Feedstocks: what is actually available

In one paragraph. This chapter asks a simple question about each of Vietnam's agricultural residues: could you actually buy it, in quantity, cheaply, in one place, and would it feed a microorganism. The answer that matters is that the best feedstock in the country is the wet pulp left at cassava starch factories, because it is large, concentrated at about 120 known addresses, currently sold cheaply as cattle feed, and needs only the enzyme step those factories already run. The answer that surprises people is that Vietnam has no spare molasses, which is the substrate everyone assumes a tropical country has. The general rule that emerged is that dry and concentrated residues are available and wet, scattered ones are not, and that a wet residue must be used where it is made.

8.1 Cassava starch-mill residues: pulp, peel and processing wastewater#

The resource. Roughly 10.5 million tonnes of fresh roots a year from 517,800 hectares. About 120 industrial-scale starch factories operate in 27 provinces, with a combined design capacity of 11.3 million tonnes of fresh tubers a year and an actual processing capacity of about 8.62 million tonnes. The size distribution is 70 to 74 percent small scale, 16 to 20 percent medium and about 10 percent large. Note that 8.62 million tonnes is a capacity figure rather than a measured throughput.

How much pulp that produces is not published, and it is the most important unmeasured number here. An external reviewer offers the first estimate this project has seen, at 1.7 to 2.1 million tonnes of wet pulp a year, roughly 400,000 to 500,000 tonnes dry, derived from a pulp yield of 10 to 15 percent of wet root mass. That estimate does not reconcile with itself: 10 to 15 percent of the 8.62 million tonnes processed is 0.86 to 1.29 million tonnes, and reaching 1.7 to 2.1 million tonnes would require 16 to 20 percent of the entire national crop. The wider literature generally puts pulp at 15 to 20 percent of root weight, so the tonnage may be the sounder half of the estimate, but one of the two figures is wrong and neither is quoted in this report. Open question CF-G1, closed by action 4, which is a question three factories can answer directly.

Composition.

Stream	Total carbohydrate	Fermentable sugar	Crude protein	Ash	Inhibitors
Cassava pulp	72.72 percent of dry weight	48.63 g/L glucose in hydrolysate	0.93 percent	1.52 percent	Furfural not detected
Cassava peel	64.51 to 69.6 percent	not reported	10.60 percent	6.54 percent	Cyanogenic glucosides present, concentration not reported
Cassava processing wastewater	9.6 to 37.5 g/L starch	not reported	nitrogen 780 to 1,300 mg/L	not reported	Cyanogenic compounds, nitrates, nitrites, sulfates, heavy metals, concentrations not reported

A Vietnamese commercial supplier lists cassava residue pellets at 2 to 4 percent crude protein, 12 to 18 percent crude fibre, 30 to 35 percent starch, moisture at or below 13 percent and ash at or below 8 percent, sold as an energy and fibre feed for cattle. That is direct evidence that a market for this residue already exists, which the academic literature did not surface. [VN-direct, commercial source]

The nitrogen problem. Cassava pulp, the main stream, is starch-rich and nitrogen-poor, at 0.93 percent crude protein, and cannot support microbial growth without nitrogen being added. This is a real cost line that gets left out of arguments about free waste feedstock. Peel is a different matter at 10.60 percent, and the wastewater stream at 780 to 1,300 mg/L of nitrogen is interesting alongside the solid residue rather than instead of it.

An external estimate puts the cost of that nitrogen at USD 80 to 130 per tonne of protein produced, buying it as urea or ammonium sulfate at a carbon-to-nitrogen ratio of 8:1 to 12:1. The arithmetic is checkable by anyone who can price urea in Vietnam, and doing so would convert this caveat from a warning into a line in a model. It is not quoted as a finding here.

Performance as a substrate, from chapter 10: cassava peel hydrolysate supports Candida utilis at 49.1 percent crude protein and 0.52 g/g biomass yield on acid hydrolysate, and 56.7 percent at 0.44 g/g on enzymatic hydrolysate. Cassava pulp fermented with S. cerevisiae and C. utilis raised crude protein from 2.59 to 33.34 percent. Cassava bagasse in submerged culture gave 8 to 12 g/L of biomass at 40 to 50 percent protein. Nothing above bench scale. Open question F-G8.

Verdict. This is the strongest feedstock candidate in Vietnam and the only one where volume, concentration, existing processing infrastructure, low value in its current use and demonstrated substrate performance line up at once. It is also the candidate whose headline volume nobody has measured, which is a reason to make the phone call before writing a business plan.

8.2 Cassava roots and dried chips#

Storage is the binding constraint. Fresh roots deteriorate within two to three days of harvest. Running a plant year round therefore requires conversion to dried chips immediately after harvest, which adds cost, and the accessible literature gives little hard data on drying energy or safe moisture targets. Fresh roots are about 50 percent moisture, with roughly 30 percent starch and 5 percent sugars; dried cassava is reported at about 80 percent fermentable material.

One source reports a warehouse storage assumption of 50 percent weight loss per month for dried chips. That figure is almost certainly wrong as transcribed and is not used anywhere in this report. See contradiction CF-C4.

Verdict. Buying roots means competing with 120 starch mills for the same tonnage and inheriting a two to three day clock. Buying chips means paying for someone else's drying. Neither is as good as taking pulp at the mill gate.

8.3 Cassava-derived glucose#

Covered in section 10.8 and repeated here because it is the bridge between this chapter and precision fermentation. Enzymatic hydrolysis of cassava starch reaches 82.1 to 94.1 dextrose equivalent, with 176.41 to 267.9 g/L of reducing sugars and 2.45 g/L/h for raw starch hydrolysis. Cassava glucose syrup has supported recombinant D-amino acid oxidase production in Trigonopsis variabilis at 166.9 units per gram of cell dry weight, which is an enzyme rather than a food protein. In ethanol fermentation cassava performed close to corn: 57.0 g/L at 93.5 percent efficiency in 36 hours, against corn flour at 61.0 g/L and 95.1 percent in 48 hours. No like-for-like industrial cost or purity comparison against corn or sugarcane glucose exists.

8.4 Rice straw#

The resource is not reliably quantified, and this matters. Four estimates of Vietnam's annual rice straw are in circulation and the extremes differ by more than a factor of two:

roughly 43 million tonnes, from the IRRI and agriculture ministry basis of a one-to-one straw-to-paddy ratio applied to 43.12 million tonnes of paddy
42 to 47 million tonnes, from a 2025 Vietnamese mushroom industry report
70.7 million dry tonnes, in one Vietnamese paper
approximately 97 million tonnes, in an energy-potential study that itself notes "there is still a significant lack of data about its availability and produced quantities"

The same energy study gives 9 million tonnes of rice husk a year.

The position taken here: quote roughly 43 to 47 million tonnes, state that this rests on a straw-to-paddy ratio rather than on measurement, and note that the two higher figures exist and are not corroborated. See contradiction CF-C1.

Competing use. Across Southeast Asia rice straw is split between open-field burning, ploughing back into the soil, livestock feed, mushroom substrate and small industry. In Thailand in 2015/16, of 26 million tonnes generated, 56 percent went to animal feeding, fermentation and mushroom cultivation, 20 percent was left in the field and 22 percent was openly burned. In the Philippines in 2015, 60 percent was burned in fields and the source states that no market was established. In Punjab, India, more than 80 percent of 23 million tonnes was burned. The equivalent Vietnamese shares could not be confirmed; see contradiction CF-C2.

Processing burden. Rice straw is woody and needs pretreatment before the sugars can be released. Best reported results: about 94 percent glucose yield and 97 percent of theoretical ethanol for straw pretreated with ammonia and carbon dioxide, and about 0.567 kg of sugars per kg of straw for popping or steam-type pretreatment. Enzyme loading for well-opened material runs 6 to 10 filter paper units per gram of biomass, but some rice straw systems needed 23 units of cellulase plus 62 international units of xylanase per gram. Sugar concentrations of 50 to 100 g/L are reported after pretreatment. Composition is 38.33 percent cellulose and 19.73 percent hemicellulose.

Storage. Ensiling works: up to six months, and years under maintained airless conditions, at 50 to 70 percent moisture, avoiding extremes at or below 25 percent and at or above 75 percent. Dry matter loss of 10 to 15 percent occurs in the initial storage period.

Verdict. The largest residue in the country by tonnage, the least committed to an existing buyer, and the most expensive to use. Collection and pretreatment are both unsolved at Vietnamese cost levels, and the resource estimate itself is unreliable. A candidate for later, not for a first plant. Open question CF-G7.

