commercialization Officer

The leading indicators offered were a roughly 15% quarter-over-quarter increase in total prescriptions and consistent gains in new prescribers, against acknowledged weakness in new patient starts and repeat prescribing.

Assessment: Cobenfy in schizophrenia is not the investment case; Cobenfy in Alzheimer's psychosis is, and that readout just moved to 2027. The base business gives no reason to underwrite the label expansion early. Management's own framing, that prescribing behaviour in this disease state is deeply entrenched, is a multi-year argument dressed as a quarterly update.

Legacy Portfolio excluding Eliquis: down 52.8%

Outside Eliquis the Legacy Portfolio generated $941M against $1,993M a year ago. Revlimid fell 49% to $425M, Pomalyst and Imnovid fell 71% to $204M, Abraxane fell 47% and Sprycel fell 27%. U.S. generic lenalidomide stopped being volume-limited on January 31, 2026 and generic pomalidomide entered in March.

"Within the legacy portfolio, Eliquis delivered demand-driven growth of 21%, which nearly offset declines from generic entry across the remainder of the legacy portfolio. Despite full generic availability in the U.S., Revlimid revenue remains more resilient." — David Elkins, EVP and Chief Financial Officer

Assessment: The headline Legacy decline of 4.4%, and the newly disclosed full-year guide of a 4% to 6% decline, are Eliquis-net figures. The book that is actually going away is going away at better than 50% a year and is now down to $941M a quarter, which means the remaining absolute dollars at risk from this cohort are modest. That is genuinely good news, and it is why the guide could improve. It also means the next erosion event is Eliquis itself, which is nearly five times the size of everything left in the Legacy book combined.

Key Topics & Management Commentary

Overall Management Tone: Assured and commercially fluent on the P&L, and noticeably more forthcoming on inventory and channel dynamics than last quarter. The awkward material was handled by pre-announcing it rather than defending it: both readout delays were disclosed in prepared remarks and framed as favourable before any analyst could frame them otherwise. Where the tone thinned was the same place as in April, in the financial detail nobody volunteered, with gross margin getting one sentence, the flat underlying EPS getting none, and the rebate-timing promise from Q1 going unmentioned.

1. The guidance raise and what actually changed inside it

Management raised full-year revenue guidance from approximately $46.0B to $47.5B up to approximately $49.0B to $50.0B, and non-GAAP EPS from $6.05 to $6.35 up to $6.75 to $7.00. Gross margin, other income and expense, and the tax rate were all left unchanged. Operating expenses moved up $0.2B to approximately $16.5B.

"We are increasing our full year 2026 financial guidance for both revenue and adjusted diluted EPS based upon our strong first half results and our current projections for the balance of the year, while maintaining our projections for gross margin, other income and expense and our tax rate. Incorporated into these revisions are projected higher Eliquis revenue growth of 20% to 25% and a less pronounced total legacy portfolio revenue decrease now expected to be between 4% and 6%." — David Elkins, EVP and Chief Financial Officer

The sign flip is the story. Against FY2025 revenue of $48,194M, the April midpoint of $46.75B implied a 3.0% decline for the year. The July midpoint of $49.5B implies a 2.7% increase. On EPS, the April midpoint of $6.20 implied 0.8% growth on FY2025's $6.15; the July midpoint of $6.875 implies 11.8%.

Assessment: This is a genuine and large upgrade to the 2026 P&L and it should not be minimized. It is also entirely explained by two line items on the same product, Eliquis growth going from 10-to-15 up to 20-to-25 and the Legacy decline improving. Nothing in the raise reflects a change in the Growth Portfolio outlook or in the pipeline, and nothing in it extends past December.

2. Eliquis and the mechanics behind a 1,000bps guidance revision

The most consequential exchange on the call was about what happens to the 2026 Eliquis strength in 2027. Management held the line it drew in April rather than extending the good news.

"Now we still expect the $1.5 billion to $2 billion step down in 2027 that we referenced on the Q1 call. We'll clearly provide any appropriate update in conjunction with our 2027 guidance when issued. But the U.S. sales certainly would carry forward to next year, offset by the patent expiry of Eliquis in Europe in mid-Q4." — Adam Lenkowsky, EVP and Chief Commercialization Officer

Read carefully, that answer says the U.S. price reset persists and the European business goes away, netting to a step-down of $1.5B to $2.0B against a base that is now guided 1,000bps higher than it was in April. The absolute 2027 Eliquis dollar figure implied by that framing is therefore higher than the April framing implied, but the direction is unchanged and the 2028 U.S. loss of exclusivity is untouched.

Assessment: Management resisted the obvious temptation to imply that a stronger 2026 makes 2027 easier, and reiterated the step-down without being pressed on it. That is credible behaviour. The investment consequence is that the 2026 EPS number now includes a component the company has told you reverses, which is an argument for a lower multiple on that number rather than a higher one.

