Everything About AI
NextFin News - Novo Nordisk said on Wednesday it will work with Anthropic to apply the Claude family of models, including the science-focused Claude Science, to its drug-discovery workflows and software engineering, the latest in a string of artificial-intelligence partnerships the Wegovy maker has assembled this year as it races to compress the path from lab bench to marketed medicine.
The collaboration, announced from the company's headquarters in Bagsværd, Denmark, will have Novo's scientists and computational teams work with Anthropic on specific drug-discovery challenges and biological-reasoning problems, with an initial aim of testing Claude Science in targeted research-and-development workflows. Novo will also deploy Anthropic's frontier models to strengthen AI-driven software development, which the company called a key enabler for scaling AI across its operations. The deal was designed with defined data-governance and human-oversight frameworks; financial terms were not disclosed.
The announcement matters less as a standalone deal than as the third major AI partnership Novo has signed in five months. In April, the company tied up with OpenAI to embed AI from early discovery through manufacturing and commercial execution. In August, it named Amazon Web Services its preferred cloud provider and strategic AI partner, opening a co-innovation hub in London built around Amazon Bio Discovery and Amazon Bedrock. Days later it added a partnership with H1 on AI for drug discovery and clinical trials, and it has also worked with NVIDIA and Denmark's sovereign AI supercomputer, Gefion. Taken together, the deals amount to a partnership stack rather than a single bet — and they arrive as Novo tries to defend a growth story that its own 2026 guidance no longer promises.
A Partnership Stack, Not a One-Off Bet
The clearest read of Novo's AI activity is that the company is building infrastructure, not placing a wager on one model. Each partner covers a different layer of the stack. OpenAI provides broad enterprise and discovery capabilities across the value chain. AWS supplies the cloud foundation, the London co-innovation hub, and purpose-built life-sciences services. Anthropic brings a science-specialized reasoning model and agentic software engineering. Novo is effectively insuring itself against picking the wrong winner in a frontier-model race that, from a pharmaceutical company's vantage point, is still being run.
The financial backdrop explains the urgency. Novo reported second-quarter 2026 net sales of DKK 78.5 billion, up 3% at constant exchange rates, with adjusted sales up 7% and adjusted operating profit rising 11% to DKK 33.4 billion, helped by GLP-1 volume growth and favorable U.S. rebate adjustments. But reported operating profit fell 16% at constant rates to DKK 27.1 billion, hit by non-cash impairments to pipeline assets and a prior-year reversal of a U.S. 340B rebate provision. For the full year, the company now expects adjusted sales and adjusted operating profit to be flat to down 6% at constant rates, with free cash flow of DKK 45 billion to DKK 55 billion and capital expenditure of around DKK 55 billion.
That guidance is the real story behind the partnership flow. Novo is navigating U.S. pricing pressure, most-favored-nation agreements with the U.S. administration, and loss of exclusivity for semaglutide in some international markets, even as Eli Lilly continues to gain ground with tirzepatide. Against that backdrop, AI is the growth narrative Novo can still control. The company has said it wants to become the world's most AI-driven healthcare company, and this week's deal with Anthropic is the most explicit signal yet that it intends to make good on the phrase.
"This partnership is another testament to our ambition to become the world's most AI-driven healthcare company, building on Novo's current AI initiatives with other technology partners," said Mike Doustdar, president and chief executive officer of Novo. "Joining forces with Anthropic will supercharge our R&D organisation and help us on our mission to bring new, transformative health solutions to people living with chronic diseases."
Novo's research budget gives the AI push real scale to work with. The company spent DKK 52.0 billion on R&D in 2025, equal to 16.8% of sales, and second-quarter 2026 R&D costs rose 2% to DKK 11.5 billion. If AI can lift the yield of even a fraction of that spending — more viable candidates per krone, faster iteration between hypothesis and experiment — the option value is substantial. But option value is not earnings, and the market will eventually demand evidence that the partnerships are producing assets, not just announcements.
Why Anthropic, and Why Now
From Anthropic's side, the timing is equally revealing. The company has raised roughly $130 billion ahead of an initial public offering that market participants expect as early as late September or early October 2026. A high-profile collaboration with one of the world's largest drugmakers by market value is precisely the kind of enterprise validation that supports a blockbuster IPO narrative. Anthropic has been building deliberately into life sciences: it launched Claude for Life Sciences, expanded into Claude for Healthcare, signed a collaboration with Genmab in January 2026 to build agentic AI for clinical development, and acquired Coefficient Bio, an AI drug-discovery startup founded by former Genentech scientists, for about $400 million.
"AI's increasing capability brings with it the potential to compress a century's worth of biological and medical breakthroughs into a decade," said Dario Amodei, co-founder and chief executive officer of Anthropic. "By giving leading researchers access to safe, capable, and trusted frontier models, we can shorten research timelines, improve outcomes, and discover medicines and treatments that significantly improve human life."
Amodei's language — compressing a century into a decade — is the kind of ambition that sells an IPO. It is also, on its own, a claim with no near-term proof attached. The useful question is not whether AI will eventually transform drug discovery; most of the industry already believes it will. The useful question is whether Novo's particular stack of partners converts into a measurable productivity advantage before its pricing and patent headwinds convert into a measurable earnings disadvantage.
The Mechanism: Where AI Actually Touches the Pipeline
To judge the deal, it helps to separate the drug-development pipeline into the parts AI can reach and the parts it cannot. A new medicine typically takes 10 to 15 years and well over $1 billion to reach approval. The early stages — target identification, hit finding, lead optimization — are information-intensive and hypothesis-driven, exactly the kind of work a reasoning model can accelerate. A science-specialized model can triage the literature, surface non-obvious targets from omics and clinical datasets, propose molecular hypotheses, and help computational teams iterate faster. Agentic software engineering, the second stated aim of the Anthropic deal, shortens the internal tooling cycle, which compounds into faster experiments.
