Fermentation: growing protein in a tank

In one paragraph. This is the chapter on growing microorganisms in tanks. Two things come out of it. First, biomass fermentation, where the organism itself is the product, is industrially mature elsewhere, has a plausible route to commodity prices, and has never been done in Vietnam: no Vietnamese study in the sources used any of the organisms that dominate the international literature. Second, precision fermentation, where the organism manufactures one specific protein, is much less mature than its press coverage implies, but two things changed in 2024 that make it approachable without building anything. The chapter also carries the cost benchmarks, which are quoted here with the grade of plant each one assumes, because otherwise they are not comparable.

10.1 Filamentous fungal biomass protein, also called mycoprotein#

No documented Vietnamese activity. Across the 40 sources screened in a search restricted to Vietnamese institutions, no Vietnamese study used Aspergillus oryzae, Rhizopus oryzae, Fusarium venenatum or Trichoderma species, despite these organisms dominating the international literature on microbial protein from agro-industrial waste. No Vietnamese study used food-processing wastewater as a substrate for fungal or yeast protein. [VN-direct]

Substrate performance, verified against the source. Filamentous fungi grown on 5 percent sugarcane vinasse, in 250 millilitre flasks with 100 millilitre of liquid, over 72 hours (Karimi et al., Fermentation, 2019):

Two cautions, and the second is stronger than it may look. These are flask numbers, not bioreactor numbers, and should always be quoted with that caveat. And the top figure is exceptionally high for flask culture, several times what commercial mycoprotein processes report. Nobody in this project or in the external review has read that paper's methods. It is the best substrate number in this report and the least tested, and anyone building an argument on it should open the paper first.

Other demonstrated substrates. Cassava starch and tuber in a 100 litre fermenter with 50 litre working volume (Sukara 1989). Rasped fresh cassava roots in 200 and 3,000 litre fermenters (Santos et al. 1983, CIAT). Sugar-ethanol vinasse at 45.55 percent crude protein (Nitayavardhana and Khanal 2010). Starch processing wastewater at 46 to 50 percent protein and 0.85 to 0.92 grams per cubic decimetre per hour in pilot airlift bioreactors (Jin et al. 2001). Rice polishings at 49.50 percent crude protein with Trichoderma harzianum in 5 and 75 litre fermenters (Ahmed et al. 2017, Pakistan). Rice straw pulp with a local Trichoderma reesei in solid-state fermentation, reaching 22 percent crude protein. [general]

State of the art. Metabolic engineering of Fusarium venenatum reached 61.9 percent protein with a 57 percent improvement in the rate of protein synthesis (Tong et al. 2023). [general]

10.2 Yeast single-cell protein#

Vietnam has strain and fermentation expertise but has not pointed it at bulk protein. The Food Industries Research Institute, FIRI, maintains a yeast culture collection, screens strains for industrial use, and has cloned and expressed industrially important enzyme genes in Pichia pastoris (Thanh 2009). FIRI also developed Saccharomyces cerevisiae A112 for zinc-enriched yeast biomass, reaching up to 12.88 milligrams of zinc per gram of dry biomass. This is functional-food work rather than bulk protein, but it is the closest existing Vietnamese capability to precision fermentation. [VN-direct]

Substrate performance. [general]

• Candida tropicalis on sugarcane bagasse hemicellulosic hydrolysate: 16.97 grams per litre biomass, 0.1767 grams per litre per hour, 60.05 percent protein, biomass yield 0.28 grams per gram, submerged culture.
• Candida utilis on cassava peel hydrolysate: 49.1 percent protein and 0.52 grams per gram yield on acid hydrolysate; 56.7 percent protein and 0.44 grams per gram on enzymatic hydrolysate. Submerged culture, scale not reported.
• Cassava pulp fermented with S. cerevisiae and C. utilis raised crude protein from 2.59 percent to 33.34 percent.

10.3 Microalgal and cyanobacterial biomass, adjacent to scope#

Growing algae in open ponds using sunlight is not fermentation. It is reported here because it is the only protein production that has actually been scaled in Vietnam, and because it shows what equipment exists in the country.