8.5 Rice bran#

Volume is an open question. Vietnam milled 26.95 million tonnes of rice from 43.12 million tonnes of paddy in MY2024/25. At the usual bran fraction of 8 to 10 percent of paddy, that implies roughly 3.5 to 4.3 million tonnes of bran. But USDA records rice bran and broken rice demand as feed at only 730,000 tonnes in CY2024, rising to a forecast 740,000 tonnes in CY2026. The gap is unexplained by the sources consulted and could be direct on-farm feeding, bran oil extraction, export, or bran that is never separated at small mills. See open question CF-G3.

Composition and stability. Rice bran is reported at 59.06 percent total carbohydrate. It goes rancid quickly through enzyme activity after milling and needs prompt stabilisation; the accessible sources give no storage life, moisture threshold or loss rate.

Competing use. Rice bran is a traded feed and oil co-product across Southeast Asia rather than a discard, so a fermentation plant would be outbidding existing buyers. No Southeast Asian price was found in the literature.

8.6 Rice husk#

Approximately 9 million tonnes a year in Vietnam. Husk is widely used as fuel, in mills and in industry. One source reports 70 percent incineration in rural Vietnam, which could not be confirmed in the underlying paper; see CF-C2. The only price found anywhere was 1.15 Philippine pesos per kilogram in 2013, roughly 1,150 pesos per tonne. Husk is high in silica and poor as a fermentation carbon source; it is listed here for completeness rather than as a candidate.

8.7 Sugarcane molasses#

Vietnam does not have spare molasses. Molasses trade in 2022: imports of USD 78 million, up 36 percent on 2021, from India at 46 percent, Indonesia at 20 percent and Cambodia at 15.3 percent; exports of USD 5.4 million, to Korea, the Philippines and Australia. Quantities are not given in the source, only values. A country that imports about fourteen times more molasses by value than it exports does not have an unused molasses stream.

The upstream reason. Vietnamese sugarcane is small and shrinking: 127,000 hectares and 7.498 million tonnes of cane in the 2020/21 season, with 25 mills of total crushing capacity 110,000 tonnes of cane per day, producing 0.763 million tonnes of sugar. By 2025 the industry was reported at approximately 25 to 30 operating mills, down from more than 40 previously, with 2024/25 sugar output over 1.2 million tonnes by the end of May 2025 and inventories above 70 percent of output. The GFI APAC and Hawkwood study scored Vietnamese sugar capability at 26 out of 100, on 2024 figures of 1.35 million tonnes produced, none exported and 110,000 tonnes imported. These sources use different years and methods and should not be assembled into a trend line; see contradiction CF-C5. All of them point the same way.

Composition, for completeness. Molasses is 45 to 50 percent sugar, with 56.8 g/L sucrose, 31.5 g/L glucose and 33.8 g/L fructose reported after pretreatment. It carries calcium, magnesium, iron, copper, manganese and zinc, reduced by pretreatment. It stores better than any other feedstock here but still loses 2.65 to 11.20 percent of its fermentable sugar, most of that in the first two months, and should not be held above 40 degrees Celsius, which is a live constraint for uninsulated tank storage in southern Vietnam. Availability is seasonal, at 200 to 240 days a year.

Verdict. The best substrate chemistry of any feedstock considered, and the one Vietnam cannot supply. A Vietnamese plant running on molasses runs on imports, which is defensible on financial grounds but gives up the food-security argument.

8.8 Sugarcane bagasse and vinasse#

Bagasse is already committed. It is the boiler fuel and cogeneration feedstock at the mills, and is also used in alcohol processing and paper. One source reports about 70 percent of Vietnamese bagasse going to cogeneration with over 30 percent described as excess with no practical use, but those shares could not be confirmed in the cited paper; see CF-C2 and open question CF-G8. Across the region, bagasse in the Philippines and Cambodia is reported as sold to neighbouring countries, and in Myanmar as largely burned in the open.

Composition: 65 to 86.9 percent total carbohydrate, 24.49 g/L xylose in hydrolysate, with hydroxymethylfurfural at 0.37 g/L and furfural at 0.38 g/L, both of which inhibit microbial growth. One source reports 12.50 percent crude protein for bagasse, which is implausible for this material and is flagged in CF-C3 rather than used. Lime-pretreated bagasse yields about 0.384 kg of glucose per kg; dilute acid routes are reported at near 70 to 90 percent glucose or total sugar recovery.

Vinasse, the liquid left after distilling ethanol, is 3 to 7 percent solids of which about 75 percent is organic matter, with glucose and fructose present but unquantified, plus glycerol, lactic acid, ethanol and acetic acid. It is available only during the cane season, 200 to 240 days a year. Despite that, vinasse produced the highest fungal biomass figures of any substrate in this study: Aspergillus oryzae at 118.5 g of dry biomass per litre and 44.7 percent protein, in flask culture. That figure is exceptionally high and carries a caution in section 10.1.

