The US-China split now runs through ownership, not molecules

To compete with Chinese biotechs, you have to run as fast as they do, which takes a bigger round and more risk, or build something Beijing won’t let them sell. 

How worried should a US biotech be about Chinese competitors? It was one of several topics, alongside AI and cross-border biotech, at a small panel ZAKA hosted in Singapore last week with the Central and Eastern European Chamber of Commerce, joined by Si Hui Tan, who runs the Singapore site and pipeline of our portfolio company Cirrus Therapeutics.

The answer depends on what you sell. If you have one drug, you should be very worried. If you have a cell, gene therapy or siRNA platform, you can worry less than you did a month ago. In these fields pharma has often licensed a platform first and paid the big money later, when it bought the company. On 16 October Beijing published draft rules that keep the licensing route open to Chinese platforms, but close the sale option.

Single drugs compete on speed, and China is faster

If you are building one drug, your real competitor is a Chinese team that will reach data in people before you do. At the panel, an attendee who works with Chinese researchers said what struck them most was that Chinese hospitals and research institutions share data that elsewhere stays inside each one. Add the country’s size, and Chinese teams find patients and get answers faster and more cheaply than anyone else. Several people on the panel argued that US biotech still leads in new biology, thanks to the depth of its research ecosystem, and I agree. That lead lasts until the first data in people. Once a target is proven, a Chinese team often gets a better molecule against it into patients first.

In most industries China won on the cost of labour and materials. In drugs it wins on time, and it gets it by doing multiple steps in parallel. Chinese teams start toxicology, manufacturing and trial preparation together, if the drug fails, much of that work is wasted, but it was cheap. McKinsey estimates that early discovery work in China costs a third to half of what it costs elsewhere. A US seed company pays American prices out of a round sized for one milestone, so it does one step at a time.

Pharma is buying the output. According to IQVIA, 40% of the assets big pharma licensed in last year came from Chinese companies. In lung cancer, Akeso’s ivonescimab beat Merck’s Keytruda head to head, and Bristol Myers Squibb paid $1.5 billion upfront to share a similar drug, BNT327, which BioNTech had bought from China’s Biotheus.

Don’t count on Washington to fix this. In August Representatives John Moolenaar and Ben Cline asked the FDA to reject Chinese trial data from sites it hasn’t inspected in the past year. Even if it does, pharma uses that data mainly to decide what to license, and it can repeat the early trials at home.

The damage shows up at the Series A. Pharma can license a Chinese drug that already has data in people and pay most of the price only if it works. In 2025 Chinese companies signed 157 out-licensing deals with a headline value of $135.7 billion, according to PharmCube, but only $7 billion of that was paid upfront. Against that, a Western drug with only animal data is worth less to pharma, and thus also to Series A investors, who price a company off what acquirers will eventually pay.

One way out is to use China’s speed yourself. Several of our portfolio companies already use China for parts of their work, including early trials in Chinese hospitals. Cell and gene therapy has a fast route there. An investigator-initiated trial needs approval from the hospital’s ethics committee instead of the national regulator. EsoBiotec, a Belgian company, got the first human data for its cell therapy this way. Its chief executive, JP Latere, told Endpoints: “We gained probably two years.” AstraZeneca then agreed to buy it for $425 million upfront. For small molecules the gain is smaller, because most of China’s lead comes before the first patient, in fast chemistry against known targets. Either way, the data won’t get a drug approved in the US, and a Chinese team that started first and knows the hospitals will still be ahead.

So price yourself against Chinese assets with data in people, not against other US startups. To match their speed at home, raise a bigger seed and spend it the way they do, running toxicology, manufacturing and trial preparation at once instead of one after another. That burns more cash, and every failed bet costs more than it would in China, so it only pays when the first result in people is worth enough to cover the extra spend. Plan as if no rule will protect you.