3. The acquired IPRD comparison, and the earnings growth that is not there

The 40% non-GAAP EPS increase is arithmetic from a base-period charge. In Q2 2025 the company booked $1,508M of acquired IPRD; in Q2 2026 it booked none. Its own summary table prints the net EPS impact of acquired IPRD and licensing income at $(0.57) last year and $0.01 this year, a swing of exactly $0.58, which is exactly the increase in reported non-GAAP EPS.

"Our effective tax rate in the quarter was 16.5%, reflecting jurisdictional earnings mix. Overall, diluted earnings per share were $2.04 for the quarter." — David Elkins, EVP and Chief Financial Officer

That is the entirety of the prepared commentary on earnings per share. Excluding the acquired IPRD and licensing item from both periods, EPS was $2.03 against $2.03. Non-GAAP pre-tax profit, adding the $1,508M charge back to the prior year, was $4,994M against $5,069M, down 1.5%.

Assessment: Nothing here is hidden; the company disclosed the item on the first page of the release and footnotes that both GAAP and non-GAAP EPS include it. But no analyst raised it across eleven questions, and the framing on the call and in the coverage that followed was uniformly about a 40% increase. An investor buying the multiple on the raised full-year guide should understand that the second quarter's operating earnings did not grow.

4. Gross margin above the range, with the range unchanged

Non-GAAP gross margin was 71.4%, against 72.6% a year ago and against a full-year guide of approximately 69% to 70% that management explicitly left unchanged. First-half gross margin was 70.9%.

"Gross margin in the second quarter was 71.4%, reflecting product mix." — David Elkins, EVP and Chief Financial Officer

The implied second half, holding the guide at its 69.5% midpoint, is approximately 68.1%. That is 280bps below the first-half run-rate and 90bps below the bottom of the guided range. No question in the Q&A touched gross margin.

Assessment: One quarter above a range is noise. Two quarters, with the range reaffirmed and a sharp implied step-down that nobody explained, is either genuine conservatism or a mix effect management expects to land abruptly. We lean toward the former and carry 69.9% for the year, above the guide. If we are wrong, the second half loses roughly 280bps of gross margin against the run-rate just delivered, which is worth about $0.28 of EPS.

5. The operating expense ramp arrived on schedule, and then grew

Non-GAAP operating expenses were $4,142M, up 7.5% sequentially from Q1's $3,854M and 4.8% year over year. In April we computed that hitting the then-$16.3B full-year guide required the remaining three quarters to average $4,149M. The second quarter came in at $4,142M, within $7M of the required run-rate.

"We have slightly increased our projected operating expenses for the year to account for accelerated prelaunch activities, supporting our CELMoDs, iberdomide and mezigdomide, the expanding pumitamig development program and opportunities to further support the growth portfolio." — David Elkins, EVP and Chief Financial Officer

The raised $16.5B guide requires the second half to average $4,252M per quarter, a further 2.7% above the level just delivered.

Assessment: This resolves cleanly in management's favour. The ramp we treated as possible conservatism in April turned out to be a committed spend, it landed on the number, and the incremental $200M is attached to named launch and development programmes rather than to a vague investment posture. It is also a reminder that the EPS raise is smaller than the revenue raise: $2.75B of incremental revenue produced $0.675 of incremental EPS, roughly $1.4B of net income, because the company is spending part of the upside.

6. Milvexian atrial fibrillation slips to Q1 2027

The single largest binary in the story moved out of 2026. Management disclosed it in prepared remarks rather than waiting for the question.

"For milvexian, we now expect the atrial fibrillation study to read out in the first quarter of 2027. This is an event-driven study and the updated timing from late 2026 reflects the pace of events. As a reminder, the study is being monitored by an independent data monitoring committee, which continues to endorse the study." — Christopher Boerner, Board Chair and Chief Executive Officer

The framing offered was that slower event accrual is favourable, because a lower background stroke rate in the control arm makes the non-inferiority margin easier to clear while a lower bleeding rate has never been the concern. The Chief Medical Officer disclosed the event thresholds: 430 events for the primary stroke and systemic embolism endpoint, 530 for bleeding, with both required before database lock.

Assessment: The clinical logic is coherent and we largely accept it; slower events on a non-inferiority trial genuinely is a better problem than fast events. But the interpretive argument was available in April too, when this readout was guided to late 2026 and the same slowing event rate was being described. In April the Chief Medical Officer's phrasing was that we will see later in the year if the event rate will change. It changed, in the direction the question anticipated, and the readout is now roughly five months later than the date that framed our initiation. We flagged in April that any slip past year end re-rates the whole second-half catalyst set. It slipped.

7. Cobenfy in Alzheimer's psychosis: the ADEPT programme moves to 2027

The second of the four catalysts also left the calendar, and for a different reason.