There is already a hint that Novo is seeing productivity gains at the administrative end of the pipeline. In the AWS partnership, the company reported that AI had reduced clinical-documentation time and that more than 25,000 employees were using AI tools. That is a concrete, measurable win — but it is a cost and speed win, not a discovery win. It improves the denominator of R&D productivity without guaranteeing anything about the numerator.
The hard constraint is downstream. Phase 3 clinical trials and regulatory review consume the majority of a drug's timeline and cost, and no language model compresses patient recruitment, follow-up, or a Food and Drug Administration review clock. AI can improve trial design and patient stratification — the Genmab collaboration is explicitly aimed at that — but it cannot eliminate the biological uncertainty that makes late-stage trials fail. The honest mechanism, then, is bounded: AI can raise the hit rate and shorten the front end of the funnel, but the back end of the funnel still moves at the speed of human biology and regulation.
That boundedness matters for valuation. If the market prices Novo's AI stack as a near-term earnings driver, it is pricing a story the pipeline cannot deliver on this cycle. If it prices the stack as a long-dated real option on R&D yield, the framing is more defensible — but options that far out are volatile assets, and they do not offset a guidance cut in the current fiscal year.
The Counter-Thesis: AI Has Not Yet Delivered a Single Approved Drug
The strongest case against the bullish read is simple and empirical: after more than a decade of AI-drug-discovery hype and billions of dollars of investment, not one AI-designed medicine has reached regulatory approval. Companies such as Recursion Pharmaceuticals, Exscientia, and Insilico Medicine have spent years building AI-native discovery platforms. Google DeepMind's Isomorphic Labs has signed deals with Novartis and Eli Lilly carrying a combined potential value of nearly $3 billion, excluding royalties. The capital is real; the approved drugs are not.
Applied to Novo, the counter-thesis sharpens. The company's near-term earnings problem is not that its scientists cannot generate hypotheses fast enough. It is that realized prices for GLP-1 therapies face political and competitive pressure, that semaglutide exclusivity is eroding in some markets, and that Lilly is executing aggressively. None of those problems yields to a better reasoning model. "The world's most AI-driven healthcare company" is a strategic aspiration, not a margin bridge, and a string of technology partnerships does not repair a 2026 guidance range of flat to down 6%.
The counter-thesis is correct on timing, but it mistakes the nature of the bet. Novo is not claiming that Anthropic will fix next year's earnings. It is assembling optionality across the model stack because no single vendor has yet proven decisive in pharma. The first measurable wins from AI in drug discovery are almost certain to be productivity gains — documentation, coding, literature triage, trial design — rather than novel approvals, and Novo's own AWS disclosure points in exactly that direction. The partnerships are exploratory by construction; stacking them is the rational response to an exploratory technology.
The falsifying signal is specific. If, by the end of 2027, Novo cannot point to at least one AI-nominated or AI-advanced candidate moving into Phase 1 trials, or if R&D productivity — measured as viable candidates advanced per krone of R&D spending — fails to improve while R&D intensity remains above 16% of sales, then the AI transformation is mostly narrative. Investors should treat the partnership announcements as the beginning of a measurement period, not as evidence of a result.
What Comes Next: Three Horizons
The implications split cleanly by time horizon, and collapsing them into one verdict is the most common mistake.
In the short term, the sentiment channel is supportive. Novo's shares have recovered through 2026 and were up roughly 13% year to date as of September 16, trading in the mid-$40s after falling more than 50% from their 2024 peak. A steady drumbeat of AI partnerships keeps the stock in the technology-adjacent conversation and can sustain a modest re-rating even while fundamentals lag.
Over the medium term — the 2026 to 2027 window that the current guidance covers — fundamentals will be driven by GLP-1 volume growth against pricing pressure, not by AI. The AI contribution to earnings in this window is likely to be negligible, and the stock will respond to quarterly sales, rebate dynamics, and competitive share shifts with Lilly far more than to any Claude workflow.
Over the long term, the structural question dominates. If AI lifts R&D yield even modestly across a DKK 52 billion annual research base, the compound value of an improved hit rate is enormous — a single additional blockbuster therapy would dwarf the cost of the entire partnership stack. That is the structural case, and it is the only one that justifies treating these deals as more than public relations.
Three scenarios frame the path. In the base case, the partnerships deliver measurable productivity gains and one or two AI-assisted candidates enter clinical testing by 2027 or 2028, supporting a gradual re-rating. In the upside case, Claude Science or another partner model surfaces a novel target that becomes a genuine clinical asset, and Novo becomes the benchmark for AI-native pharma. In the downside case, AI delivers only incremental productivity while pricing and patent pressures dominate earnings, and the partnership stack is remembered as expensive narrative.
What to watch is concrete. Novo holds a Capital Markets Day on September 21, 2026, where management will review strategy and operations; any detail there on AI milestones would be the first real calibration of the ambition. The company reports third-quarter results on November 4, 2026. Anthropic's IPO filing and pricing — expected in the coming weeks — will test how much of the pharma-validation narrative is already reflected in its own valuation. And the falsifying signal above is the one that matters most for the structural thesis.
Novo is not buying AI to fix next year's earnings. It is buying insurance against becoming the incumbent that a faster, AI-native rival out-discovers — and the premium on that policy will only be worth paying if the partnerships produce assets, not just announcements.
Explore more exclusive insights at nextfin.ai.
