The New CDMO Moat: Why Committed Capacity Is Starting to Matter More Than Announced Capacity

Over the past eight weeks, four large CDMO expansions across Switzerland, the United States and India were announced with a customer, a product family or a prepayment already attached. Over the same period, two capable providers reported growing pipelines and worsening near-term economics. The gap between those two groups is the most useful thing in the market right now.

The metric that stopped being useful

Installed capacity is the number the industry still leads with. Liters. Reactor volume. Suites. Lines. Square meters of classified space.

It is a real number and it is increasingly a weak one, because it answers a question almost nobody is exposed to. Very few sponsors are worried that the world lacks manufacturing floor space. They are worried about whether a specific process can be transferred, validated, inspected and released on a specific date, at a cost that does not move.

The number that speaks to that is different. How much of this capacity is already spoken for, by whom, at what stage, and under what terms if the program changes.

Over roughly ten weeks of announcements between June and mid-August 2026, that distinction started showing up in how capital is being committed. Not everywhere, and not as a rule. But often enough, across enough modalities and regions, to be worth naming.

Four expansions, three regions, one capital logic

On 21 July, Bachem announced an agreement to supply large volumes of peptides and said it would invest more than CHF500 million in a new facility at Sisslerfeld. The company explicitly linked the greenfield project to strategic collaboration prepayments and long-term commitments. Commercial production is expected in 2030. [S1]

On 30 July, Resilience and Lilly announced $750 million of investment to expand injectable device production in Ohio. The project adds KwikPen production for diabetes and obesity medicines, is expected to create at least 400 skilled jobs, and is scheduled to reach full operations in early 2027. The two companies had already produced more than 150 million Lilly doses together. [S2]

On 22 July, Lonza disclosed further expansion of commercial-scale aseptic ADC drug product capacity at Stein, Switzerland, backed by a long-term customer commitment. The same H1 update showed Advanced Synthesis sales up 27.7% at constant exchange rates on a 48.1% CORE EBITDA margin. [S3]

On 1 July, Morepen reported the first commercial dispatch under an INR8.25 billion CDMO mandate, after the program moved through validation and qualification into execution. Reactor capacity is expected to reach approximately 600 KL, with phased expansion toward 800 KL and 1,000 KL. [S4]

Four companies. Switzerland, the United States and India. Peptide API, injectable devices, ADC drug product and small-molecule manufacturing.

The shared feature is not geography and it is not modality. In each case the investment decision came after something reduced the utilization risk: a prepayment, a joint investment, a long-term customer commitment, or a mandate already in commercial supply.

Morepen is the clearest sequencing example. The commercial proof and the multi-year mandate came first. The larger reactor build follows.

That ordering is the whole point.

The counter-evidence is just as strong

A thesis built only on the announcements that worked would be worthless. The two most instructive disclosures in the same window came from companies whose demand is growing.

OXB's 7 August update reported H1 revenue up approximately 9% to about GBP80 million. The company added 17 new clients, more than 30% above the number signed in all of 2025, and its non-risk-adjusted new business pipeline rose about 30% year over year to approximately $713 million. [S5]

It then reduced FY2026 revenue guidance to GBP180 million to GBP200 million. The reasons given were deferrals following changes in client strategy or clinical data, a procurement and approval-pathway change at a larger client, clients staging work packages, and a six-month delay to operational readiness at its Durham, North Carolina site. Net cash fell from GBP55 million at the end of 2025 to GBP21 million at 30 June 2026. [S5]

Read that sequence again. Record client acquisition. A materially larger pipeline. Lower guidance.

Fujifilm shows the same problem at a different scale. Bio CDMO revenue grew by more than 6% in the quarter ended 30 June 2026, driven by the ramp-up of new large-scale facilities. The company also reported delays in that ramp-up, lower revenue from small and medium-scale facilities following unplanned shutdowns associated with regulatory inspections, and increased Bio CDMO costs. The Healthcare segment recorded a JPY12.7 billion operating loss. [S6]

Demand can be real and still arrive too late, too unevenly, or with too little contractual protection to absorb a manufacturing network's fixed costs.