• My An Joint Stock Company of Tourism, Phu Vang district, Thua Thien Hue: a 20 square metre open raceway pond, run semi-continuously on My An mineral water at 1,055 milligrams per litre bicarbonate. About 10 grams of dry biomass per square metre per day, at 68.32 percent protein and 7.32 percent lipid. Supporting equipment: two 3 cubic metre stainless steel tanks, 200 litre inoculum containers, a paddle-wheel mixer, a 0.02 millimetre filter net and a spray dryer (Vu et al. 2017). [VN-direct]
• Vinh Hao Company, Binh Thuan: a 5,000 square metre Spirulina installation on mineral water, at 8 to 10 grams per square metre per day. Further small and medium installations exist in Hanoi, Nghe An and Thanh Hoa, with operators not named. [VN-direct]
• Scenedesmus sp. on rice straw hydrolysate, at the Graduate University of Science and Technology, Hanoi: 25.7 to 45.0 percent protein, 70 to 94 percent of the sugar used, in 500 millilitre flasks (Do Thi Cam Van and Pham Thi Mai Huong 2024). [VN-direct]
• 28 photosynthetic bacterial isolates from coastal Huế, four of them at 36.25 to 50.75 percent protein of dry biomass, in 250 millilitre flasks (Liên et al. 2025). [VN-direct]

The reading. The spray dryer, the harvesting equipment and the food-grade handling experience exist in Vietnam. The sterile stirred-tank fermentation train does not.

10.4 Gas fermentation on hydrogen-oxidizing bacteria#

No Vietnamese activity found. A hydrogen-oxidizing bacterial protein has been on sale in Singapore since 2022 under the Singapore Food Agency novel food framework (Tan et al. 2024). [VN-adjacent] The transfer assumption is that the Singapore approval shows the product class is acceptable to at least one ASEAN regulator, not that Vietnam could host production.

Economics. The lowest minimum selling price found for any dedicated protein process in this study belongs to this route: USD 2,070 per tonne, which is USD 2.07 per kilogram, for gas fermentation using wind-powered water electrolysis to make hydrogen and oxygen, with carbon dioxide from corn ethanol production (Jean and Brown 2024). This is a feed-grade product aimed at replacing fishmeal and soybean meal.

Lower figures appear in section 10.6, at USD 1.20 and USD 0.81 per kilogram, but those come from a microbial oil model rather than a protein model, so the two sets are not directly comparable.

Vietnam has neither the hydrogen supply chain nor the pressurised gas systems this route requires, and it is the least transferable of the categories considered. [general]

10.5 Precision fermentation: dairy, egg and heme proteins#

No Vietnamese activity found. What the published literature reports, one row per process with a published numeric titer:

Note on naming: Pichia pastoris and Komagataella phaffii are the same yeast under an old and a current name. Papers use both, and this table follows each paper.

Alpha-lactalbumin has no published titer at all. It is produced commercially by Perfect Day in Trichoderma reesei, with the process details held as trade secrets.

The economic threshold. Reviews state that cost-efficient food protein manufacturing likely requires titers above 50 grams per litre (Nielsen et al. 2023). The best figure in the table, 7.27 grams per litre, is about one seventh of that threshold; the weakest, 35 milligrams per litre, is about one fourteen-hundredth. [general]

Why the threshold bites downstream. At low titer the problem is not the tank, it is everything after it: recovering one kilogram of target protein means pushing 200 to 1,000 litres of dilute broth through microfiltration, ultrafiltration and chromatography. That is why the crude-against-pure cost gap in section 10.6 is a factor of thirty, and why cheap Vietnamese power and labour help least on exactly the step that dominates precision fermentation cost.

Intellectual property, which is a harder constraint than the science. [general]

• US patent 9924728B2 covers recombinant beta-lactoglobulin and alpha-lactalbumin production in Pichia pastoris / Komagataella phaffii and their use in food.
• Impossible Foods holds a patent covering not just the soy leghemoglobin gene but the P. pastoris production strain itself.
• All standard commercial Komagataella phaffii lineages descend from strain NRRL Y-11430, which is not available without restrictions on use, creating royalty and distribution constraints.
• OPENPichia is a licence-free K. phaffii chassis and toolkit published in Nature Microbiology (Claes et al. 2024). It traces back to the NCYC 2543 type strain from the UC Davis Phaff Yeast Strain Collection and is engineered to be easier to transform. It carries no royalty for industrial recombinant protein production, no constraints on distributing the strain, and liberal end-user distribution licences for both academia and industry. This is the single most actionable finding in the study for a Vietnamese entrant. It removes a licensing barrier. It does not touch the downstream recovery cost above, so it makes entry possible rather than cheap.
• FIRI already works with Pichia pastoris for enzyme expression (Thanh 2009), which is the same chassis organism. [VN-direct]

10.6 Capital intensity and production cost#

Every figure below is reported with the grade of plant it assumes, because that is what makes these numbers comparable or not. They must not be averaged. See contradiction F-C4.

