On Tuesday, 8 September 2026, Novartis shares fell about 9.4 percent in Zurich, and briefly more than ten. Measured by one-day loss, that is the worst trading session in the company’s history. Roughly 24 billion Swiss francs — about 29.6 billion dollars — of market value was gone by the afternoon. In early September the stock had been trading near 131 francs and a market capitalisation of some 234 billion francs, close to its twelve-month high of 132.48.
The trigger was a trial readout. Novartis said the Phase III HARBOR study of del-desiran (delpacibart etedesiran) in myotonic dystrophy type 1 had missed its primary endpoint. No statistically significant improvement over placebo.
That is the news. The number that actually matters is a different one: 12 billion dollars. That is what Novartis paid in October 2025 for Avidity Biosciences, the company del-desiran came from. In a single session the market wrote off two and a half times that purchase price. That is not a price for a molecule. That is a price for a strategy.
Three failures in fifteen days — and each one on a different route
Tuesday was not the first bad day of this run. It was the third. And the real story is not that there were three, but which three.
On 24 August, Novartis halted eight clinical trials of its experimental cell therapy rap-cel after being notified of three cases of a severe, life-threatening immune reaction. All three patients died. The diagnosis was IEC-HS, a known CAR-T complication in which the engineered T cells expand uncontrollably inside the body and the immune system turns on the patient’s own organs. The paused programmes covered lupus, rheumatoid arthritis, vasculitis, multiple sclerosis and myasthenia gravis. Bristol Myers Squibb paused autoimmune CAR-T work of its own in the same week. Rap-cel was in-house research — the continuation of the cell-therapy line on which Novartis had won the first CAR-T approval anywhere.
On 4 September came pelacarsen. The Lp(a)HORIZON study, with 8,323 patients, tested whether lowering lipoprotein(a) reduces major cardiovascular events — a composite of cardiovascular death, non-fatal myocardial infarction, non-fatal stroke and urgent coronary revascularisation requiring hospitalisation. The drug lowered Lp(a). It did not lower risk. Shreeram Aradhye, president of development and chief medical officer at Novartis, put it plainly in the release: although lower Lp(a) levels were observed with pelacarsen, the findings did not demonstrate that this translated into reduced cardiovascular risk in the overall study population. Pelacarsen was in-licensed, from Ionis Pharmaceuticals, with worldwide rights sitting at Novartis.
And on 8 September, del-desiran — acquired, for 12 billion dollars.
There are exactly three ways to fill a drug pipeline: discover it, license it, or buy it. Within fifteen days a leading candidate failed on every one of them. That is why the market on Tuesday did not write down a drug. It wrote down a method.
What HARBOR actually measured
Myotonic dystrophy type 1 is an inherited muscle disease that causes progressive stiffness and weakness. There is no approved treatment addressing the cause. HARBOR’s primary endpoint was video hand opening time, or vHOT: how long a patient needs to open the hand again after squeezing it shut. That sounds trivial and is not — myotonia, the delayed relaxation of a contracted muscle, is the defining symptom of the disease, and the test can be measured objectively on video.
Del-desiran missed it. Novartis phrased the outcome conservatively: no statistically significant improvement versus placebo.
The drug comes from the AOC platform Novartis bought with Avidity — antibody-oligonucleotide conjugates. An antibody acts as the delivery vehicle carrying an RNA-active payload into muscle cells, precisely where conventional RNA therapeutics have struggled to arrive. That was the technological bet behind the price. Novartis paid 72.00 dollars a share in cash, a 46 percent premium to the close two days before the announcement, making it the group’s largest acquisition in more than a decade. Del-desiran was the platform’s most advanced candidate — the one that would show first whether the technology works.
The arithmetic: why 24 billion is more than the drug was ever worth
This is where it gets interesting, and this is the core of the day.
Barclays modelled del-desiran at roughly 3.1 billion dollars in peak annual sales with a 60 percent probability of success. Vas Narasimhan, chief executive since 2018, had told Bloomberg TV in July the medicine had peak sales potential of five billion dollars and more.
