Every pharma and CDMO leader is being asked to make more product, faster, without a capital program that takes years to deliver. The default answer has always been physical: build a new line, a new suite, a new site. But 52% of CDMOs are already capacity constrained or effectively sold out, and capital is no longer cheap.
Aizon co-founder and CEO Pep Gubau and Independent Director Scott Waldman sat down to make the case for a different answer: most of the capacity you need is already inside the facility you own.
More than half of your capacity is already built
The industry treats a facility as fully utilized at 80%. The real picture is worse. Around 56% of physical capacity is held hostage by three losses:
- Availability loss (–30%): Manual documentation, long changeovers, cleaning
- Performance loss (–20%): Fear-based set-points, micro-stops, conservative machine speeds
- Quality loss (–6%): Batch rejections, deviations, manual sampling delays
That puts effective capacity at roughly 44% of theoretical.
Capacity is not square meters, it's output over time
Physical capacity means high CapEx, low ROCE, and delayed single-step growth. Digital capacity means little or no CapEx, faster implementation, and immediate incremental growth, pushing the limits of the equipment and people already in the building.
The hard part isn’t the model
"Anyone can train a model," Pep says. "That's the easy part."
The difficulty is operationalizing AI in a GxP environment, at scale, across sites and countries, traceable, auditable, explainable, validated. Which data created which model, which model created which insight, which insight drove which decision, and how that affected the batch.
That is why 87% of pharma AI pilots never reach production.
Two examples from the conversation show what changes when it's solved. A leading pharma manufacturer with declining yields and no identifiable root cause integrated 6Bn+ data points, isolated the parameter driving 55% of the variability, and lifted average yield 4% in three months worth more than $30M. A biopharma manufacturer rejecting one batch in four cut rejections to nearly zero, with models predicting deviations at 91% accuracy before they occurred.
Don't wait for the perfect program
The industry's habitual failure mode is the never-ending transformation initiative: 3+ years to first ROI, $10–50M spent, no quantified outcome.
The alternative is progressive. Land one painful use case in a quarter. Prove the ROI. Expand. As Scott puts it: "We don't have to wait for the perfect program. We can make those changes today."
Watch the full conversation below:
