The next US blockbuster may be discovered in Shanghai, developed in Boston and launched by a company that did not invent it. That is already the pattern in the deal data. In 2025, Chinese drugmakers signed 157 cross-border licensing deals with a combined headline value of $135.7 billion, against $51.9 billion in 2024. In the first half of 2026 they signed about $110 billion more.
Most commentary treats this as a story about R&D productivity or about geopolitics. Both matter, and I will get to the politics. But I think the more immediate consequence lands on the commercial side. When the molecule, the evidence and the people who understand it all start somewhere else, the company that launches it knows less about its own product than launch teams used to, and it has less time to learn. Commercial readiness, not discovery, becomes the bottleneck.
The headline values are huge, and the cash is much smaller
Start with what the numbers actually say. PharmCube, the Chinese data provider behind the most-cited tallies, counted $135.7 billion of total deal value from mainland China in 2025, but only $7.0 billion of it was paid up front. The rest is milestones and royalties that arrive only if the drug succeeds. Roughly 5 cents of every headline dollar is cash on signing.
The individual deals tell the same story. Pfizer paid 3SBio $1.25 billion up front in 2025 for ex-China rights to a PD-1/VEGF bispecific antibody, with up to $4.8 billion more in milestones; at the time, a record upfront for a Chinese asset. GSK paid Hengrui $500 million up front in July 2025 for a COPD candidate plus options on 11 more programs, with about $12 billion in potential milestones. In May 2026, Pfizer agreed to pay Innovent $650 million up front for a portfolio of 12 early-stage cancer programs, with up to $9.85 billion in milestones.
The share is what makes it structural. Evaluate expected Chinese-origin assets to account for nearly 40% of global licensing deals in 2025. The IQVIA Institute reported in March 2026 that deals between China-headquartered and international companies reached an all-time high in 2025. I read this as big pharma buying a large number of options at a modest premium. That is a rational response to a patent cliff that needs many replacement assets, and a reasonable price to pay for speed.
The inventor, the developer and the launcher are now different companies
For most of the industry's history, one company discovered a molecule, ran its trials, wrote its label and launched it. The commercial team grew up alongside the asset. Medical affairs had sat in on the trial design. The key opinion leaders had usually been investigators. By launch, the brand knew its product the way a founder knows a company.
That sequence is coming apart. In the new pattern, one company discovers the molecule and generates early data, often in China. A second runs the global trials. A third, sometimes the same as the second, owns the US label and the brand. Each handoff loses context. I think of the result as a split brand: a product whose knowledge was created in one place and whose revenue has to be earned in another.
The pool of candidates for this pattern is growing. The IQVIA Institute counted thirty new active substances launched only in China in 2025. Some of those will be licensed for other markets, and they will arrive with a Chinese label, Chinese prescribing experience and Chinese clinical experts, but no US physicians who have used them. The launcher starts with the molecule and very little of the context around it.
Takeda's Fruzaqla shows how fast the split can move. Takeda licensed fruquintinib from Hutchmed in January 2023, with $400 million up front, and FDA approved it for metastatic colorectal cancer in November 2023. The approval rested on FRESCO, a trial run in China, and FRESCO-2, a larger multiregional trial. Takeda had less than ten months between signing and a US label for a molecule it did not discover.
Summit Therapeutics is a sharper version. It licensed ivonescimab, a PD-1/VEGF bispecific, from Akeso in December 2022 for $500 million up front and up to $5.0 billion in total. FDA accepted Summit's application in January 2026 on the basis of the multiregional HARMONi trial. As of October 2026 the decision is pending, with a goal date of November 14.
The launcher inherits evidence it did not design
The hardest inheritance is the evidence. FDA has been explicit that data from one country does not automatically apply to US patients. In February 2022, its oncology advisory committee voted 14 to 1 that Lilly and Innovent's sintilimab needed more trials before approval, because the registrational ORIENT-11 study had enrolled 397 patients, all in China. FDA issued a complete response letter in March 2022 and recommended a multiregional trial against the US standard of care.
The industry learned that lesson, which is why the better deals now come with global trials. But even a successful multiregional program leaves the launcher with evidence designed by someone else, for someone else's regulatory strategy. The comparator, the endpoints and the patient mix were chosen years earlier, by people who now work at a different company and often in a different language. Every promotional claim the US brand makes still has to be consistent with the US label, and the brand team has to build that claim set from a dossier it did not write.
