X. Drivers, shocks, signals and wildcards to 2050

What this appendix contains. The evidence behind Drivers to 2050 and Scenarios 2050. Sections X.1 to X.5 give the drivers by domain (trade, disease, energy, demography, climate). Section X.6 sorts them into near-certain trends and critical uncertainties and explains the two scenario axes. Sections X.7 to X.11 are the monitoring registers: 16 signposts, 43 signals, 26 wildcards, 22 external forecasts and 17 past shocks.

How to use it. To see which 2050 world is unfolding, start with X.7. To check a number in chapter 15, use X.1 to X.5. Status is as of 24 September 2026 unless stated.

---

#X.1 Trade and geopolitics

#X.1.1 US tariffs on Vietnam, 2025 to 2026

A commercial website says 46% applied until February 2026; the joint statement says 20%. Position: 20%. The legal basis changed three times, so exporters should expect tariff volatility to about 2030 VN-directMedium signal.

#X.1.2 Purchase commitments

Vietnamese firms signed 20 memorandums to buy "over $2.9 billion" of US farm goods ² VN-directHigh; the United States already supplied 43.6% of Vietnam's soybeans and 72% of its DDGS in 2025 ^7,8 VN-directMedium. China committed to buy at least 25 Mt of US soybeans a year in 2026 to 2028 (compliance not verified) ⁹ generalHigh projection (official commitment), which pushes Brazilian soy towards other buyers, including Vietnam. In 2026 to 2028, deals as well as prices will set Vietnam's origins generalMedium signal.

#X.1.3 China's soy demand and feed targets

China as swing buyer (our calculation). Reaching 10% by 2030 from 13.4% cuts about 3 to 3.4 points of roughly 300 Mt of feed: about 9 to 10 Mt of meal (12 to 13 Mt of soybeans) a year, more than Vietnam's whole 2025 use of 7.2 Mt ¹⁵ generalLow estimate. Position: report 13 to 16% for the current share; do not assume the 2030 target is met.

#X.1.4 Exporter baseline to 2035

OECD-FAO Agricultural Outlook 2026 to 2035 baseline ¹⁶ generalMedium projection (OECD-FAO baseline).

For Vietnam, whose protein-meal imports reach 10.3 Mt in 2035 in the same baseline ¹⁷, the risk to 2035 is volatility and concentration of origin, not a rising price trend VN-directMedium projection. Brazil's flat soy exports may be conservative; we report them as published. We found no soy or maize price path beyond 2035 ¹⁸ generalHigh.

#X.1.5 Argentina's export duties

Decree 423/2026 (3 June 2026) ¹⁹ generalMedium projection (official schedule).

Argentina is Vietnam's main soybean-meal origin and supplied 46.5% of its maize in 2025 (Brazil 34.8%) ²⁰ VN-directMedium. We expect the cuts to push Vietnamese feed costs down to 2028 (our reading) VN-adjacentMedium estimate. Duties move both ways: meal duties rose from 31% to 33% in March 2022 ²¹ generalLow.

#X.1.6 EU rules

• EUDR applies from 30 December 2026 (large and medium operators) and 30 June 2027 (small firms) to soy entering the EU, not to meat or fish from soy-fed animals (our reading) ^22,23 generalHigh. Uncertified soy sold to Asia could be slightly cheaper in 2027 to 2030; no study measures this generalLow estimate.
• CBAM covers cement, steel, aluminium, fertilisers, electricity and hydrogen; the proposed extension adds 180 to 457 metal-intensive goods, reviewed yearly from 2028, and no food or feed ^24,25 generalHigh projection. The 2028 review is the signpost (WC-15).

#X.1.7 Chokepoints, export restrictions and the 2026 Middle East shock

#X.1.8 Import duties

Soybean meal and feed maize carry 0% MFN duty since 31 March 2025 ³³ VN-directMedium. Food-protein ingredients carry 3% (HS 3504), 5% (HS 2106.10) and 7 to 10% (yeasts and dead single-cell organisms, HS 2102) ^34,35 VN-directMedium. Preferential rates for Chinese microbial protein under ACFTA or RCEP are unverified (X.12).