Platforms can lose the licence and still win the exit

Chinese platforms are good, and pharma wants them. In August 2025 Gilead’s Kite bought Interius, a Philadelphia company that engineers a patient’s immune cells inside the body, for $350 million. Two months later it licensed the same kind of technology from China’s Pregene, paying $120 million upfront in a deal worth up to $1.64 billion. A Western platform company should expect to lose deals like that one.

What it keeps is the exit. Pharma has often licensed a platform first and bought the company once the field worked. Celgene signed a ten-year partnership with Juno worth about $1 billion in 2015 and bought it for about $9 billion three years later. Novo Nordisk began working with Dicerna on more than 30 targets in 2019 and bought it for about $3.3 billion in 2021. Sanofi partnered with Translate Bio on mRNA vaccines in 2018 and paid $3.2 billion for it in 2021, after Pfizer and Moderna had proved the technology. In each case the partner’s own drugs were still early and the buyer was paying for the platform.

Pharma buys because the purchase brings the next products, the people who run the platform and the manufacturing. In cell and gene therapy (the field Cirrus works in), regulators in effect approve the manufacturing process, so whoever sells the drug has to control it. For a venture investor, the purchase is the step that matters. The licence keeps a platform company alive, but it is the acquisition that returns the fund.

Beijing’s draft removes that step for Chinese platforms. It puts cell therapy, gene therapy and siRNA platforms on the list of technologies that may not leave China, and counts a takeover of the company that owns one as the platform leaving. Licensing a drug, or the use of a platform, stays allowed, but AstraZeneca’s $1 billion purchase of the Chinese cell therapy company Gracell would likely have been blocked.

Washington pushed the same way six days later. The US Treasury’s draft under the COINS Act limits American equity stakes and joint ventures in Chinese biotech and largely leaves licences alone. Since June 2025 the FDA has also refused to let new trials send American patients’ cells to China for genetic engineering unless they meet its standards. Read together, the rules let a Western drugmaker rent a Chinese platform and stop it owning one.

A licence to a Chinese platform now comes without the option to buy, so pharma should pay less for it than for a comparable Western deal. Alejandro Tortosa of The Health Shift newsletter has suggested European platforms could earn a premium for this reason.

Moving won’t get Chinese platforms round the rules. At the panel, someone asked whether Chinese biotech would do what Chinese tech companies have done for years and relocate to Singapore. Shein moved its headquarters there in 2021. Manus, an AI start-up founded in China, did the same in 2025. Meta bought it for about $2 billion in December, though Beijing ordered the deal unwound. The biotech draft counts moving a platform to an offshore subsidiary as an export, and Treasury’s draft treats a Singapore company that Chinese owners control as Chinese.

Pharma will look for ways round it anyway. In early October Novartis paid Abogen, a Chinese mRNA company, $575 million upfront for one drug and took options to license the next ones from Abogen’s platform. mRNA isn’t on Beijing’s list, but the structure gets a buyer much of a platform’s future output without owning the company. Expect more deals like it. They narrow the gap without closing it, because options give pharma neither the team nor the manufacturing.

So build a platform pharma will want to own. Make the path from partner to acquirer, the one Juno and Dicerna took, part of every partnership pitch. Keep manufacturing and key suppliers out of China, even if you run trials there, so the process is something a buyer can take over. And make the technology good enough to own, because pharma can still rent a Chinese one. Discovery and AI platforms get nothing from this. 

How I look at deals now

Most of our portfolio is US biotech, and this has changed how I screen new deals. For a single drug, I ask whether a Chinese team starting tomorrow would reach data in people first, and I benchmark the company against Chinese assets that already have that data. For a cell, gene therapy or siRNA platform, I ask whether pharma will want to own it, and I now give a trade sale better odds than I would for a Chinese peer.

Three things would change my mind: Beijing dropping the ban when it finalises the rule, Chinese platforms winning licences on the same terms as Western ones after the rules take effect, or Washington blocking licences too. The Biotech Investment National Security Act, introduced by Moolenaar and Debbie Dingell in June, would put licensing deals under Treasury review.

I did not always ask these questions, and some of our investments would not pass all of them.

Jan Buza, Partner at ZAKA VC