By 2036, a child born in Europe has a shorter life expectancy than one born in the United States or China. Christina von Messling, Senior Foresight Advisor at Foresight Factory, shares why this is likely to happen - and what can still be done to prevent it.
A 52-year-old woman in Helsinki with a breast cancer diagnosis. Her oncologist prescribes the new AI-developed drug, the one making headlines because it slashed mortality rates in US trials by 34%. But six months in, her tumour is still growing. He checks the trial data. The US cohort. He already knows the answer before he finds it.
The drug is not failing. It is working, precisely as designed, for the population it was designed for. Finnish patients metabolise key compounds differently. The genetic variants that determine how the body processes this class of drug are distributed unevenly across populations, and those distributions were not adequately represented in the training data. She is not an outlier. She is simply Finnish, and that was not part of the model. A screening existed, but it was not done.
Treatment for the general population has not moved and that is the problem. American and Chinese patients have access to a new generation of drugs for the diseases that kill most people: cardiovascular disease, type 2 diabetes, and several of the most common cancers. AI-driven discovery, concentrated in oncology, metabolic disease, and autoimmune disorders, has moved from filing to approval. The pipeline that looked experimental in 2026 is now standard of care. In Houston and Shenzhen. Not in Paris. Europe's pharmaceutical industry is still, in the precise language of its own strategy documents, "transitioning."
This would be serious in any era. In Europe's, it is potentially irreversible. The continent is aging faster than any domestic pipeline can compensate for, and the diseases accumulating in that population are precisely the ones where Europe has fallen behind. The workforce that might have cared for them is already spoken for: across the continent, family members have stepped in where trained carers were never prepared, quietly withdrawing labour from an economy that could not afford to lose it. More patients, sicker patients, fewer hands, slower growth. Europeans are trapped.
By 2036, the pharmaceutical industry as Europe knew it no longer exists in its previous form. The patent cliff that stripped $236 billion from global revenues between 2025 and 2030 removed the financial buffer that might have funded transformation. Roche, Sanofi, GSK and Bayer survive in reduced form with depleted pipelines and their most valuable remaining assets increasingly the target of acquisition by the companies that have replaced them at the top of the value chain. Those companies are biological operating systems: a small number of AI-native platform businesses, predominantly American and Chinese, that own the foundation models, the biofoundry networks, and the training data that determine what drugs get designed, for whom, and on whose terms. The science happens elsewhere. Europe fills out the forms.
By 2029, surface indicators still looked reassuring. Europe had moved. The EU's AI factory network was operational across sixteen member states, and the InvestAI gigafactory programme’s twenty billion euros for AI Gigafactories had begun to deploy, which was enough to sustain the narrative of progress, not enough to shift the underlying architecture of dependency. The EHDS had entered into force, and health data was, in principle, beginning to flow across borders for research purposes for the first time. In the synthetic biology space, there was activity: research programes, startup formation, incremental capital allocation.
But the foundational investments that would have made Europe a builder rather than a buyer were not made. The AI gigafactories, built primarily around research consortia rather than commercial actors, were not generating the private sector innovation velocity their architects had promised. The chips powering them were Nvidia's, procured under the EU-US trade framework, adding a sovereign dependency at the base of the infrastructure Europe was building to reduce its dependencies. The EHDS was opening data access for patient summaries and prescriptions, but genomic and omics data, the categories that matter for training biology-specific models, remained locked until 2031 at the earliest. And public anxiety about data sharing, amplified by disinformation and eroding institutional trust, made it politically costly to accelerate.
In 2029, Europe's biological research still ran on infrastructure others had built. The models had evolved well beyond AlphaFold and Evo: agentic systems now ran autonomous drug design campaigns, executing entire design-build-test-learn (DBTL) cycles without human intervention, continuously learning from each iteration. The models worked but they were trained on someone else's priorities, someone else's populations, and someone else's values about what transparency means. The biofoundry infrastructure that would have allowed Europe to run its own cycles at competitive speed did not exist. This dependency had been visible long before it became structural. Public budgets that might have seeded the foundational infrastructure were consumed elsewhere: defence commitments absorbed the fiscal space that technology sovereignty required. On the private side, the calculus was equally rational and paralysing: no single company could build the ecosystem alone, and without public commitment to anchor it, none tried. The only remaining option was to buy or lease access to the foundational ecosystem, adding a new layer of strategic dependency in a geopolitical environment that was becoming less forgiving of exactly that.
The shift was not a failure of ambition. It was a failure of comprehension. Drug discovery used to compete on the quality of its science. Synthetic biology flipped that: biology became a manufacturing platform, and the competitive advantage moved to whoever could run the fastest design-build-test-learn cycle. That speed is determined by the models you run, the data those models were trained on, and the biofoundry infrastructure executing the cycle. Europe built world-class capability in the old paradigm: precision chemistry, rigorous manufacturing, deep scientific expertise. Those strengths did not disappear. They were made contingent on a foundation layer Europe did not control and had not built. What matters now is not who holds the patent but who can run the next design-make-test-learn cycle fastest. The ground shifted underneath capabilities Europe legitimately owned, quickly enough and invisibly enough inside existing workflows that most decision makers did not register it until the dependency was already structural.