#X.2 Disease

#X.2.1 African swine fever

Three live attenuated vaccines are licensed and nearly 6 million doses were distributed by June 2024, but uptake is "limited" ^37,41 VN-directMedium. Genotype II vaccine strains were "unable to completely protect" pigs against the recombinant strain ⁴² VN-directMedium; reversion, shedding and sow safety remain concerns ⁴³ VN-directMedium. Official promotion implies control; peer-reviewed work shows partial protection. Position: assume a major wave every two to three years to 2035, each cutting feed demand for a few quarters and speeding consolidation VN-directMedium trend.

#X.2.2 Zoonoses and H5N1

Past H5N1 shocks (Vietnam 2003 to 2004, US dairy cattle from 2024, Vietnamese zoo tigers in 2024) are in X.11.

Vietnam holds about 585 million poultry and 31 million pigs ⁴⁰ VN-directMedium. A mammal-adapted H5N1 lineage spreading in pigs or cattle in 2030 to 2050 has no published probability (WC-02) VN-directLow wildcard.

#X.2.3 The antibiotic ban and the additive market

• Growth promoters were banned in feed from 2018; from 1 January 2026 antibiotics may only treat diagnosed disease ⁴⁷ VN-directMedium. About 71.7% of antimicrobials used in Vietnam in 2015 went to animals ⁴⁸ VN-directMedium.
• World use in food animals is projected to rise 8%, from 99,502 t (2020) to 107,472 t (2030), with 67% of hotspots in Asia ⁴⁹ generalHigh projection (published model).
• The livestock strategy encourages "biological products that replace antibiotics" ⁵⁰ VN-directMedium. We found no 2030 target for antimicrobial use in animals.
• Reading. The ban opens a 2026 to 2030 market for fermentation-made probiotics, yeast products and enzymes, the lowest-regulatory-risk use of Vietnamese fermentation capacity VN-directLow estimate. Its effect is not yet measured.

#X.3 Energy

#X.3.1 Plans

Revised Power Development Plan 8 (Decision 768/QD-TTg, April 2025) and hydrogen strategy (Decision 165/QD-TTg, February 2024). All rows projection (official targets).

In 2023, coal gave 45% of 274 TWh, hydro 29%, solar and gas 10% each, wind 4% ⁵⁵ VN-directMedium; the grid factor was 0.6592 t CO2 per MWh ⁵⁶ VN-directHigh.

#X.3.2 Plans versus delivery

The IEA lists a 240 MW Vietnamese electrolyser as under construction ⁶², but it does not match the stalled plant. Position: treat Vietnamese green hydrogen before 2030 as unproven VN-directLow.

#X.3.3 Power prices

Reading (our judgement). To about 2035, grid prices follow imported coal, LNG and the dollar; the hedge is a direct renewable contract. If growth runs nearer 5.7% than 10%, planned capacity could run ahead of demand in the 2030s and create cheap off-peak power; this is unquantified VN-directLow estimate.

#X.4 Demography, income and labour

#X.4.1 Ageing and urbanisation

UN World Population Prospects 2024 (medium variant) and World Urbanization Prospects; shares are our calculation from age groups ^68,69 VN-directMedium projection (UN projection).

The 65-and-over share crosses 14% (an "aged society") around 2036, and population peaks around 2049 ^68,70 VN-directMedium projection. Headcount adds only about 8% from 2025 to 2050, so income drives protein demand (Protein balance 2050). The 2025 provincial merger and new UN urban definitions could change urban shares with no change on the ground.

#X.4.2 Growth paths: official and others

Real GDP per person, constant 2025 USD (our calculation) VN-directLow estimate.

Steps. GDP per person in year t = 5,026 x (1 + g)^n / (population in year t / 101.6 million). Base USD 5,026 (2025, edition 1.0 macro_indicators.csv); UN population as above. Example, official path 2030: 5,026 x 1.611 / 1.026 = about 7,900. Records MDR-040 to MDR-048 in macro_drivers_2050.csv.

