The Throughline

Radiopharmaceuticals: sixteen years of deal data on where the field goes next

Buyers have concluded that radiopharmaceuticals are a platform business, and they are paying platform prices. Takeout premiums have roughly doubled since Novartis set the benchmark in 2017. The underlying mix of what gets bought has not moved in nine years.

Athithi Verma  ·  8 September 2026  ·  Monthly edition  ·  Synopulse

Seven quiet years, then a step change that has held for nine

Dealmaking ran at four transactions a year through 2016 and above twenty every year since.

Radiopharmaceutical dealmaking ran at four transactions a year from 2010 through 2016. In three of those years it was three. The field had approved products, a functioning supply chain and almost no corporate interest.

From 2017 the average has been above twenty a year. That is a fivefold step change, and it has now persisted through nine consecutive years including a funding downturn that cut deal volume across most other modalities.

The inflection point is precisely dated. Novartis acquired Advanced Accelerator Applications in October 2017 for $3.9bn, then Endocyte fourteen months later for $2.1bn. Those two transactions moved radioligand therapy from a European speciality business into a category that every large oncology franchise had to take a view on.

Everything after 2017 is the industry responding to a decision Novartis made on its behalf.

Novartis paid 47% and 54%. Everyone since 2023 has paid at least 87%.

Nine transactions disclose a one-day premium, and they split cleanly by era.

Every platform buyout since 2023 has cleared 87%, and three of the four cleared 97%.

Scarcity is the straightforward explanation. The number of clinical-stage radioligand platforms with proprietary isotope access is finite, and the pool shrank with each acquisition. Bristol Myers Squibb paying 104% for RayzeBio in 2023 and AstraZeneca 97% for Fusion in 2024 were not independent decisions. They were consecutive bids into a thinning market.

The alternative explanation is less comfortable. Novartis bought at 47% and got Pluvicto and Lutathera. No asset acquired at a 100% premium has yet produced anything comparable. That question resolves over the next three years and it is the most consequential open item in the field.

Theranostics is the story the field tells. The deal data does not show it winning.

The paired diagnostic share stood at 42% in 2017 and 39% in 2024.

Pairing an imaging agent with a therapeutic so the target is confirmed before treatment is the argument the field makes for itself, and it is a strong one. Radiopharmaceuticals are the only modality that can see the target before dosing it.

Deal behaviour does not reflect that. The paired share has shown no directional movement across nine years. Buyers are not systematically preferring paired assets to therapy-only ones.

Two readings fit the data. The pairing may be treated as a development decision rather than an acquisition criterion, meaning acquirers buy the therapeutic and build the companion afterwards. Or the commercial advantage is real but not yet priced into what people will pay.

Each has a consequence. If pairing is built rather than bought, the diagnostic assets in this dataset are undervalued against their strategic role. If the advantage is simply unpriced, that is an arbitrage with a closing window.

Specialists run this market and every large company bought its way in

Telix and Novartis appear more often than the rest of large-cap pharma combined.

Telix appears on one side of a transaction twenty times. Novartis, counting Advanced Accelerator Applications, appears twenty-three. Lantheus, ITM, Orano Med and Curium follow. Only after those names do large diversified companies enter the table.

One pattern holds across every large company in the table. None built a radiopharmaceutical capability internally. Each bought one: Novartis in 2017, Lilly and Bristol Myers Squibb in 2023, AstraZeneca in 2024. The premium data is the price of that choice.

The 2026 development is the arrival of financial buyers. Curium’s $8bn take-private of Lantheus is the largest transaction in the dataset, and Nordic Capital’s purchase of BWX Technologies’ medical business sits alongside it. Sponsors consolidating infrastructure implies cash flows legible enough to lever, which describes a different phase of a market.

Announced values cleared $2bn while upfronts fell from 30% to 2%

Committed capital is concentrating in acquisitions, not licences.

In 2014, Merck KGaA’s agreement with Pfizer carried an $850m upfront against $2.85bn announced, roughly 30% committed at signing. Recent licensing deals of similar headline size commit a fraction of that.

Across the full dataset the median announced value is $259m and the median upfront $56m. Acquisitions pay in full at closing, while licensing structures carry progressively more of their value in milestones that may never be earned.

For anyone reading announced values as a proxy for conviction, the gap matters. A $2bn licence signed in 2026 represents materially less commitment than a $2bn licence signed a decade ago, and league tables built on announced value will overstate recent activity against historical comparisons.

Prostate cancer accounts for more named indications than the next five combined

Thirty of the specified indications are prostate. Alzheimer’s is the only non-oncology cluster.

That concentration is the field’s largest unhedged exposure. PSMA-targeted therapy functions simultaneously as the commercial proof, the clinical proof and the investment case. A safety signal, a reimbursement change or competitive displacement in prostate cancer would reprice the whole modality, including assets with no connection to it.

The Alzheimer’s cluster deserves separating out. Those transactions are diagnostic tracers rather than therapies, and they mark the one area where radiopharmaceutical activity has moved decisively outside oncology. If the field diversifies, the data says neurology imaging is where it starts.

The product decays, so the supply chain is a clinical constraint rather than an operational one

Twenty infrastructure transactions point at what actually limits uptake.

An actinium-225 therapy has to be produced, released, shipped and administered within a window measured in days. No other modality has that property.

The infrastructure transactions in this dataset are where that constraint is being addressed: Curium acquiring Nucleis and Abscint, SHINE taking Lantheus’s SPECT business, GE HealthCare acquiring Nihon Medi-Physics, PharmaLogic buying Agilera. None are drug transactions. All of them determine whether the drug transactions work.

A paired theranostic product requires both chains functioning in the same hospital in the same week, the diagnostic isotope for the scan and the therapeutic isotope for the dose. That dual dependency explains slow launch curves in this field better than pricing does.