"With respect to the Cobenfy development program in Alzheimer's psychosis, based on the pace of enrollment in the ADEPT-2 and 4 studies and relapse events accruing more slowly than projected in ADEPT-1, we now anticipate top line data readouts from the ADEPT program to begin in early 2027 and be spread across the year with the potential interim analysis for ADEPT-1 later this year." — Christopher Boerner, Board Chair and Chief Executive Officer

Two distinct causes are bundled here. ADEPT-1 is event-driven and slow events carry the same benign interpretation as milvexian. ADEPT-2 and ADEPT-4 are enrolment-driven, and the delay there was attributed to a deliberate tightening of patient-selection and trial-conduct standards. The consolation prizes offered for later this year are open-label lead-in data from ADEPT-1 and data from ADEPT-3, the rollover study.

Assessment: An enrolment delay is a harder fact than an event-rate delay, because it reflects execution rather than epidemiology, and the explanation offered is one an investor cannot verify from outside. The interim analysis and open-label data now scheduled for later this year are uncontrolled and will not resolve anything. Cobenfy in Alzheimer's psychosis has moved from a 2026 catalyst to a 2027 sequence that is spread across the year, which is a materially different asset in a valuation framework.

8. Admilparant becomes the year's main event by default

With milvexian in atrial fibrillation and ADEPT out of 2026, the Phase 3 ALOFT-IPF readout in idiopathic pulmonary fibrosis is now the largest binary remaining on this year's calendar. It is also the one with the most favourable structure.

"It is an important aspect I want to highlight in Phase III, we have 2 shots on goal in each of the studies, IPF and PPF because we are testing 2 doses, 60 milligram, that is the dose that emerged from the Phase II, but also 120 milligram that was introduced later. We know there is a dose relation activity." — Cristian Massacesi, EVP, Chief Medical Officer and Head of Global Drug Development

On the commercial side, management sized the current market at roughly $4B and argued it could double to $8B to $10B by the middle of the 2030s with better-tolerated options, pointing to a competitor's launch where over 50% of new starts come after generic entry as evidence that a second-line add-on and switch market is forming for the first time. Recruitment is complete and the data monitoring committee has cleared the higher dose.

Assessment: This was the readout with the best setup in April and it is better now, because it is fully enrolled, dual-dosed and no longer competing for attention. It is also now carrying a disproportionate share of the year's catalyst weight. A disappointing ALOFT-IPF would remove the year's largest single catalyst, which is a concentration of risk that did not exist three months ago.

9. The CELMoD franchise and the August 17 PDUFA

Two protein-degradation assets now have regulatory dates. Iberdomide has a PDUFA date of August 17, 2026 and mezigdomide has one of May 13, 2027, based on the SUCCESSOR-2 trial that showed a 52% reduction in the risk of progression or death against carfilzomib and dexamethasone alone.

"We are awaiting the August 17 PDUFA date for iberdomide and are ready for the launch of this important medicine, which has the potential to be the first commercialized CELMoD. This would represent a brand-new class of treatment for myeloma, a historically difficult to treat illness with a patient population that could significantly benefit from new innovative options." — Christopher Boerner, Board Chair and Chief Executive Officer

The commercial framing is a direct substitution: iberdomide is intended to replace Pomalyst and Revlimid in leading second-line triplet regimens, in a setting where 70% to 80% of myeloma patients are treated in the community. Progression-free survival data is expected within a few months of approval, which management acknowledged some physicians will wait for.

Assessment: This is the most tangible near-term item on the calendar and it is three weeks away. It is worth noting what it is and is not: a CELMoD replacing Pomalyst and Revlimid recaptures revenue the company is already losing to generics rather than adding a new franchise on top. The economics of that substitution, at branded pricing against generic lenalidomide, are the thing to model, and no pricing framework was offered.

10. Cash flow recovered, and the promise from April did not arrive

Operating cash flow was $3,393M against $1,104M in the first quarter, free cash flow was $3,088M and the dividend was covered 2.4 times. Net debt fell $1,951M in the quarter to $31,656M.

"We remain in a strong financial position with approximately $11.5 billion in cash equivalents and marketable securities as of June 30. In the second quarter, we generated approximately $3.4 billion in operating cash flow and paid down an additional $1.2 billion in debt." — David Elkins, EVP and Chief Financial Officer

Against the year-ago quarter, however, operating cash flow was down 13.4% and first-half operating cash flow was down 23.4% to $4,497M. The specific item is visible in the cash-flow statement: rebates and discounts consumed $3,135M of cash in the first half against a $254M build in the comparable period. In April the Chief Financial Officer said roughly $1.2B of lower net cash collections from the January list-price reduction would be more than offset later in the year through lower rebate payments. Through June 30 it has not been, and the subject was not raised on this call.

Assessment: The dividend-coverage concern we opened in April is resolved for now; first-half free cash flow of $3,845M covers the $2,570M of dividends paid 1.5 times. The rebate promise is a different matter. It was a specific, quantified commitment made on a prior call, it has not appeared in the cash statement at the halfway point, and it went unmentioned. That is the single clearest instance this quarter of the say-to-do framing management invokes not being tested.

11. Capital allocation: a $5B authorization and no buyback for six quarters

The company discloses approximately $5B of share repurchase authorization remaining as of June 30, 2026. The first-half cash flow statement contains no repurchase line, and full-year 2025 repurchases were zero. Diluted shares rose from 2,038M to 2,048M year over year, and full-year EPS guidance assumes 2,050M.

"Business development remains a priority, and we are regularly evaluating opportunities in the therapeutic areas we know best, while continuing to return cash to shareholders through our commitment to the dividend." — David Elkins, EVP and Chief Financial Officer

On business development specifically, the CEO characterized the posture as opportunistic rather than urgent, saying the company does not feel any compulsion to chase deals given the strength of the late-stage pipeline, and named the criteria as therapeutic areas the company knows well, compelling science and a financial case, ideally with the effect of de-risking the near-term growth profile.

Assessment: Capital allocation here is coherent but conservative: pay down debt, protect the dividend, hold the buyback in reserve, keep balance-sheet capacity for a transaction that de-risks the 2028 gap. It is defensible with net debt at $31.7B. It also means shareholders get no support from share-count reduction while the company carries an authorization it does not use, and the share count drifts up half a point a year on equity compensation.

12. The AI partnerships and the rewiring framing

Management devoted a slide and a section of prepared remarks to operating-model change, anchored on newly announced agreements with two AI providers, including an expansion with a compute supplier announced in July and an enterprise deployment agreement announced in May.

"Concretely, this will help our scientists do 3 things: understand disease biology more deeply, design and test candidate molecules faster and make earlier, better informed decisions about which programs to move forward." — Christopher Boerner, Board Chair and Chief Executive Officer

The Chief Medical Officer separately connected the theme to trial execution, describing the integration of AI and data into a per-programme probability-of-success improvement effort in the context of a question about historical Phase 3 study-conduct problems.

Assessment: There is no quantified benefit attached to any of this and none was claimed, which is the correct posture. It carries no modelling consequence in 2026 and we assign it none. Note the timing: the operating-model narrative expanded on the same call that pushed two flagship readouts into 2027 for reasons that included enrolment execution. That juxtaposition is worth watching rather than dismissing.

Guidance & Outlook

Full-year guidance assumes 2,050M diluted weighted-average shares and excludes acquired IPRD charges and licensing income incurred after June 30, 2026. April guidance was set on mid-April exchange rates and July on mid-July rates.

Implied second half. First-half revenue was $24,462M, so the new midpoint implies a second half of $25.04B, 2.4% above the first half and 1.3% above the $24,724M delivered in the second half of 2025. First-half non-GAAP EPS was $3.62, so the new midpoint implies $3.26 for the second half, a 10.1% sequential decline, or roughly $1.63 a quarter against the $2.04 just reported. First-half operating expenses were $7,996M, so the $16.5B guide implies $8,504M in the second half, an average of $4,252M a quarter.

Street at: The raised revenue range sits roughly $2.5B to $3.5B above where the consensus revenue estimate for the year had been anchored, since the pre-print quarterly estimate of $11.75B was itself derived from the April guide. The EPS range brackets the level consensus had been carrying for 2027 rather than 2026.

Guidance style: Two data points now suggest a conservative pattern. Gross margin has printed above the guided range in both quarters of 2026 with the range never adjusted, and the April Eliquis growth guide was revised up by 1,000bps one quarter after it was set. Against that, the operating-expense guide has been raised twice and each increase was attached to specific programmes, which reads as a genuine spending decision rather than sandbagging. We carry EPS of $6.94, above the $6.875 midpoint and below the $7.00 top end.

Analyst Q&A Highlights

Whether the two readout delays reflect a problem with the underlying programmes

The first question of the call went straight to the timing changes and asked whether something in the patient populations had turned out different from the original assumptions. Management separated the two programmes, treating the milvexian and ADEPT-1 delays as event-rate phenomena with a benign reading and the ADEPT-2 and ADEPT-4 delays as an enrolment consequence of tightened quality standards. Confidence in the underlying science and the study designs was reaffirmed without qualification.

Q: "I guess I just wanted to talk about the ADEPT and milvexian studies going into 2027. I know they're event-driven, but maybe can you talk a little bit about what this means relative to your initial assumptions? Is there something in the patient population maybe that has changed or that is perhaps different than what you guys initially assumed?"
— Geoffrey Meacham, Citibank

A: "So this scrutiny in preserving quality and increasing quality impacted the enrollment pace. The second point I want to make is about where we are now. I think we have implemented now several measures to accelerate this enrollment. And I think the pace and the projections on time lines are solid."
— Cristian Massacesi, EVP, Chief Medical Officer and Head of Global Drug Development

Assessment: The answer was substantive and the distinction between the two causes is real, but it also concedes the more uncomfortable half. An event-rate delay is epidemiology; an enrolment delay caused by raising the bar on patient selection mid-programme is an admission that the original conduct standard was not the one the company now wants. That is a defensible fix and a genuine delay, and the phrase about projections being solid is the third timeline this programme has carried.