• The paths exclude price and exchange-rate effects, so they do not compare with the nominal official aims of about USD 8,500 per person in 2030 ⁷¹ and about USD 38,000 by 2050 ^75,76 VN-directHigh projection (official targets).
• Disagreement. Official 10% against USDA's 5.7%. Position: 10% is a target; the USDA-type path is the central case, as in the balance model. On it, Vietnam becomes an aged society in about 2036 at about USD 8,700 per person: old before rich (our calculation) VN-directLow estimate.

#X.4.3 Farm labour and smallholder exit

2025 Rural and Agricultural Census (official results, April 2026) ⁷⁷ VN-directHigh.

• Backyard farms once supplied about 80% of pork; small household farms now supply 35 to 40% ³⁷ VN-directMedium. Household livestock falls 5 to 7% a year; in 2022 to 2023 foreign-invested firms held 43% of the pig market, households 38%, domestic firms 19% ⁷⁸ VN-directMedium trend.
• Extrapolation (our calculation). Farm workers fell 1.75% a year in 2016 to 2025 (19.34 to 16.5 million). Continued: 16.5 x 0.9825^10 = 13.8 million (2035) and 16.5 x 0.9825^25 = 10.6 million (2050) VN-directLow estimate. A mechanical trend, not a forecast.

#X.4.4 Just transition: feedstock incomes, not factory jobs

• Small pig and poultry households are already exiting (X.2, X.4.3). A shift on the S-ALT scale (10% of meat demand replaced by 2050) would add little to that pressure VN-directLow estimate.
• Fermentation creates few jobs: labour is 7% (biomass) to 17% (precision fermentation) of cost in regional plant models ⁷⁹ VN-adjacentMedium.
• The income at stake is feedstock. If all S-ALT glucose came from cassava, 2050 would need about 5.37 Mt of fresh roots on about 262,000 ha, mostly from smallholders (balance_outputs.csv, BLO-1130, BLO-1131) VN-directLow estimate. Contract terms for cassava, rice by-products and brewery residues decide who gains; about 58.8 million people will still be rural in 2050 ⁶⁹ VN-directMedium projection.

#X.5 Climate

#X.5.1 Agricultural and methane caps

All targets projection (official targets, or a published recommendation where stated).

Numbers behind the caps (our calculation) VN-directLow estimate.

• Disagreement. The inventory gives livestock 18.5 MtCO2e in 2016 ⁸⁸; GIZ gives 27.2 MtCO2e in 2014 including manure N2O ⁸⁷. Position: inventory for shares, GIZ only for growth rates.
• Pork is about 60% of meat and pigs emit little enteric methane, so the 2030 livestock cut must come mostly from manure VN-directLow estimate. Rice holds 30.7 of the 45.9 MtCO2e agricultural methane cap ⁸² VN-directHigh.

#X.5.2 NDC 3.0 status

NDC 3.0 (2026 to 2035) was in draft in January 2026, with approval requested for Q1 2026. A reported draft caps 2035 energy emissions at 380.2 to 519.1 MtCO2e; agriculture pathways were "to be updated". We found no 2035 agriculture cap and no Vietnamese NDC 3.0 in the registry pages we could read ^89,90,91,92 VN-directMedium signal. A media claim that agriculture already beat its unconditional target is unverified ⁹³ VN-directLow.

#X.5.3 Mekong subsidence and salinity

• Our calculation. At about 0.9 cm a year (slower-growth rate around 2023), pumping lowers the delta about 28 cm from 2018 to 2050 (0.9 x 32), comparable to or more than global sea-level rise VN-directMedium estimate. The low south-west, now Ca Mau and An Giang (former Kien Giang), falls below sea level this century in every pathway ⁹⁶ VN-directMedium projection.
• Disagreement. Media blame climate for salinity; the modelling finds human drivers dominate to 2050 ⁹⁸. Position: human drivers to 2050, sea-level rise after; official flood figures are a lower bound. The worst event (2019 to 2020) is SHK-015 in X.11.
• The Mekong plan moves land from rice to aquaculture and restricts groundwater ¹⁰⁰ VN-directHigh projection, and farmers are already switching to shrimp ¹⁰¹. This raises aquafeed protein demand and cuts rice by-products in 2030 to 2050 VN-directMedium estimate.

