Europe’s Biotech Asymmetry

If China owns the cheap-fast clinical loop, Europe needs a different kind of learning asset

Europe is unlikely to win a brute-force speed-to-first-in-human race against a Chinese system built around regulatory delegation, dense clinical infrastructure, cluster-level operating support, and low-cost execution. But that does not mean European biotech is strategically disadvantaged everywhere. It means the European asset has to be constructed differently. Part 3 of Beyond the Model.

The interesting European companies are not slower versions of Chinese or US comparables. They are companies built around forms of evidence the speed race does not naturally produce: higher-fidelity preclinical systems, registry-linked human data, regulator-defined evidence pathways, and early US capital and BD legibility. The asset is not any one of those components. It is the stack.

In 2019, Irv Weissman’s lab at Stanford published CD24 as a novel immuno-oncology target. By 2022, leading US life-science investors had backed a $76 million Series A for Pheast Therapeutics, the spinout built around translating that paper into a clinical asset. By 2023, Antengene, a Chinese biotech, was running a CD24 programme in clinical trials at four US sites, before Pheast had filed an IND. The story, recounted by Jacopo Gabrielli, Kush Desai, and Jamie Rintoul, is American [1]. The same pattern is now confronting European venture-backed biotech, often less visibly but with the same implication: discovery priority is no longer enough.

The pattern matters because it shows what the modern race in biotechnology increasingly looks like. The bottleneck is not always discovery, IP, or seed capital. It is the time and cost required to move from a credible preclinical asset to first-in-human evidence. The first two pieces of this series argued that AI-bio is bottlenecked twice: upstream by predictive validity and downstream by clinical throughput. China has compressed the second bottleneck, and in some categories has built enough preclinical and clinical iteration capacity to make the output commercially valuable to Western pharma.

This piece asks the geographic question that follows. Once you accept that AI-bio value depends on predictive validity and clinical throughput, the next question is where those learning loops can actually be built. China has industrialised one answer: the cheap-fast clinical loop. The US is trying to rebuild a domestic version of that loop. Europe needs a different answer.

The European companies I would want to back are those built around categories where the speed race is less decisive: higher-fidelity biology, registry-linked human data, regulator-defined evidence pathways, and early US capital and BD legibility. A preclinical platform with credible registry linkage and a defined regulatory niche, capitalised through US co-leads, is a different bet from a European preclinical platform that is just slower than its US or Chinese comparable.

What China has actually built, and what Western pharma is paying for

A recent BioPharma Drive analysis of Chinese biopharma found that 41 of 57 named 2025 cross-border outlicensing deals were at preclinical or Phase 1 stage, worth $46 billion in total deal value [2]. The pattern is no longer anecdotal. Western pharma is increasingly buying access to Chinese-origin assets before they are clinically mature, effectively paying for the output of a system that can generate and advance many programmes quickly and cheaply.

Two mega-deals show the scale of the shift. In May 2026, Bristol Myers Squibb signed a strategic agreement with Hengrui worth up to $15.2 billion in total potential value across thirteen early-stage programmes in oncology, haematology, and immunology [3]. In July 2025, GSK announced a separate Hengrui partnership worth up to $12 billion across respiratory, immunology and inflammation, and oncology [4]. Hengrui is not an AI company. It is an industrial pharmaceutical company with the operating capacity to move multiple programmes through development at high throughput and relatively low cost. Western pharma is paying for that capacity.

The Chinese system is also proving it can produce category-defining assets, not only cheaper versions of established ones. Akeso’s PD-1/VEGF bispecific antibody, ivonescimab, beat Keytruda head-to-head in a Phase 3 trial in PD-L1-positive non-small-cell lung cancer, reducing the risk of disease progression or death by 49% versus pembrolizumab [5]. But the advantage is uneven. China is strongest in categories where speed, scale, engineering iteration, and clinical infrastructure compound; it still trails materially in areas such as neurology and psychiatry, where translational uncertainty is higher [6]. It has built a powerful speed-and-scale engine. It has not solved every kind of biology.

The deeper point is that these deals are not just asset transactions. They are transactions in early human evidence. That is the argument Jacob Becraft, CEO of Strand Therapeutics, made in March 2026 testimony to the US House Select Committee on the CCP [7]. In his framing, the defining competitive metric in modern biotechnology is no longer who discovers a promising therapy first, but who can turn that discovery into first-in-human clinical data fastest. The country that generates early human data most efficiently is where the rest of the biotech ecosystem grows.