Verdict. Both are tied to a sugar industry that is itself contracting. Vinasse is chemically interesting and seasonally constrained; bagasse already has a buyer.

8.9 Seafood processing side streams#

Vietnam generates approximately 1 million tonnes of seafood by-products a year, more than 90 percent of it solid, comprising shrimp and crab shells, fish skin, bones and organs. Current processing value is about USD 275 million a year against a stated potential of USD 4 to 5 billion. Projected component streams include roughly 900,000 tonnes a year of fish heads, bones and entrails, about 490,000 tonnes of shrimp heads, 150,000 tonnes of fish oil, 146,000 tonnes of shrimp shells for animal feed and 100,000 tonnes of pangasius offcuts. By 2030, shrimp by-products are projected at 650,000 tonnes worth USD 80 to 100 million and pangasius by-products at 1.3 million tonnes worth over USD 500 million.

Why this belongs in a feedstock chapter. These streams are nitrogen-rich, which is precisely what cassava pulp lacks. The obvious pairing is a carbon-rich cassava residue with a nitrogen-rich seafood side stream, and it is geographically plausible because both the starch mills and the pangasius plants concentrate in the south. That pairing was subsequently tested against the literature in section 11.6, and the result tempers it: see open question CF-G5, now partly closed. A further caution from an external reviewer is that raw seafood waste carries variable salinity, 0.5 to 3.0 percent sodium chloride, and active spoilage bacteria, both of which are awkward in a continuous sterile fermentation. Early plants will probably buy mineral nitrogen and treat the seafood pairing as a later refinement.

8.10 Coffee, coconut and fruit processing#

Coffee is large: MY2025/26 production of 31.7 million 60-kilogram bags, roughly 1.90 million tonnes of green coffee, forecast at 32.5 million bags for MY2026/27, with robusta at 31.4 million bags against arabica at 1.1 million, on about 730,000 hectares of coffee area. Husk and pulp volumes were not established in this study; the literature search returned coffee husk as a solid substrate for fungal protein enrichment with numbers too thin to compare. Coconut and fruit processing side streams were likewise not quantified. See open question CF-G6.

8.11 Imported inputs, and the price a new protein must beat#

Vietnam produced 21.5 million tonnes of industrial compound feed in CY2024, up 3.4 percent, and was forecast above 22 million tonnes for CY2025. Total feed grain imports in CY2024 were 21.75 million tonnes. Corn: 4.2 million tonnes produced against 12 million tonnes imported in MY2024/25, forecast at 4 million against 11.8 million for MY2025/26. Wheat imports were 5.45 million tonnes. Soybean: 41,000 tonnes produced against 3.2 million tonnes imported, which is roughly 98.5 percent of supply.

A note on the two headline totals. Feed grain imports of 21.75 million tonnes and compound feed output of 21.5 million tonnes are close to each other, but they are different quantities and their near-equality is a coincidence rather than an identity: the import figure covers grains that go to uses other than compound feed, and the output figure covers feed made partly from domestic inputs. The claim they support is that Vietnam's feed sector runs on imports, which is robust, rather than that every tonne of feed is imported, which is not stated.

The reading. Imported feed inputs are not a distortion to design around, they are the market. A locally-fed microbial protein competing on feed-grade terms competes against delivered imported soybean meal and corn, at scale, in a market that already moves more than 21 million tonnes a year. That is simultaneously the hardest price target in this report and the largest opportunity in it. The delivered prices themselves were never established; see open question CF-G4. An external reviewer gives delivered soybean meal at Cai Mep and Ho Chi Minh City at USD 350 to 450 per tonne across 2024 and 2025, which if confirmed would be the first hard number on this line. It is not quoted as a finding until someone checks the basis.

8.12 Feedstock scoring table#

Dimensions: D1 resource volume and how reliable the estimate is, D2 geographic concentration and collection logistics, D3 competing use and what a plant would have to outbid, D4 processing needed to make it fermentable, D5 composition quality as a substrate, D6 seasonality and storage, D7 evidence quality.

