A Dependency Investigation (#11) – October, 2026
About This Series
The Magician’s Hands is a series of dependency investigations. Each report examines a single case in which a structural dependency, between a state and an infrastructure owner, a farmer and a seed company, a continent and an energy supplier, was created, normalised, leveraged, and converted into power. The cases span domains and decades. The grammar beneath them does not change.
The series takes its name from a simple observation: the most consequential things happening in the world are rarely the things that take centre stage. While we watch the visible hands, something else is being built in the structural layer underneath. These reports are an attempt to make that layer legible.
The founding article, “”, sets out the full grammar. Each investigation that follows applies it to a specific case.
Scene-setter
A pharmacist in the Netherlands who opens a drawer for a tablet of levothyroxine, lithium or phenobarbital is relying on one of the most reassuring promises of modern health care: that the medicines which keep people stable are cheap, interchangeable and always there. Generic medicines are the quiet workhorses of that promise. The European Commission puts their share at more than 65 percent of the sales volumes of critical medicines, and the visible story is one of success. Patents expire, competition arrives, prices fall, and health systems buy more treatment for less money. In the Netherlands, insurers run preference policies that reward the cheapest supplier, and the savings are presented, with good reason, as a public good.
The visible story about the failures is told in a different register. In 2025 the Dutch pharmacists’ association KNMP counted 1,134 medicine shortages, down from 1,563 the year before, with just over 3.5 million people affected. The explanations offered in public are the ones that fit a story about procedure: insurers whose preference policies narrow supply, manufacturers who exit unprofitable products, and a global market that has become fragile. Each is partly right. None of them describes what sits underneath.
This investigation looks at that underlying layer: where the active ingredients and their precursors are made, who decided that they would be made there, and what the arrangement does to the people at the end of the chain. It is not a story about a villain. Its thesis is that Europe built this dependency itself, one rational sourcing and purchasing decision at a time, and that the same price logic that made generic medicines cheap also made the system that supplies them thin, opaque and hard to reverse, with patients absorbing the consequences of choices made far upstream of them and years before.
Move 1: Creating the Dependency
The entry point here is creation, with elements of insertion and no demonstrated capture. The pattern is familiar from textiles, from the auto industry and from printing: components and ingredients moved abroad first, then whole manufacturing followed, helped by access to raw materials and lower labour cost. The Commission itself says that companies have moved production or outsourced key ingredients outside the EU, especially to Asia. The structural shift runs over roughly the past two decades*, and there is no single decision to point at, no acquisition and no treaty. That absence is itself the finding.
The hook was real utility. For a European generic manufacturer, buying the active pharmaceutical ingredient (API) from an Asian producer cut the cost of the single largest input, and in a market where the product is chemically identical to every competitor’s, cost is the only lever. For a purchaser, a cheaper product meant lower premiums and budgets that stretched further. Nothing about any individual decision looked reckless, and the certification system gave it cover. A producer of an API can obtain a certificate of suitability from the European Directorate for the Quality of Medicines (EDQM), and the certificate does the work of admitting the ingredient into European products. The data on four molecules shows what that produced. For levothyroxine there are 10 valid certificate holders, six in India and four in Europe, and the European certificates appear, from their certificate numbers, to date from 1998 to 2011 while the Indian ones date from 2016 to 2025. For lithium carbonate there are four valid holders and none is in Europe. For phenobarbital there are six, two in China, two in India, one in Hungary and one in Germany. Sotalol has three valid holders, in Germany, Spain and India, and two withdrawn ones, in China and Switzerland.
The insertion element is visible in the levothyroxine dates. Asian producers did not replace European supply by announcement; they entered an existing European flow through cost advantage and certification, and arrived in numbers after the European holders had long been established. The Dutch numbers show what the combined move did to the market underneath: suppliers per generic medicine fell from 3.3 in 2016 to 2.6 in 2025, and the share of medicines with a single manufacturer rose from 32 to 38 percent.
The dependency creator is not one actor. European generic manufacturers and marketing authorisation holders moved production and sourcing abroad, and public and insurer purchasers rewarded the move through price-based procurement. Together they built the arrangement. The holder of the dependency, as distinct from its creators, is the Asian production system, with China upstream in starting materials, intermediates and APIs, and India in APIs and formulation while depending on China itself for intermediates. The moment at which the arrangement became costly to reverse is not dated by any record. It is the point at which the European capacity that could have produced the ingredient no longer existed, and rebuilding it meant more than reversing a contract.