The decision-relevant reading. Bulk biomass protein has a plausible route to commodity pricing, in the USD 1 to 3 per kilogram range, but only at tens of thousands of tonnes a year and only with cheap power. Precision-fermented dairy protein is an order of magnitude above conventional milk protein, and the gap is dominated by purification after the tank rather than by the fermentation itself. A country entering now enters the bulk category with a cost argument and the precision category with a development argument.

10.7 Infrastructure, utilities and the tropical question#

Substantially revised in this version. This section carries the cost argument for Vietnam, and the first version of this report stated it more confidently in the summary than the evidence here supports.

• Utilities are 18 percent of total production cost, and costs that scale with the facility itself are 58 percent, in the one single-cell protein techno-economic assessment in this study that resolves its cost structure section by section (Vlaeminck et al. 2023). [general]
• That assessment prices electricity at EUR 0.18 per kilowatt-hour, water at EUR 1.5 per cubic metre and labour at EUR 20 per hour. Vietnam's comparable figures are USD 0.08 per kilowatt-hour and USD 342 a month for manufacturing labour, which at a 40-hour week is about USD 1.97 an hour. Vietnamese power is therefore around 40 percent of the modelled price and Vietnamese manufacturing labour around a tenth of it. Those ratios compare a USD figure with a EUR figure directly, so they are approximate; at prevailing rates the power ratio is closer to 41 percent than 44 percent. No Vietnamese water price was found, so that line of the model cannot be substituted. Treat this as an indicative transfer of prices into a foreign model, not a costed result. [VN-direct] inputs applied to a [general] model
• What the power saving is worth, in the model's own terms. Utilities are 18 percent of production cost. Cutting the electricity portion of that by roughly 60 percent is worth on the order of 6 to 7 percent of total production cost. That is a real advantage and it is not the headline it is often made into. The first version of this report headlined "Vietnam's cost advantage is larger than the published models show" in the executive summary and put the qualifications here, which invited exactly the misreading an external reviewer made.
• The tropical penalty now has a first number, and it is the same order of magnitude. Added in this version. Aerobic fermentation of carbohydrate is strongly exothermic, releasing roughly 14 to 16 megajoules of heat per kilogram of dry cell weight produced. In a temperate climate that heat goes out through a cooling tower with little parasitic energy. In southern Vietnam the wet-bulb temperature means an evaporative tower cannot deliver water cold enough to cool a broth held at 30 to 35 degrees for a mesophilic organism: the temperature difference across the reactor jacket collapses towards zero and mechanical vapour-compression refrigeration has to run continuously. At a coefficient of performance around 3.5, removing 15 megajoules per kilogram of biomass costs about 1.2 kilowatt-hours, or USD 0.095 to 0.12 per kilogram at USD 0.08 per kilowatt-hour. Against a product selling at USD 1.20 to 2.07 per kilogram that is 5 to 10 percent of cost.
• Putting the two together. A 6 to 7 percent saving on power against a 5 to 10 percent cooling penalty means tropical cooling may consume the whole of the electricity advantage on the fermentation step. The honest position is that Vietnam's cost case rests on labour at roughly a tenth of European rates, on feedstock at the mill gate, and not on the power tariff. Both of these figures are desk estimates and neither has been checked against a plant, which is what action 13 is for. Open question F-G2.
• Two things work the other way. Reactor choice matters more than it looks: a continuous stirred-tank reactor converting acetate to single-cell protein uses about five times the electricity of an airlift bioreactor, because of the stirring (Vlaeminck et al. 2023), and airlift designs are what the pilot-scale fungal protein work used (Jin et al. 2001). Choosing airlift cuts the electrical load the chiller is competing with. And co-locating at a cassava starch mill puts waste heat from the site's own biogas engines within reach for drying, which is the other large thermal cost.
• Downstream cost hotspots, in order: harvesting and centrifugation, then dewatering and drying, then breaking the cells open where that is needed, then reducing RNA content by heating to about 60 degrees Celsius, then texturising. Cost rises sharply when the biomass is dilute, because centrifugation, filtration and drying all scale with the water that has to be moved or removed. Drying a slurry from 15 percent solids to 95 percent dry matter takes roughly 3.5 to 4.5 megajoules per kilogram of water evaporated, which is why co-location with a heat source matters as much as the power tariff. [general]
• In one algal case, 85 percent of production cost came from producing the starting biomass and 15 percent from processing it afterwards, with centrifugation the largest item in that 15 percent (Lina et al. 2022). [general]
• Location changes environmental impact by a factor of three or more. The carbon footprint of cellular-agriculture milk protein ranged from 5.5 to 17.6 tonnes of carbon dioxide equivalent per tonne of protein across New Zealand, Germany, the United States and Australia, and the water scarcity footprint from 88 to 5,030 cubic metres world-equivalent per tonne, depending on location and purification method (Behm et al. 2022). Vietnam was not among the modelled locations. Open question F-G9. [general]
• No published study gives a tropical-climate correction factor in megajoules per kilogram. The estimate above is a desk calculation from an external review, not a published figure. Grid carbon intensity does not change the physical kilowatt-hours a centrifuge or dryer needs, but it changes the climate impact of those kilowatt-hours and the economics under any carbon price. [general]