This is the same split we found across the CRO sector earlier this month: aggregate sector growth telling you almost nothing about which specific sites are absorbing their cost base.

Announced, contracted, and the space between

The useful reframe is to stop treating installed capacity as one number.

At minimum it splits four ways:

• Operating and contracted. Running, with committed volume against it.
• Operating but uncommitted. Running, sold in work packages, exposed to sponsor cash discipline and clinical timing.
• Under construction and customer-backed. Not yet producing, but with a named anchor, a prepayment, a reservation, a minimum volume or a joint investment against it.
• Under construction without disclosed commitment. Capital already spent against a demand forecast.

Those four categories carry completely different risk, and completely different hiring profiles, which is the part that matters for anyone building the team that has to run the asset.

Backlog needs the same treatment. A proposal pipeline is not backlog. Backlog is not protected utilization. The variables that decide which is which are development stage, cancellation rights, minimum volume provisions, price resets, technology transfer status, and the elapsed time between a signed order and recognized revenue.

None of this is new as a mechanism. Large CDMOs have built against contracts for decades. What appears to be changing is how much of the difference it makes, in a period where capital is expensive, sponsors are protecting cash, programs are reprioritized faster, and complex facilities carry high fixed costs.

There is also a selection bias to hold onto. Companies disclose a commitment when the commitment supports the investment narrative. The absence of a disclosed anchor is not proof that no anchor exists. That is one reason the negative evidence from OXB and Fujifilm matters as much as the positive cases.

For anyone doing this work properly, the questions below are the ones worth putting to a provider, a target or a board. This follows the same method as the CDMO capacity work we published earlier this year.

Eight questions, scored. Answer them about a provider, a target, or an announcement you are reading. The weightings are ours, drawn from how these deals tend to behave rather than from any published benchmark, so treat the band as a prompt for the next question rather than a verdict.

Modality explains where the money goes

Modality tells you where the capital is going. It does not explain why the capital moves the way it does. Both parts are worth separating.

Peptides have moved from a growth theme to a strategic platform. Samsung Biologics offered CHF44.31 per share for PolyPeptide, implying an equity value of approximately CHF1.46 billion and a 40% premium to the undisturbed share price, describing the rationale as specialized peptide expertise plus a manufacturing network across the United States, Europe and India. [S7] Two weeks later CordenPharma closed its acquisition of AmbioPharm, adding peptide manufacturing in South Carolina and Shanghai and taking the combined workforce to approximately 3,500. [S8] Set alongside Bachem's project and the Resilience-Lilly expansion, the pull runs the length of the chain: API synthesis, purification, fill-finish, device assembly and packaging. [S1] [S2]

ADC and bioconjugate capability is attracting commercial-scale capital. Lonza's Advanced Synthesis growth and margin is the strongest public operating signal, announced alongside customer-backed aseptic ADC drug product expansion. [S3] In June, Shilpa Biologicals commissioned an integrated ADC drug substance facility in Dharwad, India, covering payload, linker, monoclonal antibody, conjugation and GMP-scale purification. [S18] The binding constraint in ADC outsourcing has never been a single production step. It is coordinated control of containment, linker chemistry, antibody supply, conjugation, purification, sterile drug product, analytics and transfer, which is exactly the coordination problem our standing work on ADC manufacturing constraints keeps returning to.

Cell and gene therapy is recovering scientifically faster than it is recovering economically. The Q2 2026 ASGCT and Citeline review counted 4,487 gene, cell and RNA therapies in development, eight approvals in the quarter, and eleven new mRNA-engineered CAR-T entrants, matching the total for all of 2025. [S13] Lonza reported Specialized Modalities sales up 22.6% at constant exchange rates with a sixth commercial therapy approved within its Cell and Gene Technology Platform. [S3] Against that: OXB's order staging, 93 Catalent positions at Harmans and three in Baltimore disclosed in Maryland WARN records in February, and Charles River's decision to divest CDMO and Cell Solutions businesses that generated $143 million of revenue in 2025. [S5] [S15] [S16] Two things are true at once. The pipeline is expanding, and manufacturing demand within it remains concentrated, phase-dependent and easy to defer.

Radiopharmaceuticals sit at the sharp end of the same logic, and here we are reading structure rather than citing a transaction from this window. A decaying product cannot be inventoried. Capacity is only useful if it is scheduled, licensed, staffed and within reach of the patient on the day the batch exists. That makes an uncommitted radioligand suite a weaker asset than an uncommitted vial line, and a committed one considerably stronger. It is also why radiopharmaceutical manufacturing and supply is the area where we most often see hiring pulled forward ahead of the concrete, rather than after it.

M&A is buying capability, customers and options

The transaction pattern supports the commitment thesis without reducing to it.

Samsung's proposed PolyPeptide acquisition is a platform extension: a scaled biologics manufacturer buying specialized peptide technology, trained teams, customer relationships and a six-site network, in less time than building the capability organically would take. [S7]

CordenPharma's AmbioPharm deal adds a second US peptide site, an Asian manufacturing presence and complementary upstream and downstream capability. Technical capability and three-continent supply design in the same transaction. [S8]

Earlier in 2026, Samsung completed its acquisition of GSK's Rockville site, adding 60,000 liters of clinical and commercial drug substance capacity, more than 500 experienced employees, and a continuing supply relationship under which Samsung keeps manufacturing for GSK. [S12] Siegfried's acquisition of Noramco, Purisys and Extractas added approximately 400 employees and three small-molecule drug substance sites across the United States and Australia. [S21]

The recurring asset package is validated capability, an experienced quality and operations workforce, existing customer or product relationships, development-to-commercial continuity, and more than one viable geographic supply path.

Empty floor space is still traded. It is simply a weaker thesis than a capability-and-commitment bundle, and it is priced accordingly.

Worth saying plainly: an acquisition of an operating site is substantially a talent acquisition. An established GMP workforce with inspection history is harder and slower to recreate than the building around it.

Geography is becoming optionality

US capacity is attracting investment. Evonik announced a $100 million modernization of Tippecanoe Labs in Indiana, citing demand for US-based drug substance services, geopolitical uncertainty and growing API complexity, across 170 m3 of HPAPI and 860 m3 of general API capacity. [S9] Resilience is adding at least 400 skilled Ohio jobs. [S2] Bristol Myers Squibb announced a $2.3 billion Houston site for small molecules, biologics and ADCs with approximately 500 permanent skilled jobs. [S19]

The evidence does not support a clean West-replaces-China reading.

CordenPharma acquired South Carolina and Shanghai peptide sites in the same transaction and described the result as regionally aligned supply across three continents. [S8] Samsung's PolyPeptide offer adds sites in Europe, the United States and India. [S7] WuXi AppTec reported H1 2026 revenue of RMB28.90 billion, up 38.9%, with continuing-operations backlog up 25.2% to RMB66.43 billion, and raised 2026 capital expenditure guidance to RMB7.5 billion to RMB8.5 billion, while stating a dispute over its inclusion on the US Department of Defense 1260H list. [S11]

The US tariff framework announced in April creates a real incentive for domestic manufacturing. [S20] Company behavior still shows cross-border networks being retained and extended.

What that produces is regional optionality rather than relocation: US capacity for proximity, policy exposure and resilience on critical products; European capacity for established regulatory, technical and talent ecosystems; Asian capacity for scale, specialized chemistry, cost and a growing innovation base; and deliberate redundancy where the product economics justify paying for it.

Funding quality sets order quality

Biotech financing improved in the first half of 2026, but the distribution matters more than the total.

BioPharma Dive counted at least 68 venture-backed biotechnology companies raising more than $9.1 billion in H1, the highest first-half total in its tracked group since 2022. Approximately 76% of that capital came through rounds of $100 million or more, and about two-thirds of rounds went to companies that already had a drug in human testing. Cell and gene therapy and nucleic acid companies raised materially less than biologics and small-molecule developers. [S14]

That is the mechanism behind a recovery that some specialist CDMOs are not feeling. Money is concentrating around clinically de-risked assets, established teams and larger rounds. It is not spreading evenly across the preclinical market.

