The findings in brief

In one paragraph. Vietnam has no legal route to sell a new kind of protein as human food, because the relevant regulation contains no category for it. Two ways around that were found, and both can be tested with a letter. Setting regulation aside, the category that fits Vietnam best is growing fungus or yeast on the wet pulp left over at cassava starch factories and selling the result as shrimp feed. Vietnam's real advantages are cheap labour, cheap electricity, a very large and almost entirely imported animal feed bill, and concentrated agricultural side streams. Its real disadvantages are no sugar surplus, no industrial fermentation tanks suited to this, and no regulatory framework. The most promising surprise in the study is duckweed, which needs ponds rather than tanks and already has thirty years of Vietnamese research behind it, though only as feed.

The eight findings that matter#

1. The binding constraint is legal, not technical. Decree 15/2018/ND-CP contains no novel food category. There is no defined route to market for a fermentation-derived or otherwise new protein ingredient intended for human food. Every category assessed in this report is gated by that single fact. See chapter 7.

2. Two routes around it were found, and both are cheap to test. Revised in this version. The animal feed regime under the agriculture ministry has a named authority, a list of permitted ingredients and a defined dossier. The food regime has none of those. The published thirty-working-day clock applies to feed additives and to raw materials already on the permitted list; how a genuinely new raw material joins that list is not documented in any source consulted, so the feed route is a defined process of unknown duration rather than a fast one. Separately, Aspergillus oryzae, the koji mould behind Vietnam's own tương tradition, has centuries of documented food use on substrates that also have a history of use in Vietnamese food, which may place a modern koji biomass product outside the novel food question entirely. Two letters, to two parts of the same ministry family, would resolve both. See chapter 7.

3. On feedstock, the conventional view is wrong in both directions. Vietnam is a net importer of molasses, which confirms the sugar shortage and extends it to the one ready-to-ferment sugar stream the country has. But the counter-argument about cassava survives in a narrower and more useful form. The crop itself is not available, because most of the 10.5 million tonne harvest already goes through about 120 industrial starch factories. What is available is the wet pulp those factories leave behind, at those 120 addresses. Brewer's spent grain adds a second concentrated stream that is already 15 to 30 percent protein and is produced by only four brewery groups. See chapter 8.

4. Vietnam's cost advantage is real, and smaller than an unadjusted comparison suggests. Revised in this version. Substituting Vietnamese electricity and labour prices into the one cost model that resolves its structure line by line puts Vietnamese power at around 40 percent, and Vietnamese manufacturing labour at around a tenth, of what the model assumes. Two things bound what that is worth. Utilities are 18 percent of production cost in that model, so a 60 percent cut to the power line is worth roughly 6 to 7 percent of total cost. And a tropical site pays part of it back: an external estimate puts continuous mechanical chilling at USD 0.095 to 0.12 per kilogram of biomass, which against a USD 1.20 to 2.07 product is 5 to 10 percent of cost. The advantage that survives both is labour, which no chiller erodes. This remains an indicative transfer rather than a costed result, and no Vietnam-specific study exists. See chapter 10 and open question F-G2.

5. The first market is shrimp feed, not livestock feed and not human food. Revised in this version. Fishmeal reached USD 1,600 per tonne in late 2025, while soybean meal trades in the hundreds. Microbial protein modelled at USD 1,200 to 2,070 per tonne is in range for the first and not the second. Vietnam's aquafeed market is 3.4 to 4.0 million tonnes a year, but that is not the addressable figure: fishmeal inclusion is concentrated in shrimp and marine finfish, while pangasius is formulated around cheap plant meals and carries very little marine protein to displace. The pool of fishmeal a competing ingredient could actually displace is on the order of 350,000 to 450,000 tonnes a year, and that is the number to quote. There is a second reason to start here, independent of price: this is the market where a regulatory process exists at all, though its duration for a novel ingredient is unknown. See chapter 9.

6. Precision fermentation is further behind than the commercial noise suggests, but the way in has changed. Published food-protein titers run from 35 milligrams to 7.27 grams per litre, against a stated economic threshold above 50 grams per litre, so the best published result is about one seventh of the threshold and the weakest about one fourteen-hundredth. Precision-fermented milk protein costs USD 210 to 310 per kilogram against USD 15 to 25 for the conventional article. What changed is access rather than performance: OPENPichia removed the licensing barrier to the standard production yeast in 2024, the Food Industries Research Institute in Hanoi already runs that same yeast for enzyme work, and ScaleUp Bio in Singapore operates a licensed food-grade facility of up to 10,000 litres that takes third-party clients. The way in is a contract, not a construction project. See chapter 10.

7. On cultivated meat the answer is to watch and not to build, and the reason is sharper than cost. Revised in this version. Achieved cell densities are 10^6 to about 10^7 cells per millilitre against cost models that assume above 10^8. Demonstrated vessels run from tens to low thousands of litres against models assuming 20,000 or more. Growth factors are 90 to 95 percent of the cost of the culture medium. For cultivated seafood, where Vietnam would have a species advantage, there is no food-grade production cell line for catfish, tilapia or salmon. Continuous research lines do exist in public repositories, including thymus and gill lines from striped catfish deposited in India, and a hybrid shrimp line, but all are serum-dependent and unadapted to suspension culture. The advantage is real in theory and the production platform does not exist. See chapter 12.

8. One candidate needs no fermentation infrastructure at all, as feed. Revised in this version. Duckweed grown on the nutrient-rich water leaving pangasius farms, harvested and fed back into aquafeed, has a Vietnamese research base going back three decades, protein yields per hectare above ten tonnes a year against well under one tonne for soybean, and a European novel food dossier precedent scoring PDCAAS 0.89. Every other category in this report is held up by the absence of fermentation tanks. This one needs ponds. The qualification is not a footnote: duckweed grown on farm effluent takes up heavy metals and pathogens from the water and is a feed material only. Food-grade duckweed requires closed cultivation on clean water, which is a different and much more capital-intensive business. See chapter 11.

And the food-security case is stronger than it first appears. Soybean is about 98.5 percent imported by supply. Soybean meal feed consumption of 7.0 million tonnes is met from imported meal plus meal crushed domestically from imported beans, against 41,000 tonnes of domestic soybean production. Essentially all of Vietnam's soy protein is imported, in one form or the other. See chapter 14, claim 3.