The cost of a molecule at industrial scale, known before the factory is built.
Arsenale Bioyards is building Industrial Biological AI,
the intelligence layer that makes industrial biology scalable.
Arsenale starts from the molecule a client brings, designs the entire production process around it, from fermentation to purification, predicts what the molecule will cost to produce at industrial scale, and confirms the prediction in its own equipment. The client receives kilograms at a known cost, through a process that keeps learning.
WHY NOW?
AI can design new molecules faster than ever.
Almost none of them can be produced at scale.
This is an old problem getting worse.
More than 180 companies now design molecules with AI, and the number grows every month. Production has not moved: developing a process to produce one molecule still takes tens of millions of dollars and up to twenty years. Most molecules are never even attempted.
INDUSTRIAL BIOLOGICAL AI
Language AI needed a corpus.
Physical AI needed a robot. Biological AI needs a factory.
There will be no AlphaFold moment for biomanufacturing. Whoever wants the model has to make the data first.
Reading biology was the bottleneck, then writing it, then designing the molecule. Each fell to an intelligence trained on a public body of data. Producing molecules has none: the runs that would fill it happen inside commercial operations and stay there. Arsenale is building Industrial Biological AI around proprietary, sensorized equipment that makes its own data.
Three things that set biology apart
-
Biology has no first-principles simulator. Arsenale's intelligence learns from real runs.
-
A fermentation takes days, so each run has to carry more. Arsenale designs its runs as interventions and measures them densely, so that cause can be told from coincidence.
-
Biology behaves differently as conditions change. A prediction counts once it has been validated at each scale.
→ Read our CEO and Co-Founder’s essay ‘Biological AI: There Will Be No AlphaFold Moment for Biomanufacturing’ here.
THE PRECEDENT
In 1987 TSMC split designing chips from making them.
Arsenale is doing it for biology.
Before TSMC, a chip had to earn its own factory, so only the largest volumes were made. TSMC owned the factory and the process, sold the process qualified, and two orders of magnitude more chips followed. Today a molecule has to earn its own factory. Arsenale breaks that link: it owns the factory and the process, and sells the process qualified, with the cost attached.
→ Read our CEO and Co-Founder’s essay ‘The Semiconductor Lesson Biology Has Not Learned’ here.
THE ARSENALE METHOD
Molecule first.
A production process is usually developed like a relay: the organism is chosen first, the fermentation is built around it, the purification takes whatever the fermentation delivers, and what the molecule costs to produce is known with certainty only after the industrial run. Arsenale starts from the molecule a client brings and the properties and the cost it needs to deliver, designs the process around it, and chooses the organism last, according to the process specification.
Arsenale uses intelligence to start from the molecule, design the whole process, predict its cost at industrial scale, and prove it.
Our Biological AI is one system, an intelligence layer designed as a whole, creating upfront cost transparency and a molecule-specific, optimized and continuously self-improving process.
Production has no intelligence in it, so each step of the process creates compromises that compound. Each step is optimized on its own, within the limits of the hardware it inherits.
Arsenale Intelligence
decides what’s worth testing.
Before any experiment, Arsenale's intelligence lists every plausible way of producing the molecule and eliminates what chemistry, the limits of separation, regulation and the target cost already rule out. The short list that survives is built on intelligence rather than on empirical testing; experiments then confirm it. Sometimes none survives, and finding that out before a single experiment is the least expensive outcome in the whole sequence.
HOW WE WORK
One intelligence, from the molecule to the factory.
Arsenale uses intelligence to start from the molecule, design the whole process, predict its cost at industrial scale, and prove it. The work follows four phases, connected by one intelligence layer, and every molecule begins with the first.
01 DERIVATION
Can this molecule reach its target cost, and by what method?
Answered in silico, on the computer, before anything is cultured.
02 DEVELOPMENT
Tested in Arsenale’s bench and 500L reactors under the conditions of an industrial tank.
Which organism, fermentation and purification perform?
03 VALIDATION
Run in steel, in the equipment that will produce the molecule, and the cost measured rather than estimated.
Does the method perform at industrial scale?
04 PRODUCTION
Kilograms at a known cost, from a method that keeps learning.
Production lines replicate the validated line, and every run improves the method.
THE ARSENALE STACK
Four layers. One intelligence.
Physical AI gave intelligence a body and called it a robot. An intelligence cannot act on a living culture, or learn from one, through a screen.
Arsenale is building the equivalent for industrial biology: the factory is the robot.
The brain
Intelligence
Three intelligences, one system: one designs the process, one develops it, one runs production. What production teaches returns to the design.
The nervous system
Sensing and control
Sensors read the biology itself, at every scale: the state of the cells, their metabolic activity, the chemistry of the broth. A conventional bioreactor records only the conditions in the tank.
The muscle
Hardware
Proprietary, sensorised reactors from bench to industrial scale, and their purification lines, all running industrial conditions. The hardware exists to generate the data the intelligence learns from.
The backbone
Utilities and modules
Power, cooling, skids and energy storage, built as standard modules. Reconfiguring a factory is an assembly job rather than a construction project.
The factory
Split these layers across companies and the loop that learns breaks. The stack is owned whole, from model to factory.
Arsenale Alfayard
Arsenale’s Alfayard is a fully operating pilot site based in Pordenone, Italy: the place where production methods are developed
under industrial conditions and the first data is generated.
Every molecule Arsenale works on passes through it before anything larger is built. The bench and pilot-scale reactors are designed by Arsenale and fitted with the sensing the intelligence needs: they read the biology itself, not only the conditions in the tank. The same instruments sit on every vessel, so a reading at bench and a reading at pilot scale can be compared directly, and every run returns data the intelligence learns from.
Work with us
If you are interested in working with us or learning more about our technology, contact us via
the contact form or email at info@arsenale.bio