For a CDMO, this changes the quality of demand rather than the quantity:

• Later-stage programs can support reservations, dedicated suites, prepayments and long commitments.
• Early-stage programs generate proposals and initial work, with high deferral risk attached.
• Large sponsors can underwrite capacity directly. Smaller sponsors stage work packages to preserve cash.
• Clinical success can turn a constrained specialist capability into a critical path overnight. Failed data can remove years of forecast utilization just as fast.

Counting programs without weighting them for stage, runway, sponsor quality, procurement behavior and manufacturing intensity is how a pipeline number ends up disconnected from a revenue number.

What committed capacity does to hiring

This is the part we spend our working week inside, so it gets the most specific treatment.

The workforce evidence in the window is fragmented but directionally consistent. Resilience expects its Lilly-linked expansion to create at least 400 high-skilled jobs, taking Resilience-created employment in Ohio above 1,400, alongside the nearly 1,000 people already employed at its two Cincinnati-region facilities. [S2] Samsung's Rockville acquisition brought more than 500 existing employees in with the capacity and the continuing GSK supply. [S12] CordenPharma's integration creates a combined workforce of approximately 3,500. [S8] Siegfried's acquired assets include about 400 employees. [S21] On the other side, the Maryland WARN records and the Charles River divestitures show that specialized capability does not protect a site from weak utilization or portfolio reprioritization. [S15] [S16]

The pattern underneath is that durable hiring follows funded commercial transfers and operationally proven sites more reliably than it follows capacity announcements.

Three consequences worth planning around.

The scarce roles sit at the interfaces, not in the middle of the plant. Committed demand becomes released product through technical operations, MSAT, process and analytical transfer, validation and CQV, QA and sterility assurance, QC, automation and MES, regulatory CMC, and supply chain leadership. Those are the roles where a delay does not just cost a month, it moves a PPQ campaign and therefore a launch. They are also, consistently, the hardest briefs in CDMO executive search, because the pool is small and the people in it are already inside a critical path somewhere else.

Hiring timing is a fixed-cost decision, not an HR decision. Hire early against an uncommitted build and you carry payroll against an empty schedule. Hire late against a committed one and you put validation, PPQ and inspection readiness at risk, which is the more expensive error by an order of magnitude. The commitment category the asset falls into should set which error you are willing to make.

Modality changes lead time more than headcount. An ADC CDMO recruitment brief and a general biologics brief of the same seniority are not the same search, because containment, conjugation and analytical control narrow the credible pool sharply. The same is true of cell and gene therapy CDMO recruitment, where commercial-stage process and analytical transfer experience is genuinely rare, and of radiopharma CDMO recruitment, where the qualifying pool is shaped by radiation safety, site licensing and release under time pressure. Peptide CDMO recruitment has moved fastest of all, because four separate large-scale investments are now drawing on the same purification and fill-finish talent base at the same time.

The planning unit that survives contact with reality is the committed program and its critical path, not the construction schedule. The tool below back-plans from a date you actually control against.

Pick the date the first commercial batch has to exist. The model works backwards from there, adjusting for modality and for whether you are building, expanding or inheriting a site. The final column is the one most plans get wrong: retained search plus notice periods runs to roughly five months before anyone is at a desk.

If you are sequencing a team against a transfer date and want a second opinion on the shape of it, that is a conversation we are happy to have. It is also most of what our organization and headcount benchmarking work exists to answer.

What this article does not claim

The thesis is directional. It deserves the qualifications.

This is an eight-week window. Company announcements are selectively disclosed and contract terms are usually incomplete. Treat this as an emerging capital-allocation pattern, not a market-wide rule.

Customer-backed capacity is not new. Large CDMOs have built against contracts for a long time. The claim is that the discriminator is becoming more important in this capital environment, not that the mechanism was invented this summer.

The evidence is strongest in peptides and complex injectables. It is thinner in biologics drug substance and thinner still in radiopharmaceuticals, where the reasoning above is structural rather than transactional.

Not all successful capacity has to be customer-funded. Speculative investment still works, particularly where a provider has the balance sheet to wait. The hierarchy of credible investment is what appears to be shifting.