#X.5.4 Humid heat and fermentation cooling

Heat-stress days rose 0.8 to 4.2 per decade in 1979 to 2018 ¹⁰² VN-directMedium. Annual maximum WBGT (a heat-stress index) rises 0.2 to 0.4 °C per decade across SSP1-2.6 to SSP5-8.5; the delta gains under 60 extra days a year above WBGT 32 °C in 2021 to 2040 and over 150 by 2081 to 2100 under SSP5-8.5 ¹⁰³ VN-directMedium projection. Edition 1.0's cooling penalty (tower water at 31 to 33 °C is too warm for a 28 to 32 °C broth) is USD 70 to 180 per t of dry biomass ^104,105 VN-directLow.

Change to 2050 (our calculation) VN-directLow estimate.

Warming adds only 1 to 4% to the cooling penalty by 2050; a carbon price on grid power adds far more. By 2050 a 37 °C process needs trim chillers more often, while a 40 °C process stays mostly on tower water VN-directLow estimate.

#X.5.5 Exporter climate risk, El Niño and anchoveta

• Disagreement. CMIP6 spread and an observed eastern-Pacific cooling make the El Niño doubling uncertain ⁹⁷. Position: more fishmeal volatility is plausible, not certain; no probability.
• None of these sources gives a price path. The common message is volatility: a feed-ingredient import bill near USD 10 billion a year will face larger swings in 2040 to 2050, not only a higher average generalMedium estimate.

#X.5.6 Domestic crops, aquaculture and livestock

The national rice figure conflicts with crop-model gains for parts of Asia ¹⁰⁷. Position: an official stress case, not a central estimate. Closed, cooled pig and poultry houses will carry growth to 2050, raising power demand and favouring integrators VN-directLow estimate. We found no dated SSP projection of Vietnamese shrimp or pangasius output.

#X.5.7 Macro costs and the national climate scenario

• Climate change could cost Vietnam 12 to 14.5% of GDP a year by 2050, against 3.2% in 2020 ^85,119 VN-directMedium projection (World Bank model).
• The official scenario in use is the 2020 RCP edition: mid-century warming of 1.2 to 1.7 °C (RCP4.5) and 1.7 to 2.3 °C (RCP8.5) ⁸⁴ VN-directHigh projection.
• Its replacement (fifth edition, SSP-based, 4.4 km resolution for the two largest cities, district sea-level rise) was due in Q1 2026 ^120,121,122 VN-directMedium. We found no evidence of publication by September 2026 signal. One preliminary figure: plus 2 to 3.7 °C by 2100 under SSP3-7.0 ¹²³ VN-directLow projection.

#X.6 Near-certain trends versus critical uncertainties

Scores are our judgement, 1 (low) to 5 (high), for impact on Vietnam's alternative-protein supply and uncertainty to 2050 (chart data: charts/data/drivers_matrix_2050.csv) VN-directLow estimate.

Drivers to 2050: impact and uncertainty

Our judgement scores (1 low to 5 high) for impact on Vietnam's alternative-protein supply and for uncertainty to 2050

#Why these two axes

Scenario axes should be high in impact, outside the reader's control and largely independent of each other ^131,132 generalHigh. Our tests of each critical uncertainty (our judgement) VN-directLow estimate:

Vietnam's own choices (food law, feed lists, public money, a protein statistic) are the moves tested in Robust moves, not axes.

#X.7 The 2050 signposts

All 16 rows of signposts_2050.csv. A tripped signpost shows which way the evidence moves; it does not predict a world VN-directLow signal.

Reading, September 2026. Three chronic-stress signposts (SP-01 to SP-03) show a 2026 peak, but none has held long enough, while SP-06 and SP-07 point to ample soy to 2028. Chinese capacity (SP-09) grows, but so do public plants elsewhere (SP-13), and no ASEAN mechanism exists (SP-12). The evidence points both ways on both axes VN-directLow signal.