That is the AI relevance. AI-bio does not only need more generated candidates; it needs more independent human readouts. A thousand measurements on one programme can tell you a lot about that programme. Early human readouts across many programmes tell you something different: which mechanisms translate, which biomarkers fail, which toxicities repeat, and which preclinical signals were actually worth trusting. The model needs repeated contact with human outcomes across enough independent experiments to separate signal from proxy.

Becraft’s harder point is that early human data creates an industrial flywheel. Once a therapy produces early clinical results, the programme is de-risked, investors step in, partnerships form, companies grow, manufacturing capacity expands, clinical centres gain experience, and the cycle repeats. If that loop moves offshore, the ecosystem moves with it. This is why first-in-human data is not just a clinical-operations detail. It is industrial capacity.

The rare-earth analogy is useful if it is understood precisely. The analogy is not that clinical trials are commodities. It is that China has repeatedly built strategic leverage by scaling the capacity layer that Western markets were happy to outsource because it looked low-margin, operationally difficult, or environmentally inconvenient. Rare-earth processing is the obvious example, but the same pattern appears in solar, batteries, and electric vehicles: subsidise the enabling layer, build scale, compress costs, absorb the operating pain, and then move up the stack once Western capacity has weakened.

Early human evidence could become the biotech version of that mistake. If first-in-human trials become much faster and cheaper in China, Western companies will rationally use that capacity. Investors will fund assets that already have early clinical signal. Pharma will license programmes that have crossed the first human-data threshold. Trial sites, manufacturing expertise, translational talent, and company formation will follow the flow of capital. Over time, what looked like efficient outsourcing becomes dependency on the system that generates the evidence.

That is why the flywheel matters. Manufacturing sectors took decades to decay because they were supported by revenue, assets, real estate, and capex. R&D-stage biotechs are more fragile. They are often pre-revenue and dependent on investor or partner capital. If that capital shifts toward Chinese competitors and licensed Chinese assets, the flywheel can weaken over years, not decades. The risk is not simply that China supplies cheaper trials. It is that China becomes the place where early clinical evidence is produced, priced, and compounded.

The US response is therefore to rebuild the domestic first-in-human loop: faster early-stage clinical initiation, CMC and IND requirements matched to development stage, and better alignment between manufacturing and trial sites. Europe’s question is different. It has pieces of that system but not the system itself, and the institutional preconditions for building the rest sit outside Europe’s policy reach. The more interesting European question is not how to build a slower version of China’s speed machine. It is what kinds of biotech assets are valuable precisely because their value does not depend on winning the cheap-fast first-in-human race.

Speed is a system, not a regulator

The mistake in comparing China, the US, Australia, and Europe is to reduce the speed advantage to regulatory timelines. Regulation matters, but speed-to-first-in-human is not just a regulator’s decision. It is an ecosystem property.

In a recent analysis of Chinese biotech infrastructure, Cam Watson describes the Chinese system as a set of interacting layers that absorb operational risk on behalf of incoming companies [8]. The important point is not that China has more state support in the abstract. It is that the system absorbs risk that, in a Western context, would usually sit on the company’s burn rate or cap table. In Suzhou’s BioBAY, for example, incoming biotech companies can receive rent support, equipment access, regulatory-filing help, shared CRO platforms, and introductions to government-aligned capital. The cluster does not merely host companies. It carries part of the operating burden that would otherwise slow them down.

That is why “just speed up the regulator” is not enough. The speed race requires at least four conditions to be contestable: a regulator with authority to delegate, a clinical-trials market with meaningful scale and local IRB capacity, cluster infrastructure that absorbs operational risk before revenue, and patient capital that can fund long-cycle bets through public or strategic exits.

China has built all four. Australia has the first two and a simpler early-phase trial pathway. The US has fragmented but powerful versions of all four: deep capital markets, major clinical centres, FDA centrality, and dense biotech clusters, even if the first-in-human pathway is slower than it could be. Europe has pieces of the system, but not the system itself.

The UK is a partial exception. Post-Brexit, the MHRA has tried to position itself as a faster serious regulator, and ILAP gives companies addressing significant patient need a target development profile that aligns the regulator, NICE, and the NHS around the evidence package before the company has finished generating it. The O’Shaughnessy Review also set out a credible reform agenda around site activation, contracting standardisation, and using NHS data to recruit eligible patients more quickly. Those are real advantages, but they are UK-specific and do not amount to a continental speed machine.