Feedstock	D1	D2	D3	D4	D5	D6	D7
Cassava pulp and peel	High. Implied by 8.62 Mt of industrial processing capacity; the pulp tonnage itself is not published, and the one external estimate does not reconcile (CF-G1)	High. About 120 known factory sites in 27 provinces	High. Low-value ruminant feed at present; a traded but cheap market	High. Enzymatic starch hydrolysis only, already industrial at these sites	Medium. 72.72 percent carbohydrate but pulp is only 0.93 percent protein, so nitrogen must be bought in; peel carries cyanogenic glucosides	Low to Medium. Wet and unstable; no published storage life. Argues for co-location, not storage	Medium. Composition peer-reviewed; the Vietnamese market evidence is a supplier page
Cassava roots and chips	High. About 10.5 Mt crop, well documented	Medium. Dispersed at farm level, aggregated by mills	Low. Competing directly with 120 starch mills, and with six fuel ethanol plants, for the same tonnage	High. Same hydrolysis route	High. About 30 percent starch fresh, about 80 percent fermentable dry	Low. 2 to 3 days fresh; chips need drying at cost	Medium to High
Rice straw	Low to Medium. Best estimate 43 to 47 Mt on a ratio basis, with two uncorroborated higher figures in circulation (CF-C1)	Low. Field-dispersed, bulky, low density	High. Least committed residue; burned or left in the field in much of the region	Low. Pretreatment required; no cost per tonne of sugar published, no commercial deployment for this feedstock	Medium. 38.33 percent cellulose, 19.73 percent hemicellulose, 50 to 100 g/L sugar after pretreatment	Medium. Ensiles up to 6 months at 50 to 70 percent moisture, with 10 to 15 percent dry matter loss	Medium. The competing-use shares are regional, and the Vietnamese shares could not be confirmed (CF-C2)
Rice bran	Insufficient evidence. Implied 3.5 to 4.3 Mt against a recorded 730,000 t of feed demand, unreconciled (CF-G3)	High. Concentrated at mills	Low. Traded feed and oil co-product with existing buyers	High. No pretreatment needed	Medium. 59.06 percent carbohydrate; oil content complicates handling	Low. Goes rancid rapidly after milling; no published stabilisation data	Low to Medium
Rice husk	High. About 9 Mt a year	High. At mills	Low. Widely used as fuel	Low. High silica, poor carbon source	Low	High. Dry and stable	Medium
Sugarcane molasses	Low. Vietnam is a net importer; no domestic surplus	High. 25 to 30 mills	Low. Already bid up by MSG and alcohol producers, and supplemented by imports	High. Directly fermentable, no pretreatment	High. 45 to 50 percent sugar; the best substrate chemistry in this report	Medium. Best storability here, but 2.65 to 11.20 percent sugar loss, a 40 degree Celsius ceiling, and a 200 to 240 day season	High. Trade data plus two independent industry sources
Sugarcane bagasse	Medium. Tied to 7.5 Mt of cane	High. At the mills	Low. Committed to boiler fuel and cogeneration	Low. Pretreatment required; 0.384 kg glucose per kg with lime	Medium. 65 to 86.9 percent carbohydrate, but with HMF at 0.37 g/L and furfural at 0.38 g/L	Medium. Dry, but seasonal with the cane	Medium. Vietnamese utilisation shares unconfirmed (CF-C2)
Vinasse	Low. Tied to a contracting distilling sector	High. At the distilleries	Medium. Often a disposal liability	High. Liquid, directly usable	Medium. 3 to 7 percent solids and low fermentable sugar, but the highest fungal biomass figures in this study	Low. Available only 200 to 240 days a year	Medium
Seafood side streams	High. About 1 Mt a year, with component streams itemised	High. Concentrated at processing plants in the south	Medium. Partly used for fishmeal, oil and chitosan; a large stated value gap suggests slack	Medium. Wet, perishable, needs rendering or hydrolysis; variable salinity complicates sterile operation	High as a nitrogen source, which is exactly what cassava pulp lacks	Low. Highly perishable	Medium. Government-adjacent press, not peer-reviewed
Brewer's spent grain	Medium to High. On the order of 600,000 to 860,000 t wet a year, but this is arithmetic rather than a measured stream (CF-G11)	High. Four brewery groups	Medium. Low-value cattle feed, biogas and disposal	High. No pretreatment needed for fungal upgrading	High. Already 15 to 30 percent protein in the dry matter, so no nitrogen purchase	Low. Very wet and spoils fast; must be used where it is made	Medium. Generation rate and composition general, Vietnamese volume derived
Okara	Medium. Implied by 550,000 t a year of soybean food use at near one-to-one generation	Low. Dispersed across many small tofu and soymilk producers	Medium. Feed, fertiliser and disposal	Medium. Wet, needs stabilisation	Medium to High. Good protein, and fermentation raises availability	Low. 70 to 80 percent water, spoils very fast	Medium
Coffee husk and pulp	Insufficient evidence. Green coffee at about 1.90 Mt is well documented; husk volumes are not (CF-G6)	Medium	Insufficient evidence	Medium	Insufficient evidence. Surfaced only as a solid-state protein enrichment substrate with thin numbers	Medium	Low
Imported corn and soybean meal	High. 12 Mt corn and 3.2 Mt soybean imported	High. At ports and feed mills	Low by definition. This is the market, not a residue	High. Standard	High	High	High. USDA FAS

