The most expensive thing in a fermentation factory may be the moment when nothing is fermenting. A batch ends. The product leaves. The vessel is emptied, cleaned, sterilized and started again. The steel is impressive; the pause is not. Pow.Bio, a 2019 company now based in Alameda, has built its argument in that pause. Instead of making the vessel larger, it asks whether biology can be persuaded to keep working.
The short pour
- Pow.Bio separates microbial growth from product-making in a two-chamber continuous process.
- It sells process development, scale-up and manufacturing support to ingredient, food, chemical and biotech companies.
- In a company case study at ATV Technologies, a 3,000-liter run lasted 300 hours and generated more than 3× the material of a modeled fed-batch comparison.
- The best fit is a stable process whose economics are constrained by reactor productivity. It is not a cure for every organism, molecule or downstream bottleneck.
The trouble with a living factory
Precision fermentation gives microbes a peculiar occupation: eat a feedstock and make something useful - a protein, an acid, a lipid, perhaps an enzyme. Conventional fed-batch production lets the population grow, changes its environment to trigger production, harvests the result and begins again. It is closer to baking than to an oil refinery. Each loaf has a beginning and an end.
Continuous fermentation sounds like the obvious improvement until one remembers that the workforce is alive. Keep a culture running and two saboteurs appear. A contaminant can enter with the constant flow of media. More subtly, the production strain can mutate. A cell that spends less energy making the customer’s molecule may reproduce faster than its diligent colleagues. Evolution is an excellent optimizer and a dreadful shift manager.
“There’s a burst of productivity and then it drops to zero, a burst of productivity and then it’s back to zero.”Ouwei Wang, co-founder and CTO, describing batch production
That was the first thing continuous systems failed at: remaining continuous without being conquered by contamination or genetic drift. Pow.Bio’s answer is architectural. It gives growth and production separate rooms.
In the second chamber, cells have enough metabolic life to produce but not the building blocks to multiply freely. Wang once called them “zombified,” an image of uncommon accuracy for a pitch about industrial biotechnology. The arrangement reduces the evolutionary advantage of freeloaders and makes it harder for contaminants to take over. Pow.Bio is careful on the important point: the system is more resilient, not magically immune.
The week the tank met weather
The consequential test came in Compiègne, France. Pow.Bio had already taken a high-value dairy-protein process from two liters to 300 liters in Alameda. At ATV Technologies, it attempted two awkward things at once: transfer the process into somebody else’s facility and enlarge it tenfold to 3,000 liters. The schedule allowed less than ten weeks.
The team did not replace the plant. Existing bioreactors, pumps and tanks were reused; a tangential-flow filtration skid enabled cell recycling; media composition changed for continuous operation. This detail matters. A technology that works only after a factory is rebuilt is an architectural fantasy. Pow.Bio’s commercial case depends on retrofitting what customers already own.
Cassie, Pow.Bio’s model-predictive controller and digital-twin software, built a model of the run, tested configurations and adjusted settings as the evidence arrived. Then reality supplied the sort of joke reality enjoys: evaporation was higher than anticipated. Cassie changed permeation rates while the run continued. The software is not there to sprinkle “AI” on stainless steel. Its credible purpose is narrower and more useful - detect a process leaving the expected path and recommend or execute a correction before the batch becomes a postmortem.
That qualifier is not decorative. These are results for a particular protein process and a company-published comparison, not a law of nature. Pow.Bio’s home page reports gains from 2.5× for a yeast food protein to 6× for Gram-positive biomass. The range itself is the honest clue: biology refuses a universal multiplier.
What the customer is actually buying
Pow.Bio calls its model “fermentation as a service.” A customer may arrive with several strains and ask which one deserves a factory. Another may know its titer, yield and productivity but need better economics. A third simply needs a kilogram for regulatory work or market testing. Pow.Bio validates strains, tunes media and oxygen transfer, optimizes pH and feed rates, makes sample material and designs the path to scale.
The named projects show the breadth. MeliBio chose Pow.Bio to improve a process for bee proteins and enzymes. California Cultured moved plant cells from shake flasks into controlled reactors at Pow.Bio’s 25,000-square-foot, FDA-registered Alameda facility, producing commercial quantities of food-grade cultured cocoa. A Department of Energy-backed Agile BioFoundry collaboration is adapting R. toruloides yeast to make fatty alcohols used in detergents and surfactants.
In 2023, Pow.Bio said it had 30 clients and insufficient capacity for all incoming demand. Its $9.5 million Series A, led by Re:Food and Thia Ventures, funded the Alameda facility. The business now stretches beyond contract work. A 2025 partnership with Bühler packages Pow.Bio’s process and software with industrial engineering, installation, commissioning and support. In other words, the startup supplies the unusual recipe; the 165-year-old equipment company helps make it repeatable at somebody else’s plant.
The lesson hiding between the vessels
The transferable idea is not “use fermentation.” It is to stop making one asset perform two conflicting jobs. Growth wants conditions that reward reproduction. Production may want precisely the opposite. Pow.Bio separated them, instrumented the handoff and used software to attend to deviations rather than pretend deviations would not occur.
A team attempting the same logic elsewhere can copy four habits: map the idle time across the whole cycle; isolate stages with opposing incentives; reuse standard infrastructure wherever possible; and compare economics at equal output over equal time, including cleaning and turnaround. The glamorous metric is tank volume. The useful metric is saleable product per liter per hour.
Where the economics turn
Continuous operation is most attractive when reactor time dominates cost, demand supports long campaigns, the organism remains productive and filtration can keep pace. It weakens when batch economics are already good, product demand is small, a strain cannot tolerate prolonged operation, purification - not fermentation - is the bottleneck, or regulation makes process changes unusually expensive.
Pow.Bio’s competitive field therefore includes ordinary fed-batch plants, contract manufacturers and other continuous platforms. Its difference is less a singular machine than the combination of a two-stage biological process, live control and a service team willing to cross the ugly distance from flask to factory. The Bühler partnership is the next examination. A successful pilot can be tended by inventors. A platform must survive installation by people who did not invent it.
In June 2026, the World Economic Forum named Pow.Bio a Technology Pioneer. Awards are pleasant; uptime is better. The company’s most persuasive achievement remains prosaic: for 300 hours in a borrowed French plant, cells kept arriving, protein kept leaving and the system corrected for the weather inside the tank. Industry advances sometimes look like discovery. This one looks more like refusing to stop.