Continental Europe has scientific depth and cluster density, but not China’s coordination architecture. The EMA coordinates; it does not delegate in the way China’s system or Australia’s CTN pathway can. European trial populations are large in aggregate, but fragmented across countries, sponsors, sites, languages, and reimbursement systems. European biotech capital exists, but the depth of risk capital for early platform bets remains smaller than in the US. The result is not that Europe has nothing. It is that Europe has components that do not naturally assemble into a speed-to-human machine.

The tempting response is to say Europe should fix all of this: harmonise clinical trial initiation, build EU-wide cluster strategies, push more patient capital through Horizon Europe and the European Innovation Council, and make EMA pathways more usable for early-stage companies. None of that is wrong. But it is unlikely to produce a European BioBAY, and it will not reproduce the Chinese local-government incentive architecture where officials are rewarded for company creation, tax generation, and ecosystem growth. Europe’s institutions are not built that way. Trying to out-China China on European institutional ground is unlikely to close the gap on the timescale the race is running.

The geopolitical lever could change the relative economics. The BIOSECURE Act, signed into law in December 2025, restricts US federal contractors from working with named Chinese biotech CDMOs and sequencing companies [9]. Alongside explicit no-China provisions in some US LP agreements, it points toward a capital market in which Chinese clinical infrastructure may become harder for Western investors and pharma companies to underwrite. If that hardens, the value of alternative sources of de-risked clinical evidence rises. European biotech is one of those alternatives. The European thesis is stronger in a world of continued US-China decoupling, and weaker in a world where Western capital remains fully comfortable pricing Chinese clinical infrastructure into its models.

The compound asset, with the US bridge load-bearing

The European companies worth investing in are not slower versions of the global race. They are built around categories where Europe is structurally advantaged. The differentiator is the discipline of constructing the asset to be transactable with US capital and US pharma BD from the seed round, not retrofitted at Series B.

European biotech pitches tend to be calibrated to European LPs and European pharma sponsors who, on average, accept thinner predictive-validity packages, journal-defined milestones rather than regulator-defined ones, and a narrative that treats US-style commercial outcomes as a future-stage problem. This is not universal; the best European companies already build past it. But the pattern is familiar: a company anchors its early investor base, scientific advisory board, and BD relationships locally, then attempts a transatlantic rebuild at the point where US growth capital becomes necessary. The rebuild is expensive in cycle time, which is the most valuable thing the company has.

The strongest European-origin companies are often not purely European companies in the operating sense. The science may start in Europe, but the globally valuable company is built transatlantically, with US capital markets, US clinical credibility, US advisors, or US commercial partners integrated early rather than bolted on late. BioNTech and CRISPR Therapeutics are different companies with different histories, but both illustrate the same broad pattern: European scientific depth became globally valuable through structures that could access US-scale capital, US pharma partnerships, or US commercial infrastructure before the platform was fully validated. The lesson is not that every European company needs to become American. It is that US legibility cannot be an afterthought.

US co-leads on early rounds, US scientific advisors named on the seed deck, US BD relationships warmed before they are needed, and predictive-validity packages benchmarked against US clinical comparators rather than European registry literature. None of this dilutes the Europeanness of the underlying asset. It is the condition for the underlying asset to be capitalised at the price its specifically European inputs justify.

Three other components stack on top of that bridge.

A fidelity advantage. Europe has unusual density in patient-derived organoid systems and microphysiology, and the predictive-validity argument from Piece 1 has more room to compound where those systems already exist. This is narrower than the broader claim that European translational biology beats American translational biology. It does not, at least not on volume. But it is substantive. The Dutch organoid programme built around the Clevers lab and HUB Organoids has published clinical-concordance work showing patient-derived organoids predict patient response in cystic fibrosis CFTR modulator therapy and in colorectal cancer chemotherapy regimens. That is the kind of predictive-validity evidence the rest of the field has not generated at the same depth, and it sits in Europe.

The bet is not “European biology is better.” The bet is that European platforms with this kind of evidence are systematically under-capitalised relative to the implications for Phase 2 success rates if the evidence holds.