#X.8 Signals register

All 43 rows of signals_register.csv, short form, grouped by signal type. Every row is signal (2024 to 2026). Direction is for alternative protein; relevance (High, Medium, Low) is to Vietnam.

#Science and patents

#Plants, start-ups and finance

#Policy

#Shocks

#X.9 Wildcards

All 26 rows of wildcards.csv. No probabilities unless a named source published one. "Jumps to" marks wildcards that act on a scenario axis and could move Vietnam between the 2050 worlds within a year: the six in Scenarios 2050 (WC-01, 06, 08, 09, 13, 25) plus four we add (WC-05, 07, 18, 24); the additions are our judgement VN-directLow estimate.

WC-01 moved in 2026 (X.5.5). WC-16 is a live risk for the 2027 dry season, because the 2016 and 2020 salinity records followed strong El Niño years ¹⁸⁵ VN-directLow signal. WC-03 is live because recombinant strains escape the vaccines ⁴³ VN-directMedium signal.

#X.10 External forecasts and their track record

All 22 rows of external_forecasts.csv, as published by the named forecaster projection. EXF-001 to EXF-008 come from one Rethink Priorities panel (five top Metaculus forecasters and one scientist, 2021, updated 2024) ¹⁸⁶ generalMedium.

The cultivated-meat forecast record.

• Consultancy forecasts ran far too high. McKinsey's 2030 path needs 220 to 440 million L of capacity; in 2026 the largest run is 20,000 L and nothing sells at commercial scale ^160,139 generalMedium.
• Past timelines were optimistic: "very systematic overconfidence" ¹⁸⁸ generalMedium. A fetched summary said about 3% were right; position: use the source's counts (11%).
• The only probabilistic forecast is sober: a 54% chance that world output stays below 100,000 t a year through 2051 ¹⁸⁶ generalMedium projection.
• EAT-Lancet 2019 and 2025 figures must not be mixed. We use these forecasts as benchmarks, never as inputs.

#X.11 Shock register

All 17 rows of shock_register.csv, with the lesson we draw (our judgement).

Lessons. Vietnam's protein supply met a large shock every one to two years in 2019 to 2026, so plan for recurring shocks, not a smooth trend, to 2035 VN-directMedium trend. Each shock moved output from households to integrators, the buyers best able to specify new ingredients. Energy and freight mattered as much as grain in 2026. Policy shocks move both ways. Adaptation works: 1,580 ha of rice at risk in 2024 against 26,000 ha lost in 2020 ^185,206 VN-directMedium.

#X.12 Gaps and open questions

These feed Q. Open questions.

Source issues. In signposts_2050.csv, SP-03 cites HSC-30 for the NOAA odds, which come from HSC-28 (we cite both), and SP-10 cites IND-01, which does not cover Chinese imports. WC-21, WC-23 and WC-26 have no source.

#X.13 Data files

• macro_drivers_2050.csv (92 rows): demography, income, trade, disease, energy and workforce drivers to 2050.
• climate_impacts_2050.csv (34 rows): climate effects on Vietnam, the delta, heat, cooling, exporters and fishmeal.
• emissions_targets.csv (19 rows): climate targets and caps, 2030 to 2050.
• vn_energy_inputs_2050.csv (50 rows): grid factor, power prices, power plan and hydrogen inputs.
• signposts_2050.csv (16 rows): 2050 signposts with thresholds and September 2026 status.
• scenarios_2050.csv (4 rows): the four 2050 worlds.
• signals_register.csv (43 rows): signals, 2024 to 2026.
• wildcards.csv (26 rows): wildcards with direction, horizon and signposts.
• external_forecasts.csv (22 rows): published forecasts on alternative protein and inputs.
• shock_register.csv (17 rows): past shocks to protein supply, 2003 to 2026.

Related: Drivers to 2050, Scenarios 2050, Robust moves, U. Futures method, O. Outlook data, Y. Targets and hubs, Z. Aquafeed and feedstocks 2050, Q. Open questions, R. Disagreements.