8.13 Feedstocks added in the second expansion round#

Brewer's spent grain on volume, on the rest#

Generation rate: 14 to 20 kilograms per hectolitre of beer, often simplified to about 20 kilograms per 100 litres, or 0.2 kilograms per kilogram of beer.
Vietnamese volume: beer output was 416.2 million litres in June 2025, up 5.7 percent year on year, and 2,132.6 million litres over January to June 2025, up 0.7 percent. The producers are Sabeco, Habeco, Heineken and Carlsberg. Applying the generation rate to an annualised figure of roughly 4.3 billion litres implies on the order of 600,000 to 860,000 tonnes of wet spent grain a year. That arithmetic is the author's, from a sourced rate and a sourced volume, and should be checked against a brewery. Open question CF-G11.
Composition: fresh spent grain is mostly water; the dry matter is mainly fibre-rich cell wall material plus 15 to 30 percent protein.
Current use and price: most goes to low-value local cattle feed, with smaller shares to biogas and energy, and some to landfill, composting or incineration. High moisture and a short shelf life make transport, storage and centralised processing difficult, which is the main reason higher-value protein upgrading is uncommon.
Upgrading: alkaline extraction followed by precipitation at the isoelectric point gives protein-rich fractions of around 40 to 73 percent protein at 8 to 50 percent recovery in practical food-style processes; harsher pretreatment extracts more but damages what the protein can do in a food. Solid-state fungal fermentation is the best documented feed-upgrading route. There is at least one visible commercial food-ingredient example and pilot-scale techno-economic studies, but feed and anaerobic digestion remain the routes that are ready today.

Verdict. The same co-location logic as cassava pulp applies, for the same reason: a wet residue that spoils fast has to be used where it is made. The difference is that spent grain starts protein-rich, which removes the nitrogen purchase that cassava pulp requires. Four brewery groups is a much shorter list of conversations than 120 starch factories.

Okara, the residue from tofu and soy milk , with Vietnamese scale inferred#

Okara is generated in almost one-to-one mass with the soybeans processed, is about 70 to 80 percent water, spoils very fast, and goes mostly to feed, fertiliser or disposal. Protein recovery routes are alkaline extraction, high-pressure homogenisation and enzyme hydrolysis, which raise solubility and digestibility and release peptides and isoflavones. Fermentation is the main upgrading route, with A. oryzae, Rhizopus oligosporus and Neurospora intermedia repeatedly reported to raise protein availability, improve digestibility and lower antinutrients. Most human-food work is at laboratory or pilot scale; the economic barrier is the cost of stabilising a fast-spoiling wet residue.

Vietnamese scale. Soybean food use is 550,000 tonnes a year, which at near one-to-one generation implies a large okara stream, though it is dispersed across many small tofu and soymilk producers rather than concentrated.

The catch worth stating plainly. Vietnam's soybeans are 98.5 percent imported. Recovering protein from okara recovers protein that was already imported. It is a waste-reduction and value-capture case, not a food-security case, and the project should not present it as one.

Cassava fuel ethanol stillage, and a competing bidder #

Vietnam operates six fuel ethanol plants with a combined design capacity of approximately 600,000 cubic metres a year, most of them running on cassava, and securing enough feedstock to run at full capacity "continues to face considerable difficulties". One producer has switched partly to maize, both domestic and imported from Argentina, Brazil and the United States, for more stable supply and more consistent quality.

Two implications. The stillage stream exists but is tied to plants that cannot fill themselves. More importantly, these plants are a competing bidder for the same cassava, which independently reinforces section 8.2: the opportunity is the starch mill residue and not the root. A fermentation plant buying cassava roots would be bidding against 120 starch factories and six ethanol plants at once.