01
The science is not the constraint
Strains work. Processes work. What does not exist is capacity a company can buy at a price and a speed the market can absorb. So each company builds its own plant, engineers it from scratch, and pays the engineering bill again on behalf of an industry that has already paid it many times over.
02
The green premium is real, and thinner than plans assume
Some buyers will pay something for how a thing was made, and in the right category that premium opens the first door. It is usually narrower than a business plan assumes, and it tends to thin out on the way from launch volume to real volume, where procurement compares a line item to the incumbent. So the target is set from outside: a price that already exists on somebody's spec sheet. Everything after this point is aimed at that number, because a product that only works above it stays a small product.
03
Nobody can afford to move first
A company with a working process needs investment or an offtake contract to reach commercial scale. The people who would sign either one want to see commercial scale first. So the process that works at one cubic metre stays there, and the money runs out in the gap between them. That deadlock does not break with better biology or a better pitch. It breaks when proving a process at volume stops requiring a company to bet its balance sheet on a building.
04
Scale-up has no middle
A process proves itself in a few litres. The next number anyone cares about is commercial. The equipment in between either does not exist near you or costs more than the question is worth, so companies jump: a 200,000 litre plant designed off a 200 litre run, on the assumption that mixing, oxygen transfer, shear and heat removal will behave at a thousand times the volume the way the model says they will.
Often they do not, and the discovery arrives after the concrete has been poured. The honest position is that nobody knows in advance. You can work downwards instead, with a scale-down model that mimics the large vessel, and that is worth doing, but it gets complicated and slow quickly, and the equipment is priced as though it were the answer rather than the experiment: around $200,000 for a set of benchtop reactors, or $2M for a set of twelve 15 mL vessels. Neither number makes sense to us for what it tells you.
A fleet has the intermediate steps by construction, because an intermediate step is simply a smaller module of the same design, with the same controls and the same data coming off it. You find out by running the thing you will run, one size down.
05
Everything expensive about biology is the building
A high-complexity plant is $$$$$ and 48 months. A mid-complexity one is $$$$ and 24 months. Most of that is not the tank: it is the design house, the site-specific engineering, the controls built in place and the commissioning. Those costs land in the price of every kilogram that plant will ever make, which is why biology loses to chemistry on cost long before the biology is at fault.
06
Going big has already been tried, with billions
The standard advice is to build the largest plant you can finance and let economies of scale do the rest. That experiment has been run, repeatedly, at full size. DuPont opened a $200 million cellulosic plant in Nevada, Iowa in 2015 and described it as the first of many; it closed two years later and was sold to a German buyer who converted it to make methane. Abengoa started up in Hugoton, Kansas in 2014 and shut it down in 2015. The Beta Renewables plant in Crescentino changed hands out of its parent company's bankruptcy. Amyris walked away from its own production targets and indefinitely delayed its second Brazilian plant after finding out how differently yields behave at commercial volume.
None of those plants failed because the tank was too big. They failed because one enormous single-purpose bet has to be correct the first time, committed years before anyone knows whether the process, the strain, the feedstock price or the market will hold. Scale is not the mistake. Scale as the first move is.
07
So make the plant a product
One module, engineered once, manufactured over and over to the same drawing. The engineering is paid for a single time and amortised across every unit that follows. That alone moves a plant from $$$$$ and 48 months to $$ and under three months, and it is the whole reason the rest of this thesis is arithmetic rather than hope.
08
Capacity in steps you can finance
A bespoke plant is one configuration at one volume, committed to before anyone knows which volume the market will take. The pad, the utility spine and the aisles are drawn once, and modules are added onto them as demand arrives, so capacity follows the order book instead of leading it by four years. Getting the guess wrong costs a module rather than a plant.
09
Nobody should pay for hygiene they do not need
Most fermentation capacity was built for biopharma, so a food ingredient rented on it is paying for pharmaceutical steel, pharmaceutical air and pharmaceutical paperwork. The module is configured to the tier the product actually requires. That is where a sub-$5 shelf price stops being a rounding error and starts being reachable.
10
Downstream is where the cost hides
Attention goes to the tank. The money leaves after it. Getting a product out of broth means some sequence of disc-stack or decanter centrifuge, homogeniser or bead mill, depth, micro or ultrafiltration, tangential flow, evaporation, precipitation, extraction, and then a dryer: spray, drum, ring, belt, vacuum or freeze. There are dozens of defensible combinations, and most projects pick theirs late, out of a pharmaceutical catalogue, after the vessel has already been bought.
So downstream in the fleet is standard. One specified train per hygiene tier, the same at every site, engineered and priced once like the rest of the module. A company that needs something particular adds its own step after it, rather than having the module redrawn around it.
And the reference for that train is not biopharma. Dairy, starch, sugar, corn wet milling and vegetable oil separate, evaporate and dry enormous tonnages every day at commodity cost, on equipment almost nobody in biotech thinks to price. A spray dryer that turns milk into powder for cents a kilogram is not a different machine because the powder is a protein.