What magnitude of benefit milvexian needs to be commercially transformational rather than incremental

A recurring line of questioning pressed on where the commercial threshold sits, framing the choice as cannibalizing meaningful Eliquis volume against serving an underserved slice of the atrial fibrillation market. Management declined to name a hazard ratio and redirected to the trial's design intent and to the roughly 40% of patients who should be anticoagulated but are undertreated, underdosed or discontinued because of bleeding concerns.

Q: "I think there's a debate on the street of what type of hazard ratio you need to see for this drug to get broad adoption, cannibalizing sizable pieces of Eliquis versus just targeting maybe an underserved segment of the AF market. Can you just help frame out what you think you'd need to see from that study for this to be a very large opportunity for Bristol versus what would be maybe more of just an incremental new product for the company just in terms of how strong that data has to be?"
— Christopher Schott, JPMorgan

A: "And so I don't think it makes sense to speculate on hazard ratio. But what I can say is our study, as Cristian mentioned, was designed to demonstrate a superior bleeding profile compared to Eliquis with comparable efficacy."
— Adam Lenkowsky, EVP and Chief Commercialization Officer

Assessment: This is the same refusal we noted in April, offered in the same terms, and it remains the right answer for a company with a blinded trial. The useful new information was in the surrounding commentary rather than the answer itself: management now expects milvexian to shift first-line decision-making over time rather than only to unlock the undertreated tail, which is a wider commercial ambition than was articulated three months ago and a higher bar for the data to clear.

Which assets are actually driving the raised growth-portfolio outlook

Asked to name the standout contributors behind the guidance increase, management pointed to breadth rather than to any single asset, listing four products with growth rates between 22% and 59% before returning to Eliquis demand as the anchor of full-year confidence.

Q: "On the guidance increase, can you just talk a little bit about the drivers of the upside for the growth portfolio that's implied in the guidance? I guess is there any standout assets that are driving that upside?"
— Christopher Schott, JPMorgan

A: "And on guidance, Chris, look, it was -- you saw in the results, it was pretty pervasive across the growth portfolio. We saw really strong growth. Opdualag was up 22%, Reblozyl 29%. Breyanzi, we continue to see great growth there at 41% and Camzyos was at 59%."
— David Elkins, EVP and Chief Financial Officer

Assessment: The answer is accurate and slightly misdirecting. The products named are genuinely growing and genuinely broad-based, but the guidance revision itself was quantified only on two items, the Eliquis growth range and the Legacy decline range, both of which sit outside the Growth Portfolio. The company's own guidance slide lists continued Growth Portfolio strength first and the two quantified Legacy revisions second, which reverses the order of magnitude.

Whether Cobenfy's titration problem is being solved

A question returned to the dosing issue management has acknowledged for several quarters, that too many patients remain below the effective dose. The response quantified prescriber and prescription trends but stopped short of saying the dose-mix shift has reached a level that changes the revenue trajectory, and repeated that further increases are still needed.

Q: "You've talked about one of the drags on sales being that clinicians aren't titrating up quickly enough to the therapeutic dose. So just wondering if you could give us an update on how the educational efforts on optimizing that are going and when perhaps we could see that reflected in the sales trajectory?"
— Emily Field, Barclays

A: "What we have seen, though, is an increase in 125-milligram prescribing since introducing our Phase IV switch study, which will both help maximize efficacy, but we need to see further increases in the dose shift towards the 125 milligrams."
— Adam Lenkowsky, EVP and Chief Commercialization Officer

Assessment: An honest answer that concedes the problem is not fixed. The candour is welcome and the implication is not: with the Alzheimer's psychosis readouts now in 2027, the schizophrenia base is all Cobenfy has for another eighteen months, and management is telling you that base is still constrained by a prescribing behaviour they have been trying to change for over a year.

Whether a competitor's strong launch raises the bar for admilparant

Questioning turned to whether a recently launched competitor in idiopathic pulmonary fibrosis has raised the commercial threshold. Management inverted the premise, arguing that the competitor's launch validates the market and, more specifically, has created a second-line add-on and switch segment that did not previously exist.

Q: "Just given the really strong launch of JASCAYD by BI, does this impact the bar for success in your view? Does it elevate the maybe the commercial bar for success? Or is potentially winning and differentiating versus JASCAYD mainly on tolerability? And do you see the market potentially moving to polypharmacy much like markets like pulmonary hypertension."
— Jason Gerberry, Bank of America

A: "And I think what you're getting at is, right, JASCAYD's launch validates the market opportunity and shows there's an opportunity to gain significant share as better tolerated and effective new treatment options emerge. And that's what we're seeing. We're seeing the creation of a second-line add-on market for the first time ever as over 50% of new JASCAYDs starts are coming after generic Ofev or generic Esbriet as switch or add-on treatments."
— Adam Lenkowsky, EVP and Chief Commercialization Officer

Assessment: The market-expansion argument is the strongest commercial case made on the call and the supporting data point, that a majority of a competitor's new starts are post-generic add-ons, is specific and checkable. It is also an argument that a rival is executing well in the category BMS intends to enter, and it was made in the same breath as claiming a placebo-comparable tolerability profile from Phase 2. Phase 2 tolerability claims in fibrosis have not historically survived Phase 3 at the same magnitude.