This is where I would focus. A usable claims library for an in-licensed product cannot be a translation of the originator's publications. It has to be rebuilt from the US label and the registrational data, with each claim linked to its reference and checked by people who understand both datasets. That is slow, specialist work, and it is usually invisible in the deal model.
Europe adds a second test. Since January 2025, new cancer medicines filed with the EMA go through an EU joint clinical assessment, and orphan medicines join in January 2028. Many of the largest China-origin deals are in oncology. An evidence package built for China and the US now also has to stand up at EU level.
Molecules are abundant now, and readiness is scarce
I came to this industry from ad tech, and the closest parallel I know is what happened to advertising inventory once it could be bought from almost anywhere. Supply stopped being the constraint. The value moved to whoever could verify what was being bought and package it into something a buyer would trust. I think drug licensing is going the same way. Promising molecules are no longer scarce. The ability to turn a foreign dossier into a credible, compliant US brand quickly is.
Crowding makes this sharper. Pfizer's 3SBio asset and Summit's ivonescimab are both PD-1/VEGF bispecifics that started in Chinese labs. When several large companies license similar mechanisms from similar sources, the molecules will not be what separates them. Evidence, label, payer story and how quickly each launcher can explain its product to physicians will.
The time between deal and launch also shrinks when an asset is licensed after meaningful human data exist, as Fruzaqla was. Ten months from signing to approval does not leave years for the launcher's medical team to learn the data, for US investigators to build experience, or for content to go through several review cycles. The work that used to happen in parallel with development now has to happen after the deal, under a deadline.
The field feels this first. US oncologists who never enrolled a patient in the registrational trials have no personal experience of the drug, so the medical team has to build scientific relationships from zero. Review teams feel it next. Medical, legal and regulatory reviewers are used to checking claims against references their own company generated. Now they are checking claims against a translated clinical study report, a foreign publication and a US label negotiated by a partner, and every one of those references has to be traceable before an asset clears the review process. None of that is unusual work. There is simply more of it, on less familiar material, in less time.
The risk view deserves a serious hearing
The case against leaning on China is not paranoia, and I want to state it fairly. In September 2025, the New York Times reported a draft executive order that would have sent licensing deals with Chinese companies to CFIUS, the interagency panel that reviews foreign investment, raised FDA scrutiny of data generated in China and added fees to applications that rely on it. The argument behind it is serious: that the US risks handing a strategic rival the early stage of its own innovation base. As of early October 2026 that order has not been issued, and plain licensing deals remain outside the main federal restrictions. Policy can change much faster than a development program.
Congress has already acted on part of the supply chain. The BIOSECURE provisions, signed into law with the 2026 defense authorization in December 2025, bar federal agencies, contractors and grant recipients from buying biotechnology equipment or services from designated biotechnology companies of concern. The law does not govern licensing a molecule, and the government's initial list is not due until December 2026. But many in-licensed programs depend on contract manufacturing and testing partners, and a launcher with federal contracts will need to know where its product is made.
My view is that the deals will keep coming, because the science is good and the economics work. I am less sure about the policy path than about anything else in this essay. The sensible response is not to avoid Chinese assets. It is to price policy risk into the plan: a second manufacturing source, deal structures that can survive a review, and evidence strategies that do not depend on any single country's data.
Run the commercial diligence before you sign
The practical shift is to move commercial readiness into the deal itself. Business development already diligences the science, the IP and the manufacturing. It should also ask whether the evidence supports the US claims the brand will need, how long it will take to build a reviewed core content set, and who on launch day will understand the data well enough to defend it to a physician or a payer.
I would start the claims and evidence work the week the deal is signed, not after the filing. That means getting the source documents, in English, with their provenance, and building the US claim set from them while the knowledge transfer is still fresh. Agreements that secure access to the originator's clinical and medical teams for the first years matter as much as the rights themselves, though the right structure is a question for each company's lawyers.
Three questions are worth asking before any China-origin deal closes. Can we write the US claims we need from this evidence? Who will know this molecule best on launch day, and do they work for us? What happens to the launch if Washington changes the rules halfway through? A team that can answer all three is buying an asset. A team that can't is buying an option it may not be able to exercise.
In this market, the company that invents a medicine and the company that has to explain it to physicians are increasingly not the same company. The upfront goes to the inventor. The rest of the value goes to whoever learns the molecule fastest.