BIOSECURE, supply chains, skilled labour: how geopolitics is forcing bioprocess technology toward digitalisation

Around 76 percent of the 1,250 decision-makers surveyed in the Cytiva Global Biopharma Index 2025 state that the BIOSECURE Act is fundamentally changing their sourcing strategies. By 2028, 56 percent expect a drastic increase in domestic production of biological active ingredients in BIOSECURE-compliant high-wage countries in order to continue serving the dominant US pharmaceutical market.

Western Big Pharma companies that nevertheless want to utilise China's speed advantage to develop drugs for other countries and regions are now calculating a risk premium of 20% to 30% for projects in China to maintain back-up capacities in the West (dual sourcing).

Overlap is particularly observable in markets where China has clearly taken the development lead, such as in the optimisation of CAR-T (> 50% of the global pipeline) and ADC therapies (42%). Around 83.6 percent of cell and gene therapy developers outsourced their production in 2025 – in 2024, a good 30.4 percent still produced in-house. Early development takes place in China, while pivotal studies occur in BIOSECURE-compliant countries. Cell and gene therapies present process technology with fundamentally different challenges than the mass production of antibodies: batch sizes are small, processes are often patient-specific, and the requirements for sterility, traceability, and process control are extremely high. This is precisely why dependence on specialised CDMOs is growing particularly fast here.

According to BioPlan, in-house capacities generally still account for 65 percent of global production capacity. However, 42 percent of developers plan to outsource their production at least in part to CDMOs by the end of 2027. Following this trend, the use of single-use equipment in commercial production in CDMO-typical multi-product plants has risen to more than 44 percent. For biologics produced in mammalian cells, the outsourcing rate was almost 78.5 percent.

• BioPlan Associates reports budget growth for upstream and downstream efficiency tools of 5.6 and 5.5 percent.
• Downstream processing: The purification of market-dominating antibodies and antibody constructs accounts for 50 to 70 percent of process costs. Since available GMP-compliant purification methods are reaching their limits, 34 percent of companies plan to invest in continuous chromatography this year, according to BioPlan Associates. Together with membrane filtration, this is the decisive lever for economically managing the high output of modern mAb processes.
• Although 46.5 percent of those surveyed by BioPlan from the bioprocess industry are testing continuous bioprocessing in single-use systems made of bio-based plastics—primarily CDMOs—for sustainability and efficiency reasons, the adoption rate remains low at 10 to 15 percent. This is because fed-batch processes carried out in contamination-proof single-use equipment are established and approximately 20 percent more cost-effective. Additionally, the extremely high validation effort acts as a deterrent. According to BioPlan surveys, there is nonetheless movement in the market, as 39 percent of decision-makers state they intend to invest in continuous downstream processing in 2026. Mass production continues to take place in stainless steel fermenters due to economies of scale, but increased operating and energy costs, as well as sustainability requirements, are increasingly forcing companies toward smarter solutions and outsourcing.
• Productivity peaks: In this context, Christopher Hwang, an industry veteran in continuous bioprocessing, emphasises that while perfusion is inherently more complex than fed-batch, it is far from new, having been used commercially since the 1990s. “What has changed is the level of intensification. Today, multiple groups report volumetric productivity exceeding 3 g/l per day, with Transcenta/HJB demonstrating up to 8 g/l per day in early 2024,” Hwang says. “At such high productivity level, a single 500 l single-use perfusion bioreactor, when integrated with hybrid continuous downstream processing, can generate more than 700 kg of drug substance annually. This enables a relatively small single-use facility to match, or even exceed, the output of much larger traditional plants. The result is a step-change in economics: significantly lower CAPEX and COGM, improved operational flexibility, and the ability to move toward near ‘just-in-time’ capacity for any protein biologics.”
• Sustainability & regulation: This leap in efficiency also addresses increasing regulatory pressure toward sustainable production. While stainless steel plants consume enormous amounts of highly purified water (WFI), high-efficiency single-use concepts massively reduce resource requirements per kilogram of product, even if handling plastic waste volumes remain a central innovation topic under EU regulatory pressure. These technological leaps in the midst of a necessarily traditionally conservative market environment are motivated in particular by European sustainability requirements, which in this context could prove to be innovation-promoting rather than bureaucratic-inhibiting.
• A similarly positive role could be played by recent EMA regulatory requirements from February 2026 regarding the use of causal physics-based simulation solutions, which are intended to curb purely empirical wet-lab testing as well as animal experiments and material consumption in order to gain resilience.
• While leading equipment suppliers already use physics-informed AI/ML models (PINNs) in mini-bioreactors for process development, Digital Twins and PINNs – simulation instead of empiricism – have not yet arrived in bioprocess control. This is due, in part, to the fact that real-time FT-IR sensors are simply too expensive, forcing the practical application of indirect methods to estimate crucial process parameters. However, according to BioPlan, almost 40 percent (38.8%) of decision-makers state they are evaluating in-silico tools for bioprocess control and process documentation.