Move 2: Normalising the Dependency
Narrative normalisation worked through a story about efficiency that nobody had to impose. Generics were the success of health policy: patents ran out, prices dropped, and access widened. The institutions that carried this story were the ones that benefited from it, the insurers and ministries that report savings, and the manufacturers who could point to a competitive market. The counter-reading, that cheap supply was being bought with concentration, was available early. Dutch broadcast journalism was reporting on dependence on China and India in 2019 and 2020, including NOS op 3 in June 2019 and the Pointer programme in May 2020. It did not land as a structural claim. It was received as a series of individual shortage stories, each with its own immediate explanation, and the question of what the shortages had in common was left to the viewer.
Operational normalisation is the stronger mechanism. A pharmacist does not check the country of manufacture of an API, because the product on the shelf carries a marketing authorisation, a brand, and a European company name. The cause of Dutch shortages is recorded as unknown in 70 percent of cases. That number deserves to be read slowly: a system that is relied on by millions of people cannot say, in most cases, why its supply failed. Illustrative cases show how this looks in practice. In early October 2026, levothyroxine tablets, lithium tablets and phenobarbital 50 mg were all shortage-listed from the same manufacturer, Pharmachemie, which is Teva’s Dutch company. Lithium and phenobarbital are not under preference policy, and the lithium shortage is attributed to raw material problems. Whatever the preference policy explains, it does not explain that.
The dependency stopped being a relationship and became infrastructure when the supply chain stopped being something anyone in the purchasing system looked at. Switching cost rose invisibly. A producer who left the market because the price no longer covered cost did not return when it recovered, because the plant and the certificate had gone with it.
Move 3: Leveraging the Dependency
Direct leverage by the holders of the dependency is difficult to document in this case, and honesty requires saying so. There are allegations of below-cost pricing by state-supported producers. Sandoz announced on 28 May 2026 that it had submitted a draft anti-dumping complaint to the European Commission about Chinese amoxicillin API, alleging below-cost prices and heavy state subsidy; the allegation is the company’s and is carried here as such. In March 2020 India put exports of 26 APIs and their formulations under licence, which is a fact about what a producing state can do in a crisis rather than evidence of sustained leverage.
The more consequential leverage is structural and sits on the receiving side. Dutch insurers’ preference policies determine which product gets reimbursed, and in 2025 71 percent of shortages (74 percent in 2024) involved products under that policy. The figure cannot be read as causation, because the base rate of products under preference policy is unknown, but it shows where the purchasing system’s decisions meet the shortage statistics. Germany’s discount contracts play the same role, and the German insurers dispute that they cause shortages. The effect is that the buyer’s own design narrows the field. A manufacturer who must be the cheapest to be bought at all cannot afford a second source of API, a safety stock or a European plant.
Option-space leverage is the most durable form, and here it operates on everyone. Manufacturers do not pursue resilient sourcing, because it would price them out. Insurers do not tender for resilience, because their mandate is premium. Ministries do not legislate against price competition, because it is the instrument that kept costs under control. The dependent parties have each narrowed their own options before any decision point arrived, and no one applied pressure to make that happen. The people who inherit the effect are the patients and the pharmacists. A pharmacist in the Netherlands spends time finding a substitute; a patient takes a different tablet, or goes without.
Move 4: Obscuring the Mechanism
The central question is why the pattern was not seen when it could still have been acted on. Temporal displacement answers part of it. The offshoring decisions of the 2000s and 2010s were made by managers and procurement officers who were not the officials, pharmacists or patients who faced the shortages of 2019 and after. Savings were booked in the year of the decision. The cost arrived years later and in a different budget.
Narrative capture kept certain questions from being asked. The dominant story was about price and competition, and it made unaskable the question of what a medicine’s price leaves out. Which factory makes the ingredient? What happens if it stops? Who inspects it? The questions that could be asked were procedural: is the tender fair, is the preference policy well designed, is the shortage registered. The European Court of Auditors’ special report, published on 17 September 2025, records 136 critical shortages reported to the European Medicines Agency by national authorities between 2022 and October 2024. A law firm has called that report alarmist, noting that critical shortages are rare and usually resolved*. That is the contest the narrative allows: argue about how bad the shortages are, not about why the system is built to produce them.
Structural opacity is the strongest mechanism. The border between the receiving and producing regions is held by two instruments: the EDQM certificate and the inspection. US data show what the second looks like in practice. The GAO found in 2022 that FDA foreign inspections had generally been preannounced, which it cited as a challenge to equivalence with domestic inspection, and that 8 of 20 positions in the dedicated foreign drug cadre were vacant in November 2021. These are American figures and dated, but they describe an interface that the European system may share in structure. The information on who makes what is technically available in the EDQM database. What does not exist is the institutional capacity to assemble it into a picture, and the Dutch Court of Audit has concluded, as reported by the KNMP, that the minister has insufficient insight into shortages from the patient perspective.