Whether the 2026 Eliquis strength carries into 2027

The most consequential question of the call asked directly whether this year's Eliquis growth is a tailwind, a headwind or neither next year, and how much of the 2026 revenue is sustainable. Management reiterated the step-down figure it introduced last quarter without revising it, and identified the European patent expiry as the offset to a U.S. base that carries forward.

Q: "First, how will the increased growth of Eliquis in 2026 impact sales in 2027 and beyond? Is it a tailwind, a headwind or neither in '27? And is any portion of those revenues in '26 sustainable?"
— Steve Scala, TD Cowen

A: "Now we still expect the $1.5 billion to $2 billion step down in 2027 that we referenced on the Q1 call. We'll clearly provide any appropriate update in conjunction with our 2027 guidance when issued. But the U.S. sales certainly would carry forward to next year, offset by the patent expiry of Eliquis in Europe in mid-Q4."
— Adam Lenkowsky, EVP and Chief Commercialization Officer

Assessment: The most important answer on the call, and management did not use it to sell. It would have been easy to argue that a higher 2026 base makes the 2027 step-down proportionally smaller, and that argument was not made. The disclosure investors should take from it is that the company still expects Eliquis revenue to fall next year even after raising this year's growth by 1,000bps, which means a meaningful part of the 2026 upside is a level shift rather than a growth rate.

How to read a data monitoring committee's continued endorsement of a blinded trial

A pointed exchange challenged the interpretive weight management had placed on the committee's ongoing endorsement, noting that milvexian's bleeding advantage was never in doubt and that the only question worth asking concerns efficacy. Management responded by separating the two, arguing the endorsement carries information about efficacy specifically because efficacy is the endpoint at risk.

Q: "So can you please comment further on the endorsement, including how you're interpreting study continuation as a positive signal when the DMC is evaluating both efficacy and bleeding. I ask because it's already known and clear that milvexian is associated with much lower bleeding risk than Eliquis. So that is not any surprise that lower bleeding would be positive in the trial. So I'm more interested in your interpretation on the primary endpoint of efficacy with respect to your observations."
— David Risinger, Leerink Partners

A: "And if you look not only at the most recent reviews, but really just look at this over time, we've continued to see on this program a slowing of event rates, along with the continued view that the efficacy continues to be in line with expectations to continue the study."
— Christopher Boerner, Board Chair and Chief Executive Officer

Assessment: The question was the sharpest of the call and the answer was better than a deflection without being conclusive. A committee that has not stopped a trial for futility or harm is genuine information, and the supporting anecdote about a competing study in the same indication being halted for a directional efficacy signal gives it real content. But a continuation decision is a low bar by construction, and management is inferring from it a confidence level that the bar does not support. This exchange is the clearest example of a story being built on the absence of bad news.

What the disclosed event thresholds reveal about the timeline

The final question sought clarification on whether both the efficacy and bleeding event counts are accruing slowly, or whether the bleeding threshold has already been met. The answer disclosed both required event counts and confirmed that database lock waits on the later of the two.

Q: "You previously mentioned both bleeding events and AFib events need to mature for the trial to stop. Are both bleeding and AFib tracking slower than expected? Or have we already reached the required number of events for bleeding events already?"
— Akash Tewari, Jefferies

A: "What I can tell you because this is -- we disclosed this, we need to recruit 430 events for primary endpoint, strokes and embolism and 530 events for bleeding. We want to achieve both numbers before locking the database and have the readout because the primary endpoint is very important, but also then we test the superiority on bleeding is also very important. So we are on track for this, and this will happen first quarter '27."
— Cristian Massacesi, EVP, Chief Medical Officer and Head of Global Drug Development

Assessment: A useful disclosure that also quietly answers the question that was not asked. The trial requires a bleeding-superiority test, and the bleeding threshold is the higher of the two counts at 530 against 430. A lower bleeding rate on the investigational arm therefore slows the readout while helping the secondary endpoint. That structure means the timeline slipping is mechanically consistent with the outcome management wants, which is a real point in their favour and a reason to expect further timing risk rather than less.