WuXi's performance contradicts decoupling, not regionalization. Strong economics and integrated execution can outweigh political pressure while Western capacity expands in parallel. [S11]

Sponsor insourcing is a live alternative explanation. The Bristol Myers Squibb Houston project shows some critical capacity will be internalized. [S19] The Resilience and Lilly structure shows sponsors can also get control through a deeply committed partnership. [S2] The likely outcome is a spectrum of control rather than a binary.

The thesis would weaken materially if, over the next twelve to eighteen months, speculative greenfield sites consistently reached strong utilization without anchor commitments; early-stage funding broadened into firm, non-staged manufacturing orders; customer-backed expansions underperformed non-contracted builds after adjusting for modality and timing; or M&A shifted back toward generic volume with little emphasis on capability, workforce or customer continuity.

We will be watching those four.

The diligence question

By 2028 the more useful question about a CDMO may not be how much it can make.

It may be how much of the next three years is already contracted, at what development stage, with what cancellation economics, and how much of the workforce required to deliver that is in seat today.

That last clause is the one most often left off. Committed capacity that nobody has staffed is a different asset from committed capacity with a validated team and inspection history behind it. The contract sets the revenue. The team sets whether it arrives on the date in the contract.

Frequently asked questions

Sources

Primary company releases and government records, accessed August 2026. Every bracketed reference in the body maps to this list.

1. S1.Bachem, 21 July 2026. Bachem plans to build a large-scale production facility in Sisslerfeld.
2. S2.Resilience, 30 July 2026. Resilience and Lilly Invest $750 Million to Increase U.S.-Manufactured Medicine Supply.
3. S3.Lonza, 22 July 2026. Lonza Delivered Strong H1 2026 Performance Across All Business Platforms.
4. S4.Morepen Laboratories, 1 July 2026. Commercial supplies commence under INR8.25 billion CDMO mandate.
5. S5.OXB, 7 August 2026. Half Year Trading Update and Notice of Results.
6. S6.Fujifilm Holdings, 6 August 2026. Financial Results for the First Quarter Ended June 30, 2026.
7. S7.Samsung Biologics, 20 July 2026. All-cash offer to acquire PolyPeptide.
8. S8.CordenPharma, 3 August 2026. CordenPharma closes acquisition of AmbioPharm.
9. S9.Evonik, 8 July 2026. Evonik invests $100 million to upgrade its US drug-substance CDMO site.
10. S11.WuXi AppTec, 3 August 2026, distributed via Nasdaq. WuXi AppTec Delivers Strong H1 2026 Results and Raises Full-Year Guidance.
11. S12.Samsung Biologics, 31 March 2026. Completes acquisition of GSK's Rockville facility.
12. S13.Citeline with ASGCT, Q2 2026. Q2 2026 Gene, Cell and RNA Therapy Landscape Report.
13. S14.BioPharma Dive, 13 July 2026. Biotech startup funding gap widens despite rebound in VC investment.
14. S15.Maryland Department of Labor, accessed 20 August 2026. WARN notices.
15. S16.Charles River Laboratories, 25 February 2026. Provides Update on Planned Divestitures.
16. S18.Express Pharma, 26 June 2026. Shilpa Biologicals commissions ADC GMP manufacturing facility in Dharwad.
17. S19.Reuters, 10 August 2026. Bristol Myers Squibb to build $2.3 billion manufacturing site in Houston.
18. S20.The White House, 2 April 2026. Adjusting Imports of Pharmaceuticals and Pharmaceutical Ingredients into the United States.
19. S21.Siegfried, 27 January 2026. Siegfried to acquire high-quality small-molecule drug-substance capacity in the US.

Two further sources were consulted as background and are not cited in the body: Samsung Biologics Q2 2026 results, 23 July 2026, and DCAT Value Chain Insights, CDMOs/CMOs: The Key Moves in 2026, August 2026.

Method

Evidence window 1 June to 20 August 2026. Primary company releases and government records were used for transaction terms, results, staffing and capacity figures. Company statements are treated as company-reported, not as independent validation of demand. Earlier 2026 items are marked as context. Generic market-size and CAGR forecasts were excluded.

This article is independent market intelligence and not investment, legal or regulatory advice. Company names and trademarks are the property of their respective owners. © 2026 ProGen Search Limited.