Europe does not have an awareness problem. In every conversation I have with senior leaders across Europe, the diagnosis is the same: a unified market, higher investment in innovation, greater risk appetite. The agreement on what needs to happen is not the bottleneck. Action is. The question is why.
Culture. Europe built societies the world genuinely admires on values that are not obviously wrong: craftsmanship over speed, deep expertise, performance balanced with protection. These are strengths. But they carry a specific cost in a moment that rewards iteration over perfection.
Cognition. Values, held long enough, become architecture. The same commitment to craftsmanship, expertise, and protection that built Europe's admired societies also built its institutions: regulatory frameworks designed to ensure nothing fails, funding structures that reward proven expertise over untested ideas, procurement cycles optimised for certainty over speed. In a paradigm where competitive advantage belongs to whoever runs the fastest design-build-test-learn cycle, that logic does not just slow things down. It selects against the behaviour the moment requires.
Capital. Investment follows from both. Europe has deep, patient, long-term capital. It has built a functioning early-stage ecosystem. What it lacks is the scale-up capital, the Series B and beyond, that turns promising companies into the foundational infrastructure players the moment requires. The gap is not accidental. It is structural. The biofoundries were not built. The foundation models were not trained. Not because the money did not exist. Because the decision architecture was not built to greenlight bets that will mostly fail before one succeeds.
What happened in pharma between 2029 and 2036 is not a pharma story. It is the first visible consequence of a pattern already operating in AI, earlier in its arc, where the window has not yet fully closed.
The position is structurally similar. The foundational layers of the stack, compute, infrastructure, frontier models, are almost entirely outside European control. Intel cancelled its flagship €30 billion semiconductor factory in Magdeburg in July 2025. The EU Chips Act target of 20% global market share by 2030 is now projected to reach 11.7%. Even where Europe holds genuine strength, ASML's monopoly on the chip-making technology the world depends on, that leverage is constrained by US export policy decisions Europe did not make. The dependencies are not impossible to reverse. But the same cultural and cognitive logic that allowed the pharma window to close is operating here, in real time, on a faster timeline.
We can break the cycle. But it requires collaboration at two levels simultaneously: across nations, and across the public and private sector. And it requires capital that European institutions currently cannot deploy.
On the data layer: a European health data architecture where each country retains sovereignty over its population data but contributes to a shared training commons under agreed governance terms. The EHDS provides the legal framework. What is missing is the shared infrastructure underneath it and the political will to accelerate access to the data categories that actually matter for biological AI models, genomic and omics data, currently locked until 2031.
On the biofoundry layer: a coordinated cross-European network of shared physical infrastructure, publicly seeded and privately operated, that gives European researchers and companies access to competitive DBTL cycle speeds without requiring each nation to build the full stack alone. The model exists. Airbus was founded by the governments of France, Germany, the UK, and Spain with the explicit goal of avoiding dependence on American manufacturers. It was seeded with public launch aid, handed to industry to run, and became commercially self-sustaining within two decades. The governance structure was deliberately designed so that partners had to be unanimous to stop it, not to proceed. That asymmetry is what made collaboration possible. What overcame the hesitancy was not trust. It was the shared recognition that the alternative was worse.
The capital question is equally structural. European pension funds hold €2.7 trillion in assets and contribute just 5% of continental venture capital, compared to over 50% in the US. The reasons are partly regulatory. Solvency II requires European insurers to hold capital reserves of up to 49% against unlisted equity investments. Government bonds carry no equivalent charge. Pension funds face analogous constraints under IORP II which nudge them toward safe, liquid assets. The result is the same in both cases: the rules make government bonds structurally preferable to foundational technology bets. Reforming those regulatory levers would unlock European capital that already exists.
The second lever is retail. Europeans save more than Americans. But more than a third of EU household wealth sits in deposits, compared to about a tenth in the US. That capital is not unavailable, but it is structurally misdirected. Mobilising it is a policy choice. The same logic applies to the biofoundry and data layers. And it requires one further adjustment: within these structures, we need to build in deliberate room for experimentation and the tolerance for failure that comes with it. Not the Silicon Valley version, which treats failure as performance. The European version: controlled, bounded, evaluated, and learned from.
Our culture will not change nor should it. But our cognitive patterns and our capital rules can change. The institutional architecture that prevents failure rather than learning from it can be adjusted. The regulations that lock European pension funds out of European venture capital can be reformed. These are policy choices, not “cultural destiny.” Culture is not the obstacle. Cognition and capital are. And both can be changed.
Share on X
Share on LinkedIn