A regulatory niche. EMA PRIME and MHRA ILAP are expedited regulatory pathways designed for therapies addressing unmet medical need. Both are relevant to Advanced Therapy Medicinal Products: the EMA-defined category covering cell therapies, gene therapies, and tissue-engineered products. These are the modalities where the speed race works less well, because manufacturing is bespoke, IND burden is heavy, and population-specific data matters more. Biologics dominate Chinese licensing deals partly because CDMO infrastructure is reusable across biologics and the IND burden is lighter. ATMPs, rare disease, and paediatric biology are the kinds of categories where Europe’s advantages are more likely to matter, because the asset depends less on generic speed-to-IND and more on trusted clinical networks, longitudinal patient data, regulatory design, and population-specific evidence.

A clinical-data advantage. The NHS, the Nordic registries, UK Biobank, FinnGen, and the German statutory insurance cohorts are phenotype-linked human data assets at population scale: longitudinal health records and genetic data tied together at individual-patient level, which is what makes them usable for precision medicine drug development. China is racing to build this. So far it has population-scale DNA sequencing but, on its own assessment, not the linked phenotype data that would make the sequencing useful for drug development. In a 2025 Nature interview, Rao Yi, one of the leading reformers of Chinese life-sciences research, was unambiguous about the gap: “China is lagging behind the United States in precision medicine because China is weak in human-genetics research” [10].

Europe is closer to the US position than to the Chinese one on this specific axis. What Europe has, in certain markets, is not simply more data, but public-system longitudinal data assets with unusually strong patient linkage and institutional legitimacy. Genomics England, UK Biobank, FinnGen, the Nordic registries, and the NHS are not interchangeable with each other, and none is frictionless to access. But they point to a different form of learning loop: one built less around cheap-fast first-in-human trials, and more around linked human evidence that can shape patient selection, endpoint design, natural-history comparisons, and post-approval evidence generation.

The point of stacking these together is that no single component is the asset. US capital legibility without European-specific evidence is just financing strategy. Registry data without regulatory design is just an underused public asset. Predictive-validity evidence without a route to US pharma BD is a scientific proof point trapped in a European funding market that may not pay for it. The investable asset is the combination.

A preclinical platform with credible registry linkage and a defined regulatory niche, capitalised through US co-leads, is a different bet from a European preclinical platform that is just slower than its US or Chinese comparable. In Europe, the iteration loop can close through registry-linked endpoints and ATMP or rare-disease regulatory pathways rather than through pure speed-to-IND. The structure of the iteration is different, but the principle is the same.

What this means for European biotech

For investors, the diligence question on European biotech has shifted. The relevant comparison is no longer only European platform versus Boston platform. It is whether the asset sits in a category Chinese clinical infrastructure structurally does not compete in, and whether the company has been built to be transactable with US capital long before US capital becomes necessary.

That creates a pricing distinction. European-specific assets that combine fidelity, data linkage, regulatory niche, and US legibility should be priced higher than the local market often prices them. Imitative European assets that compete on speed but lack the system to win the speed race should be priced lower. The geopolitical tailwind from BIOSECURE and US-China decoupling matters, but it is conditional on those policy moves holding. It should amplify the thesis, not substitute for it.

For operators, the compound asset cannot be retrofitted at Series B. It has to be designed in. That means explicit ILAP or PRIME strategies from the seed round, NHS or registry data partnerships before Series A, US co-leads and US scientific advisors named early, US BD relationships warmed before they are needed, and predictive-validity packages benchmarked against US clinical comparators. Pick the European fight on purpose.

For policymakers, the implication is narrower than “build a European BioBAY.” Brussels does not need to out-China China. It needs to make the European registry, regulator, and translation layers usable as a coherent stack so that companies who do this construction work can reach a Phase 2 readout that is diligence-grade for US growth capital. The opportunity is not to copy the Chinese speed machine. It is to make Europe’s slower but higher-fidelity evidence infrastructure more usable, more legible, and more financeable.

The series

This is the final piece. The first argued that AI in drug discovery is bottlenecked upstream by predictive validity. The second argued that the downstream constraint is clinical throughput. This piece argues that geography now matters because different ecosystems close the learning loop in different ways.

China has built the cheap-fast clinical loop. The US is trying to defend and rebuild its domestic version of that loop. Europe’s opportunity is not to imitate either at smaller scale. It is to build assets around the categories the global race structurally does not reward: higher-fidelity biology, linked human data, regulator-defined evidence pathways, and transatlantic capital construction.

AI-bio value will not accrue only to the best model. It will accrue to the system that generates the best feedback. Europe’s task is to build the kind of feedback loop it can actually own.