Move 5: Converting the Value
What was converted here is, first of all, financial. Cheaper generics reduced health expenditure, and the savings were real and went to premium payers, insurers and public budgets. The offshoring margin went to manufacturers that sourced cheaply, and the production value moved to the producing regions. This conversion was legitimate and visible, and it sits at the surface of the case.
Strategic conversion is the less visible one. Production capacity is a form of sovereignty, and it was converted into price. A 2024 analysis by the EU’s health emergency authority, cited by the European Court of Auditors, found that for four of the eleven APIs assessed the EU relied on non-EU sources. A European health system that cannot make its own lithium or its own phenobarbital has handed the producing regions a position it can only rebuild at cost. Normative conversion follows. Quality standards for the active ingredients reaching Europe are in practice set at the border, by certification and sparse inspection, and that is where the sartan case shows the cost. When valsartan API from Zhejiang Huahai was found contaminated with NDMA in 2018, the problem spread across class-wide recalls, and the contamination is believed to have been present for about six years. The European Medicines Agency stated that the problem had not been identified at authorisation and that marketing authorisation holders, who usually outsource API manufacture, might not have had sufficient oversight.
The ones who paid are not the ones who decided. Roughly 3.5 million people in the Netherlands were affected by a shortage in 2025, and the number was 4.5 million the year before. They had no choice in the matter, no visibility into the cause, and no influence on the purchasing logic that produced it. Pharmacists absorbed the labour of substitution, and hospitals absorbed the costs of improvisation. This is the end tier of the dependent party, the innocent bystander, and it is the tier that pays at every level of conversion. The long-term consequence still unfolding is that the system is being asked to rebuild, through the Critical Medicines Act, a capacity that its own purchasing logic spent two decades removing.
Analytical Notes
Two dependency architectures collide in this case, and the collision may be the most consequential dynamic. The first is the receiving-region architecture. Legislation and government act as regulator, price-setter and buyer, and set the framework within which commercial suppliers operate under constant price compression. This takes national forms: the Dutch preference policy, Germany’s discount contracts, and state reimbursement and pricing in the UK. The creators of this architecture are national and European legislators and the public purchasers who operate it. The second is the manufacturing-region architecture. State-shaped supply, through promotion, strategic restriction and episodic heavy enforcement, sits over commercial producers whose local oversight is of uncertain strength, and which carry a nested dependency of their own, with India relying on China for intermediates. India’s bulk-drug incentive scheme (outlay Rs 6,940 crore, running from 2020-21 to 2029-30) reads as a counter-dependency response more than the original driver; that reading is interpretive.
The two meet at the border, where certification and inspection are the only interface, and where each side’s oversight stops. The dependent party caught between them is two-tiered. Institutions are caught between a purchasing logic that rewards price and a supply logic they cannot see. Patients are caught beneath both. Purchasers occupy a dual position, as creators of the dependency and as agents acting on behalf of the patients who bear its cost, which is why the pressure to change falls on the party with the least ability to apply it.
The second analytical thread concerns resistance and failure. The sartan recall of 2018 and 2019 is the clearest documented instance of one upstream source propagating into a continental recall, and the EU review that followed (triggered on 5 July 2018, concluded on 31 January 2019) shows the system correcting itself after the fact. Policy responses are now under way. The Council reached its position on the Critical Medicines Act on 2 December 2025, a provisional agreement with the Parliament followed on 12 May 2026, and formal adoption is pending. Germany passed a shortage law in 2023, and the UK House of Lords has taken up the subject in 2026*. Whether these measures rebuild capacity, or mainly improve coordination, is open. A second axis runs through the case: contamination and quality as an effect of lost control, with the sartans as the documented instance. Ranitidine, whose NDMA problem may arise from degradation of the molecule itself, is the contrast case, a failure that cannot be attributed to the supply chain at all.
Closing
What this case reveals about the grammar is that creation does not need a creator in the singular. Cloud computing was built by three firms. Here the dependency was assembled by thousands of procurement decisions, none of which was designed to produce an outcome that no one wanted. The grammar works as well when nobody is steering as when someone is, and perhaps better, because there is no actor to name and no moment to contest. The textile case showed European capital building the dependency itself. This case adds the public purchaser to that picture, and shows what happens when the institution meant to protect the dependent party is the one that sets the price logic.
It also reveals something about where the grammar reaches its limit. The patient, who is the ultimate dependent, is the one party that can neither choose nor see, and cannot even be told why. Moves 2 and 4 in this case are less a matter of hiding than of nobody holding the whole picture, and that distinction matters for what can be done. A concealed dependency can be exposed. A dependency that is spread across too many hands to be anyone’s responsibility needs a different remedy.
The question the reader should carry forward is this: when the cheapest option has been the rational choice at every step for twenty years, who is responsible for the moment at which it stops being affordable?
Comments are closed