What They're NOT Saying

1. The flat underlying EPS. Non-GAAP EPS excluding acquired IPRD and licensing income was $2.03 in both this quarter and the year-ago quarter. The item is disclosed on page one of the release and footnoted twice, and neither prepared remarks nor eleven analyst questions mentioned it. The framing on the call was a strong quarter and the discussion of EPS ran to a single sentence.
2. The Eliquis inventory reversal promised in April. Last quarter management confirmed a Q1 wholesale build existed and confirmed it would reverse in the second quarter, without quantifying either. The Opdivo drawdown reversal was disclosed and sized this quarter; the Eliquis build was never mentioned again. With Eliquis growing 22% against what was then a 10% to 15% full-year guide, the unquantified portion matters.
3. The rebate offset in cash flow. In April the CFO said roughly $1.2B of lower net cash collections would be more than offset later in the year through lower rebate payments. Through June 30 the rebates and discounts line has drained $3,135M against a $254M build in the comparable half. No update was offered and no question was asked.
4. The implied second-half gross margin. Gross margin has printed above the guided 69% to 70% range in both quarters of 2026 and the range was reaffirmed unchanged, which implies a second half near 68.1%. Management's entire commentary was that gross margin reflected product mix. The step-down was neither explained nor challenged.
5. Why the April guide was $2.75B too low. A guidance revision of this size, three months after the guide was set, on a product the company has sold for over a decade, invites a question about the forecasting process. The guide was raised, the reasons were listed as line items, and no one asked what had been misjudged in April.
6. The economics of substituting iberdomide for Revlimid and Pomalyst. Management described iberdomide replacing both drugs in second-line triplet regimens as the commercial strategy. Those two products generated $629M this quarter against $1,546M a year ago, and the substitution is branded pricing against generics that are already in the market. No pricing, gross-to-net or timing framework was offered for the transition.
7. The unused buyback. Approximately $5B of repurchase authorization remains, no shares have been repurchased since at least the start of 2025, and the diluted share count is rising. The dividend was affirmed and business development was described as a top priority. The authorization itself was not discussed.
8. What ADEPT-2 and ADEPT-4 enrolment quality means about the earlier standard. The delay was attributed to prioritizing patient and conduct quality, which implies the prior standard was not what the company now requires. Asked separately about historical Phase 3 study-conduct deficiencies, management described a general remediation framework and said the two delays were unrelated in a certain way. The qualifier was not explored.

Market Reaction

• Pre-print setup: BMY closed at $63.10 on July 29, up 17.0% year to date against the S&P 500's 6.9%, up 37.2% over the trailing twelve months and up 9.5% over the trailing thirty days. The 52-week closing range entering the print was $42.60 to $63.60, so the stock arrived within 0.8% of its highest close of the year.
• Reaction session: The stock opened at $62.69, a 0.6% gap lower, traded as low as $62.23 and as high as $65.56, and closed at $64.86, up 2.8% or $1.76. That close is a new 52-week closing high.
• Volume: 18.0M shares against a 12.5M thirty-day average, or 1.4 times normal. Elevated but well short of a repositioning event.
• Relative: The S&P 500 rose 1.7% on the same session, so the idiosyncratic excess return on a beat-and-raise of this magnitude was roughly 110bps.

The shape of the session is more informative than the close. A company that beat revenue by 10.4%, beat EPS by 28% and raised full-year revenue guidance by $2.75B opened lower and had to build its gain intraday, finishing 110bps ahead of the index. That is not the tape of a market discovering good news; it is the tape of a market weighing good news against something else on the same page.

The something else is the readout calendar. Two of the four Phase 3 events that framed the second half moved into 2027, and they were disclosed in prepared remarks before the numbers had been digested. For a stock that had already run 37% over twelve months and arrived at the print within a point of its high, a 2026 guidance raise funded by a product with a stated 2027 step-down is worth less than it looks. The market appears to have made roughly that adjustment: it took the earnings upgrade and discounted the pipeline deferral, and the net was a hundred basis points.

Positioning matters here too. The pre-print setup was the opposite of April's. Three months ago the stock entered the Q1 print at $57.59 after a de-rating in the month before it and rallied 5.2% on a reaffirmed guide with softer language attached. This time it entered at a near-high on a 37% twelve-month gain, and a materially better print produced roughly half the move. The de-rated entry point that made the April quarter easy to trade is gone.

Street Perspective

Debate: Does the guidance raise change the shape of the cliff or only its timing?

Bull view: The bull case circulating is that a $2.75B revenue raise and a 1,000bps upgrade to Eliquis growth is not a timing detail, it is $2.75B of incremental cash that funds business development, debt paydown and the launch spend behind two CELMoDs, all in the window before the cliff arrives. The Legacy book outside Eliquis is now down to $941M a quarter, which caps the remaining erosion from that cohort at a manageable number, and the Growth Portfolio at 58% of revenue with 14% growth is compounding into the gap.

Bear view: The bear camp answers that revenue excluding Eliquis declined 1.1% this quarter, that the Growth Portfolio still does not cover Legacy erosion outside Eliquis, and that management has now told investors twice that Eliquis steps down $1.5B to $2.0B in 2027 before U.S. exclusivity ends in April 2028. On this reading the raise moves a year of revenue forward without adding one, and a company earning $6.94 in 2026 on a component that reverses is not a company that earns $6.94 in 2027.