In-silico methods devised in the US and Europe, such as physics-informed process modeling (PINN), already help digital replicas of bioreactors – so-called digital twins – in 10 to 15 percent of cases today in the development of control strategies for bioprocess development. For the control of GMP processes, the adoption rate among first-mover CDMOs is much lower at 3 to 5 percent. However, the potential efficiency gain by shortening process development time is significantly greater.

• Sartorius uses hybrid models, for example, for upstream optimization to predict the growth of cell cultures in their Ambr^® mini-bioreactors.
• Cytiva uses PINNs in chromatography (downstream) to simulate the loading and elution of proteins. Since the physical laws of mass transfer in the column are known, these are taught to the neural network. The result is a digital twin that calculates in real-time when a separation column is saturated without the need for manual sampling every time.
• Digital Twins with PINNs as the digital engine enable the use of standard sensors and the error-free calculation of critical control parameters almost in real time. Siemens positions itself via its automation platforms (such as SIMIT and PCS 7) as an integrator for hybrid modeling. Here, PINNs are used as so-called soft sensors. The use case: In a bioreactor, there are parameters that cannot be measured directly (e.g., the exact metabolic rate of the cells inside). PINNs calculate these values from existing data such as pH, dissolved oxygen, and temperature by taking the underlying thermodynamics into account. This saves the purchase of expensive spectroscopy hardware like the $US100,000 FT-IR sensors.

The fact that almost 40 percent of decision-makers in the traditionally conservative bioprocess technology industry want to test this technology shows the actual potential: Because the process is modeled in silico and – unlike standard AI – within the limits of what is physically permissible, expensive wet-lab testing is eliminated, thereby reducing raw material and material consumption as well as personnel requirements through automation.

Among decision-makers, PINN-controlled Digital Twins are therefore already considered the method of choice to scale processes more effectively and to replace experience-based process development with reproducible simulation. Although inconsistencies due to the biological variability of processes and protein micro-heterogeneities remain a major challenge, cost reductions appear realistic for the first time independently of the Chinese brute-force testing infrastructure. At the same time, the need for specialists for bioprocess development is reduced.

Looking at the integration of PINNs – largely developed in Europe – along the entire biopharma value chain, leading analytical firms forecast that they have the potential to reduce the clinical failure rate from 90 to 75 percent and to reduce the time to market even more significantly than China‘s commercial testing structure.

Whether this might result in a new distribution of roles with significantly faster timelines is currently still speculation, but appears quite realistic in light of the copy-resistance of such models.

Bioprocessing pioneer Uwe Gottschalk expects great progress from the removal of regulatory and sector-internal blockades to data-oriented bioprocessing: “The bioprocessing sector still too rarely uses data as a strategic raw material. Only through the full integration of automation, AI, and networked analytics can development times be significantly shortened, processes made more robust, and costs reduced in the course of Bioprocessing 4.0. At the same time, the sector faces central challenges, such as the high manufacturing costs of curative cell and gene therapies, which prevent broad application of fully developed therapies. This is demonstrated not least by the example of the once promising pioneer Bluebird Bio. New, result-oriented approaches are needed here to drastically reduce costs.”