Take the more conservative figure and be generous with everything else. Even crediting a fully approved, launched medicine with a present value of four times peak sales, del-desiran would be worth a little over 12 billion dollars in the success case. If 60 percent of that was in the share price, roughly 7.4 billion dollars had to come out on Tuesday.
What came out was 29.6 billion. The difference, about 22 billion dollars, was not del-desiran.
Those 22 billion are the market’s new price for the sentence we buy the pipeline we do not invent. What was repriced is not the past but the probability attached to the next acquisitions. James Eugene of Verso Investment Management said the setback would have dented confidence in the acquisition strategy. Jefferies had already written in July that HARBOR needed to deliver to justify the 12 billion dollar price tag.
The patent cliff that makes all of this urgent
Without the background, the day would be half as bad. Novartis is facing the largest patent expiry in its history in 2026.
Operationally, 2025 was a strong year: sales up 8 percent, core operating income up 14 percent to 21.9 billion francs, core earnings per share up 17 percent to 8.98 dollars, and a core margin of 40.1 percent reached ahead of the original plan. A handful of medicines carried it. Entresto rose 30 percent to 7.8 billion dollars, Cosentyx 23 percent to 6.1 billion, Kesimpta 49 percent to 3.2 billion, Kisqali 46 percent to 3.0 billion.
The problem sits inside that same list. Entresto, the heart-failure drug and the largest single seller, already met US generic entrants in 2025, as did Promacta and Tasigna. Cosentyx loses exclusivity around the end of the decade. Losing 7.8 billion dollars of annual sales that were still growing 30 percent means replacing them — and not eventually, but inside a defined window.
Avidity was meant to help do exactly that. For 2026 Novartis guided to low single-digit sales growth despite the patent cliff, and built in a one-to-two percentage-point hit to the core margin from the Avidity deal. That cost remains. The revenue it was supposed to be measured against has been pushed out, and at best made smaller.
Worth noting: full-year guidance was reaffirmed on Tuesday, as was the medium-term target of 5 to 6 percent annual sales growth through 2030. The share price loss is not about 2026. It is about the years after it.
Who else lost on Tuesday
The moves outside Basel say something about what the market considers transferable and what it does not.
After the pelacarsen readout, Ionis Pharmaceuticals fell about 10 percent and Amgen dropped roughly 5 percent in extended trading. Amgen is developing olpasiran, the direct competitor on the same mechanism; Eli Lilly is in the race with lepodisiran. Both produced deeper Lp(a) reductions than pelacarsen in mid-stage work. So the Novartis failure is not automatically the end of the idea — but it is the first dedicated outcomes trial in the class, and it is negative. Amgen reads out its secondary-prevention study in December. That is the date on which it will be decided whether the hypothesis was wrong or only the molecule.
After the HARBOR readout, the neuromuscular corner took the hit: Dyne Therapeutics lost about 30 percent, Sarepta Therapeutics more than 15. Both are pursuing related approaches — if vHOT was out of reach for a well-funded large-cap, that shifts expectations for everyone chasing the same target with less money.
What this means for a US investor, concretely
The American reference case is Pfizer. Facing its own post-pandemic revenue cliff, it paid roughly 43 billion dollars for Seagen in 2023 — the same manoeuvre Novartis attempted in smaller size, and with the same structural feature: after a deal like that, the market no longer prices the asset acquired, it prices the acquirer’s judgement. Merck faces the Keytruda expiry and has been buying steadily. Bristol Myers Squibb, which paused its own autoimmune CAR-T trials in the same week, has spent the past several years replacing Revlimid revenue with acquisitions. This is not a Swiss situation. It is the standard operating condition of large-cap pharma.
The direct readthroughs from Tuesday are Amgen, Eli Lilly and Ionis on the Lp(a) side, and Sarepta and Dyne on the neuromuscular side — but the more useful way to think about the sector is one layer down. Thermo Fisher Scientific, Danaher and IQVIA sell the instruments, bioprocessing equipment and clinical-trial services that every one of these programmes consumes. They are paid for research being done, not for research succeeding. That is the economic mirror image of Tuesday: Novartis still has to close a patent cliff, so it will keep buying and keep running trials whether HARBOR worked or not. The limit of that argument is real, though — when an industry absorbs a series of expensive failures, the reaction that follows is often a cut to research budgets, and the suppliers feel that too.