Our take: The bears have the durability argument and the bulls have the balance sheet. What genuinely changed is that the 2026 arithmetic no longer forces a decline, which removes the most immediate pressure and buys a full year of capital-allocation optionality that did not exist in April. What did not change is the 2028 exposure, and the company has been careful not to imply otherwise. The right adjustment is to raise the estimate and hold the multiple, which is what we have done.

Debate: How much did the readout slippage actually cost?

Bull view: The argument on the optimistic side is that neither delay carries negative information. Slower event accrual on a non-inferiority trial makes the primary endpoint easier to clear, the data monitoring committee continues to endorse continuation, and the disclosed event thresholds show the bleeding count is the binding constraint, which is mechanically consistent with milvexian working. A readout is worth the same in March 2027 as in December 2026 if the probability is unchanged.

Bear view: The skeptical framing is that a delayed catalyst is a cheaper catalyst regardless of the reason, because the discount rate runs for longer and because timing guidance that has already moved once is more likely to move again. On the Cobenfy programme the cause was enrolment execution, not epidemiology, and that is the delay that cannot be reframed as good news. A year of waiting has been added to a story whose entire re-rating case rests on these events.

Our take: The bulls are right on the mechanism and the bears are right on the price. We accept the event-rate explanation for milvexian and ADEPT-1, and we do not accept it for ADEPT-2 and ADEPT-4, where the company effectively said it raised its own conduct standard mid-programme. The practical consequence is concentration: admilparant in idiopathic pulmonary fibrosis now carries more of the year than any other single readout, and a disappointment there would remove the largest positive Phase 3 event and a 2027 calendar that has already proven movable.

Debate: Is 9.3 times a raised number cheap, or the right price for a peak one?

Bull view: The value case is that the stock has re-rated less than the earnings have. At the April reaction close of $60.59 the shares traded at 9.8 times the then-guided $6.20 midpoint; at $64.86 they trade at 9.4 times the $6.875 midpoint, so the multiple compressed while the price rose 7%. Add a 3.9% dividend, a balance sheet that shed $2.4B of net debt in six months, and an unused $5B authorization, and the downside is well defended for a business now guided to grow.

Bear view: The counter is that a lower multiple on a higher number is not automatically better value when the company has said the number contains something that reverses. Underlying EPS excluding the acquired IPRD comparison was flat year over year, non-GAAP pre-tax profit on a like-for-like basis fell 1.5%, and the earnings uplift is a pricing reset on one product. On a normalized 2027 base the multiple is not 9.4 times, and the stock is at a 52-week high having risen 37% in a year.

Our take: Both sides are describing the same fact from opposite ends, and the honest resolution is that the multiple is correctly low rather than mistakenly low. We raise our fair-value range because the earnings base is genuinely higher, and we narrow the top of the multiple range because 2026 is now more clearly a local peak than it appeared in April. That combination produces a stock trading close to fair value with a decent yield, which is a Hold, not a mistake.

Model Update

We initiated in April without a prior model and set the working assumptions below. The table shows what this quarter changed.

Valuation. At the $64.86 reaction close, on 2,048M diluted shares, the market capitalization is $132.8B and enterprise value is $164.5B including $31.66B of net debt. That is 9.3 times our $6.94 estimate, 9.4 times the $6.875 guide midpoint and 3.3 times enterprise value to our revenue estimate. The annualized dividend of $2.52 yields 3.9%.

Valuation impact: Applying 9.0 to 10.5 times to $6.94 gives a fair-value range of $62.50 to $72.90 with a $67.70 midpoint, raised from the $55.80 to $68.20 range and $62.00 midpoint we set in April. The estimate rises 11% and the multiple range narrows at the top, because 2026 now looks more clearly like a local peak in earnings than it did three months ago. Against the $64.86 close, the midpoint implies roughly 4.4% of price appreciation plus the 3.9% yield, or about 8.3% of expected twelve-month total return.

Thesis Scorecard Post-Earnings

Overall: The standing thesis is strengthened on the P&L and weakened on the pipeline, which is close to a wash. Three of the four bear points were challenged this quarter and the one that was confirmed, earnings quality below the operating line, is the one the market is not discussing. Against that, the bull pillar that carried the most option value, catalyst density inside 2026, lost half its content. The two tests we set in April split: the guidance language was upgraded to a full raise, and the readout calendar slipped past year end.

Action: Hold. The estimate rises 11% and the fair-value midpoint rises 9%, but the stock rose 7% to a new 52-week closing high at the same time, so the expected twelve-month total return of roughly 8.3% is barely distinguishable from a market return. The reason to own this in April was a cheap multiple plus four shots at a re-rating inside nine months. The multiple is no longer as cheap relative to a normalized 2027 base, and two of the four shots now sit in 2027. We would add on weakness toward the low $60s, and we would want either the admilparant readout in hand or a 2027 guide that does not step down before paying up.