10.8 Cassava as a fermentation feedstock#

This section exists because the site-selection study scored Vietnamese feedstock on sugar alone. [general] unless noted

• Direct protein production. Cassava pulp fermented with S. cerevisiae and C. utilis raised crude protein from 2.59 percent to 33.34 percent, with biomass yield coefficients of 0.44 to 0.52 grams per gram of sugar. Cassava peel hydrolysate in submerged culture reached 49.1 to 56.7 percent crude protein in the biomass. Cassava bagasse in submerged fermentation yielded 8 to 12 grams per litre of biomass at about 40 to 50 percent protein. The largest clearly documented scales are bench-scale bioreactors and on-farm solid-state systems.
• Cassava to glucose. Enzymatic hydrolysis of cassava starch reaches 82.1 to 94.1 dextrose equivalent, with 176.41 to 267.9 grams per litre of reducing sugars and 2.45 grams per litre per hour of productivity for raw starch hydrolysis. Dextrose equivalent is a measure of how completely the starch has been broken down to glucose, where 100 is pure glucose.
• Cassava glucose as a fermentation carbon source. Cassava glucose syrup supported recombinant D-amino acid oxidase production in Trigonopsis variabilis at 166.9 units per gram of cell dry weight. This is an enzyme, not a food protein, and it is the only recombinant example found.
• Against corn and sugarcane. In simultaneous saccharification and fermentation to ethanol, cassava gave 57.0 grams per litre at 93.5 percent efficiency in 36 hours, against corn flour at 61.0 grams per litre and 95.1 percent efficiency in 48 hours. For single-cell protein, sugarcane bagasse alone gave 29.4 percent while a sugarcane and cassava peel blend reached 43.6 percent.
• What is still missing. No like-for-like industrial cost or purity comparison of cassava glucose against corn or sugarcane glucose, and no demonstration of recombinant food protein produced on cassava glucose. Open question F-G5.
• Vietnamese scale. About 10.5 million tonnes of fresh cassava roots a year from 517,800 hectares. Vietnam exported 2.3 million tonnes of cassava and cassava products in the first half of 2025, worth USD 711.5 million, ranking third in the world. [VN-direct]

10.9 Fermentation scoring table#

Scores are High, Medium, Low, or "insufficient evidence". This is a structured judgement, not a calculation: the scores summarise the evidence in the adjacent column rather than being derived from a formula.

Dimensions: D1 feedstock availability in Vietnam or mainland Southeast Asia, D2 regulatory pathway, D3 capital intensity, D4 infrastructure and utilities, D5 technology readiness and time to market, D6 transfer to a tropical lower-middle-income setting, D7 quality and independence of the evidence base.

Fungal biomass protein (mycoprotein)#

Yeast single-cell protein#

Microalgal and cyanobacterial biomass (adjacent to scope)#

Precision fermentation: dairy proteins#

Precision fermentation: egg proteins#

Precision fermentation: heme proteins#

Gas fermentation (hydrogen-oxidizing bacteria)#