On tax, the point US holders most often miss with Novartis concerns withholding. Switzerland withholds 35 percent on dividends. Under the US-Swiss treaty this drops to 15 percent for eligible investors, which normally requires a properly filed W-8BEN with the broker; the 15 percent that remains can generally be claimed as a foreign tax credit, while amounts withheld above the treaty rate have to be reclaimed from the Swiss authorities and are not creditable. Holding the NVS American depositary receipt does not change the underlying Swiss tax; it only changes who processes it. On the domestic side, long-term capital gains are taxed at 0, 15 or 20 percent depending on income, plus the 3.8 percent net investment income tax where applicable. If you are tempted to sell into Tuesday’s drop and repurchase, note that the wash-sale rule disallows the loss if you buy a substantially identical position within 30 days either side of the sale.
The case against the market’s reading
A 24 billion franc drawdown is an opinion, not a proof. Four objections deserve serious examination.
First: none of the three announcements changes this year’s cash flow. Guidance stands. What was repriced is earnings from roughly 2030 onward — precisely the part of a valuation where the error bars are widest and sentiment has the most leverage.
Second: against core operating income of 21.9 billion francs a year, a 12 billion dollar purchase price is not a bet-the-company move; it is a bit over half a year of earnings. Novartis can absorb several failures like this without difficulty. Whether it should is a separate question — but it is not an existential one.
Third: Phase III success rates industry-wide are well below one hundred percent. Three failures in fifteen days is a cluster, not a trend break. With dozens of programmes running at once, independent events will coincide sooner or later. Narratives grow out of clusters far more readily than insights do.
Fourth: the Avidity platform has not been refuted. Del-zota, for Duchenne muscular dystrophy with mutations amenable to exon 44 skipping, is already filed with the US Food and Drug Administration for accelerated approval and has been granted priority review. For del-brax in facioscapulohumeral muscular dystrophy, Novartis plans to meet the agency on next steps after positive early-phase biomarker data. If those two deliver, Tuesday was an overreaction and Avidity was a platform purchase, not a one-molecule purchase.
There is also a counter-argument running the other way, and it is the least comfortable one: if a 60 percent probability of success on the lead asset is a normal basis for a 12 billion dollar acquisition, then the market price of bought-in pipelines is structurally too high. That would not be a Basel problem. It would be an industry problem.
How we will know who was right
The day leaves four testable dates behind, and they are unusually cleanly set.
The FDA decision on del-zota is the direct test of the claim that Avidity was a platform and not a single molecule. An approval would mark a meaningful share of Tuesday’s 22 billion as a misjudgement. Amgen’s olpasiran readout in December answers the other open question: if that programme fails too, the Lp(a) hypothesis itself was wrong, and Novartis had a science problem that would have hit anyone. If it delivers, pelacarsen was simply too weak, and the failure belongs to the selection made in Basel. Then there are the full Lp(a)HORIZON data at an upcoming medical congress — subgroups may show whether the drug did anything for anyone. And finally the outcome of the rap-cel safety review, which determines whether the eight paused trials restart.
The backdrop is not making any of this calmer. Brent traded just under 98 dollars on Tuesday after Houthi attacks on Saudi energy facilities at Jazan, Najran and Abha wounded more than seventy people and halted part of the operations. The ten-year Treasury yield stood at 4.80 percent and the thirty-year at 5.27 percent, with markets pricing close to a sixty percent probability of a quarter-point rate increase at the Federal Reserve meeting on 16 September. August consumer prices land on 11 September. Canada’s retaliatory tariffs on roughly 20 billion dollars of US goods took effect on Tuesday.
In that kind of environment, a defensive pharmaceutical name is normally where money goes to hide. Which is exactly what makes Tuesday instructive. The market did not reprice the economy at Novartis, or rates, or tariffs. It repriced the question of whether a company whose biggest medicine already faces generics can reliably buy the replacement. Since Tuesday the answer to that question is: 22 billion dollars less reliably than on Monday.
Try TradingView Free for 30 Days
Plus get a $15 discount on your first subscription through this link.