11
The biggest cost is the feed, and the plant is built away from it
Media is usually the largest line in a fermentation cost model, whatever the organism eats. Sugars for most, but also starch hydrolysates, molasses, glycerol, methanol, acetate, plant oils, protein hydrolysates and gases such as CO2, hydrogen or methane, plus the nitrogen source, salts and trace metals that go with them. All of them price by geography more than any process improvement a company will make in a year, and plants still get built where the capital and the engineering firms are, then buy the feed shipped in.
A module is small enough to stand next to the feed instead: at the mill, at the starch plant, beside a refinery vent, next to a biogas digester, or on top of a side stream somebody is currently paying to dispose of. The same organism at the same titer has a different cost per kilogram depending on nothing but where the tank is standing.
12
Say the word biotech and the quote goes up
Ask an equipment supplier, an engineering firm or a contract manufacturer for a number, and the number moves the moment the word biotech is in the room. Not because the pump is different, but because the sector is assumed to be venture-funded and in a hurry, so the same work gets priced against the round rather than against the work. Manufacturing the module ourselves takes that conversation off the table: a unit costs what a unit costs, the same for everyone, and nobody is quoted against their cap table.
13
Nobody in the supply chain is paid to make it cheaper
This is not bad faith, it is arithmetic. A vessel maker sells one tank at a time and earns on the engineering wrapped around it, so twelve weeks of design work is revenue rather than waste. An engineering house is paid for drawings, which makes reusing last year's drawings a discount it has no reason to offer. Every party is doing exactly what its business model rewards, and what those models reward is a plant that has never been built before. The only company with a reason to drive the cost of a module down is the one that has committed to building the same module a thousand times.
14
Half the steel sits idle, and pooling fixes it
A company that funds its own plant typically runs it around half the time. Campaigns are lumpy. Turnarounds, seed trains and validation eat weeks. Demand ramps more slowly than the build, so the plant is sized for a peak that arrives late and then waits through the trough. The result is the worst of both: the company feels short of capacity and is paying for capacity it is not using.
Individual demand is spiky. Aggregate demand is not. When one company is in a research campaign while another is in production, the same steel runs far closer to full and the cost per liter falls for everyone on it. That is the argument for The Fleet: one design, built repeatedly, filled from a book of demand rather than from a single forecast.
15
Contamination is a black box, and it is expensive
A lost batch is usually a lost explanation. Contamination is found days after it happened, in a vessel carrying a handful of probes and no record of what the seals, the valves, the air or the utilities were doing at the time, so the cause is a guess and the fix is a procedure change nobody can verify. It is one of the largest unbudgeted costs in fermentation and the least examined. Identical vessels, fully instrumented and reporting into one system, make it traceable: a failure seen once anywhere in the fleet becomes a known cause with a known correction, and every module gets the correction.
16
Many tanks do not fail less, they fail survivably
This is a narrower claim than it is usually made into. Running twenty vessels does not lower the chance of a contamination event; more connections and more transfers mean more chances of one. What changes is the size of the worst case. A lost 200,000 litre batch is a quarter of the year and a missed contract. One module in twenty is five per cent and a rerun.
It only holds if the failures are genuinely independent, which is a design requirement rather than a happy accident. Each module carries its own seed train, its own media preparation and its own clean-in-place, so a bad batch stays inside one unit instead of walking out through a shared utility. Scale-out does not make biology more reliable. It makes it survivable, which is what a supply contract actually needs.
17
Identical vessels are what make the AI worth having
A one-off plant produces data about itself. A fleet of identical vessels produces data that transfers. Jarvis runs the modules and learns how a vessel holds a setpoint, where a run drifts and what to correct, across every module everywhere. Every new module makes the ones already running better, which no bespoke plant can say.
18
A module can be financed, a plant gets written off
A first-of-a-kind facility is close to the hardest thing in industry to finance. No comparables, no residual value, no second buyer, so it is equity money or nothing, at the price equity charges. A unit that has been manufactured a hundred times is a different instrument: known cost, a service history across the fleet, a resale value, and somewhere else to go if the company that ordered it changes plan.
That makes it leasable, insurable and lendable against, which lowers the cost of capital sitting behind every kilogram it makes. It is also why the plants above ended as write-offs rather than as assets. Nobody else could use them.
19
Ownership is your decision, not our business model
Rent vessel time and own no steel. Buy modules and run them on your own floor. License a site: fund and host it, and run it as part of The Fleet on our design, our controls and our operating standard. The terms change; the module, the controls and the operating standard do not. A company that starts by renting is not starting over when it buys.
20
What that adds up to
Engineering paid once, hygiene matched to the product, vessels run near full and controlled by something that improves with every batch in the fleet. Compound those and biology reaches price points it has never held: commodity categories at a few dollars, and molecules that today carry a plant's capital cost in every gram made for a fraction of it. That is the bet, and it is a manufacturing bet, not a biology one.
