Editorial Article
The CDMO Industry Solved Its Demand Problem and Inherited a Geography Problem
On 3 November 2026, Lonza signed a definitive agreement to buy five contract manufacturing sites in the United States from Merck KGaA, Darmstadt, Germany — roughly 1,100 employees, and enough capacity to more than double its American footprint to nine CDMO sites and a US workforce above 6,000. It is paying an enterprise valuation in the low- to mid-triple-digit millions of Swiss francs. Eleven months earlier, Samsung Biologics had bought its way into the same country for USD 280 million: Human Genome Sciences in Rockville, Maryland, two cGMP plants, 60,000 litres of drug substance capacity and more than 500 staff, bought from GSK and closed at the end of the first quarter of 2026.
Neither company was buying steel. Both could build — Samsung had just filled Plant 4 and had Plant 5 coming. What they bought was a validated site, a regulatory history and a trained workforce, and they bought it at a fraction of greenfield cost. FUJIFILM’s Holly Springs site in North Carolina, opened in September 2025, cost USD 3.2 billion for eight 20,000-litre bioreactors in its first phase, part of more than USD 8 billion the group says it has committed to its CDMO business. Set that against a five-site network for a few hundred million francs and the arithmetic of this industry is suddenly very clear: in 2026, approved capacity trades at a discount to unbuilt capacity, and the discount is the whole strategy.
The year the demand question disappeared
Start with the top line, because it is the part nobody argued about. Lonza closed FY2025 with CHF 6.5 billion of sales, up 21.7 per cent in constant exchange rates, and CORE EBITDA of CHF 2.1 billion — a 31.6 per cent margin, up 1.4 points. Strip out the acquired Vacaville site and organic growth was still in the low teens. For 2026 the company guides to 11–12 per cent CER growth and a margin above 32 per cent.
Samsung Biologics did better than that. FY2025 revenue was KRW 4,557 billion, up 30.3 per cent, with operating profit of KRW 2,069 billion, up 56.6 per cent — and that was with Plants 1 through 4 all at full utilisation by the third quarter. Cumulative contract value passed USD 21 billion; cumulative regulatory approvals reached 420. Management guided to 15–20 per cent revenue growth for 2026, and that is before the Maryland site contributes.
And the Chinese leader, whatever Washington thinks of it, grew too. WuXi AppTec’s FY2025 revenue reached RMB 45.46 billion, up 15.8 per cent overall and 21.4 per cent from continuing operations, with a year-end backlog of RMB 58 billion, up 28.8 per cent, and 2026 revenue guided to RMB 51.3–53.0 billion. Adjusted non-IFRS net margin came in at 32.9 per cent.
Three business models, three jurisdictions, three sets of numbers all pointing the same way: demand for outsourced pharmaceutical manufacturing is not the constraint. Growth of 20 to 30 per cent in a business whose output is measured in litres and kilos is not a volume story first. It is a price, mix and utilisation story — which is exactly why the fight has moved to where the litres sit.
A note on those three numbers: Lonza reports FY2025 growth at constant exchange rates, while Samsung Biologics and WuXi AppTec are as reported. The three are not restated to a common currency basis, so the comparison above is directional rather than like-for-like.
Section 232 rewrote the map
The tariff architecture that landed during 2026 is the single largest change to the industry’s geography in two decades, and it is worth stating precisely, because most summaries of it are wrong.
Following a Commerce Department investigation into pharmaceutical supply chains, an April 2026 proclamation under Section 232 of the Trade Expansion Act set a default 100 per cent duty on patented drugs and their ingredients — active pharmaceutical ingredients and key starting materials included. It took effect on 31 July for 17 large companies named in the proclamation and applied to the rest of the industry from 29 September 2026, replacing the “Liberation Day” tariffs the Supreme Court struck down in February.
The headline rate is not the operative rate. Products from the EU, Japan, South Korea and Switzerland pay 15 per cent; UK imports pay nothing. A Commerce-approved onshoring plan cuts the rate to 20 per cent. Pair that plan with a most-favoured-nation pricing agreement and the rate falls to zero. Generics and biosimilars are excluded altogether, and on 23 September the department named orphan drugs, nuclear medicines, cell and gene therapies and antibody-drug conjugates as qualifying for a zero rate.
That schedule turns a CDMO’s address into a client’s tax position. The Pharma & Biopharma Outsourcing Association lobbied Commerce successfully to have technology transfer to a US CDMO count toward a customer’s onshoring plan — which makes hosting a client’s onshore manufacturing the cheapest available route to a 20 or 0 per cent duty rate. A contract manufacturer with capacity in Indiana is now, functionally, a tariff instrument. One with capacity only in Visp or Incheon is a 15 per cent line item.
You can see the market pricing that in the M&A. Lonza’s five-site purchase adds US bioconjugation, payload-linker and viral vector capability — the modalities where the tariffs are most likely to bite and where the zero-rate list is most generous. Samsung’s USD 280 million bought it a US flag to fly over a Korean operating model. Novo Nordisk went further and bought the asset outright: Novo Holdings closed its USD 16.5 billion take-private of Catalent on 18 December 2024, and Novo Nordisk then took three fill-finish sites — Anagni in Italy, Bloomington in Indiana, Brussels in Belgium — for USD 11 billion to fill its own injection pens. European regulators cleared that deal after finding sufficient competing alternatives. It is harder to make that finding today.
What is actually scarce
The instinct is to read all this as a capex boom. It is not, and the gap between announcements and capacity is where the next five years get decided.
Since the start of the current US administration, large pharmaceutical companies have pledged more than USD 370 billion of American investment; GSK’s own commitment, announced alongside the Rockville divestment, was USD 30 billion over five years. Contract manufacturers have been far more cautious. Trade reporting during 2026 put disclosed CDMO US capex at roughly USD 18.5 billion for 2025, falling to about USD 3.9 billion worldwide in the first half of 2026 — a fraction of it American.
The reasons are practical, and none of them are about demand. Equipment is imported, so tariffs inflate the build: PBOA president Gil Roth has described a planned USD 100 million US project becoming USD 150 million depending on which day the kit arrived on site. Sartorius set aside EUR 26 million to compensate customers for tariff surcharges after receiving refunds on duties the Supreme Court struck down. Data centres and big pharma outbid contract manufacturers for construction crews, equipment slots and — the binding constraint — trained operators. Roth’s blunter warning is that if a Lilly genuinely needs people it can pay far more than a CDMO can, and the White House’s USD 100,000 H-1B fee, extended in September, makes importing a Swiss or Korean start-up team an expensive way to open an American plant.
Which produces the industry’s central structural problem. The 17 companies with White House onshoring and pricing agreements can afford to build. The next 1,700 cannot — and they are precisely the customers for whom contract manufacturing exists. The tariff schedule has made domestic capacity strategically essential at the exact moment the domestic capacity build has become hardest to finance.
The modalities are no longer one market
The averages hide a split that is now wide enough to be a different industry on each side.
Mammalian biologics, drug product and bioconjugates are where the money went in 2025. Lonza’s growth was led by Mammalian, Bioconjugates, Small Molecules, Drug Product and Bioscience Technology Platforms, with that growth partly offset by a softer Cell & Gene business and phasing in Microbial. FUJIFILM opened Holly Springs and immediately landed commercial manufacturing with Regeneron. Antibody-drug conjugates have pulled conjugation capability into supply chains that previously stopped at antibody or payload, and the September zero-rate designation for ADCs removes the last reason to site that work outside the United States.
Cell and gene therapy went the other way. Catalent — now under Novo Holdings — moved in January 2026 to close its European cell therapy centre of excellence at Gosselies in Belgium, a campus with both autologous and allogeneic platforms that had completed a 60,000-square-foot dedicated commercial plant only in 2022, with roughly 150 jobs affected and a reported decline in production behind the decision. Charles River Laboratories announced in early 2026 that it would shed its manufacturing unit, and its chief executive said publicly in May that it likely would not be the last manufacturer to break away from cell and gene therapy. The one-stop-shop logic of 2020–21 — buy breadth, offer everything — is being unwound by the companies that bought it.
The newest fracture is radiopharma. In November 2026 PharmaLogic agreed to acquire Lantheus’s CDMO business and its Springfield, New Jersey site, including a long-term supply agreement to make lutetium Lu 177 dotatate, creating what it describes as the first transatlantic contract platform dedicated to radioligand therapies. Short-half-life isotopes do not travel; that business will be regional by physics, not by tariff.
The other half of the title
It is worth separating the C in CDMO from the CMO, because the tariff schedule treats them differently and the economics were never the same.
Development-led contract work is paid for speed, complexity and risk-sharing — the Lonza and Samsung model, priced on capability, defended by validated suites and regulatory filings. Contract manufacturing of mature small molecules is paid per kilo in a deflationary market, and the generics and biosimilar carve-out leaves that half of the industry exposed to import competition rather than shielded from it, while its customers’ pricing collapses. Both halves sit inside the same value chain that Pheonix models at USD 248.6 billion of active pharmaceutical ingredients in 2025, rising to USD 396.8 billion by 2033 at around 6.0 per cent a year. Roughly a fifth of that chain’s growth is volume. The rest is mix, complexity and price — and the ability to capture it is not distributed evenly across the sites now changing hands.
What to watch
Three signals will tell you whether 2026 was a reallocation or just a reshuffle.
First, whether the acquired sites get approved for new products. Rockville and the five Merck KGaA sites are worth what they are worth because of what is already filed on them; a site that cannot be transferred into a client’s regulatory dossier within a reasonable window is expensive real estate. Watch inspection outcomes and supplement approvals, not ribbon-cuttings.
Second, whether the tariff discount survives its own success. A rate that falls from 100 per cent to zero when a customer signs an MFN pricing agreement is a negotiating lever, not a permanent cost structure — and orphan drugs, cell and gene therapies and ADCs are already outside it. Capacity decisions justified by a rate that could be renegotiated are capacity decisions with a policy bet embedded in them.
Third, whether the labour constraint breaks. Every announcement in this piece — nine Lonza US sites, 1,400 jobs at Holly Springs by 2031, 500 retained at Rockville — assumes qualified people can be found or moved. The equipment can be bought with a tariff surcharge attached. The people cannot, which is why the companies paying triple-digit millions for sites rather than billions for greenfield are, on the current evidence, making the better trade.
The paradox of the year is easy to state. Demand for outsourced manufacturing has never been stronger, the industry’s leaders grew 20 to 30 per cent, and the thing that decided who won was not chemistry or cost per litre but the coordinates of a plant and the date on a customs schedule. Tariffs can move capacity. They cannot make a site approvable, and they cannot train an operator. That was the scarce thing in 2026, and on any forecast that runs to 2033, it still will be.
