Chapter 22Part IV: What could 2035 and 2050 look like?
22. The protein balance to 2050: what people will eat, and the imports behind it
On this page
In brief
- On trend, Vietnamese people eat more meat protein each year: about 1.01 Mt in 2025 and 1.45 Mt in 2050 (our model). Most of the rise comes from each person eating more.
- Most of that meat is produced at home on imported feed. The soybean meal behind it rises from 7.2 Mt in 2025 to about 10.4 Mt in 2050, almost all imported.
- About 70% to 78% of the meat protein people eat rested on imports in 2025, directly or through feed. On trend this rises to 79% to 87% by 2050 (our estimate).
- Replacing 1 t of meat protein avoids about 4.5 t of soybean meal and 9.5 t of maize imports (our calculation). But a plant food made from imported soybeans stays import-dependent.
- At the scale demand supports, new protein displaces about 0.4% of meat protein by 2050. It adds a second source; it does not replace meat.
22.1 What people will eat: more meat protein per person
Most of the growth comes from each person eating more. Population grows about 8%, to about 110 million in 20501Evidence: evidence about Vietnam (VN-direct); medium confidence: a reputable secondary source, or a primary source read in part. Forward-looking: projection, a published model result or an official target.. Meat demand per person rises about 32% on trendEvidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions.. These are supply figures, not measured intake (Protein diet).
What people will eat on trend (scenario S-BASE)Evidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions.
| Indicator | 2025 | 2035 | 2050 |
|---|---|---|---|
| Population, million | 101.6 | 106.5 | 110.0 |
| Meat per person, kg carcass weight | 66.5 | 81.9 | 87.6 |
| Meat protein per person, g a day | 27.3 | 33.7 | 36.0 |
| Meat protein, all people, kt | 1,013 | 1,309 | 1,446 |
| Share of meat protein resting on imports | 70% to 78% | 76% to 84% | 79% to 87% |
| Eggs per person a year | 210.6 | 247.5 | 275.1 |
| Farmed fish and seafood eaten at home, kg per person | 29.3 | 33.5 | 37.6 |
| Meat produced in Vietnam, Mt | 6.0 | 7.8 | 8.5 |
Most of that meat protein rests on imports, and the share rises. Counting imported meat plus the imported feed behind domestic meat, the share reaches 79% to 87% by 2050 (our estimate)2,3,4Evidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions.. It rises because more pigs move to factory feed and poultry, which relies most on it, grows fastest. This is an estimate, not a measured flow (Why Vietnam)5Evidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs..
79% to 87%
Readings.
- Demand is already steep. The OECD-FAO outlook puts Vietnamese meat use at 61.9 kg per person in 2035, above Korea and China2Evidence: evidence about Vietnam (VN-direct); medium confidence: a reputable secondary source, or a primary source read in part. Forward-looking: projection, a published model result or an official target..
- Our slowdown after 2035 is a judgement. The official growth path gives 101 kg per person by 2050, a stress case.
- Part of the meat is imported. About 90% of the meat supply was domestic in 20235Evidence: evidence about Vietnam (VN-direct); high confidence: a primary source was read.. The model keeps today's import shares for each meat, the more import-exposed case2Evidence: evidence about Vietnam (VN-direct); medium confidence: a reputable secondary source, or a primary source read in part..
22.2 Published numbers stop at 2035, so we use four scenarios
| Series | 2025 | 2030 | 2035 | Type | Sources |
|---|---|---|---|---|---|
| Industrial livestock feed, Mt | 24 to 25 | 30 to 32 | none | Official target | 6,7Evidence: evidence about Vietnam (VN-direct); medium confidence: a reputable secondary source, or a primary source read in part. |
| Meat output, Mt | about 6.0 | 6.0 to 6.5 | none | Official target | 8,7Evidence: evidence about Vietnam (VN-direct); high confidence: a primary source was read. |
| Protein-meal imports, Mt | 7.4 | 8.6 | 10.3 | OECD-FAO projection | 2Evidence: evidence about Vietnam (VN-direct); medium confidence: a reputable secondary source, or a primary source read in part. |
| Maize imports, Mt | 12.0 | 14.0 | 16.6 | OECD-FAO projection | 2Evidence: evidence about Vietnam (VN-direct); medium confidence: a reputable secondary source, or a primary source read in part. |
All forward values:Forward-looking: projection, a published model result or an official target.. We found no published projection to 20502,9,10Evidence: evidence about Vietnam (VN-direct); medium confidence: a reputable secondary source, or a primary source read in part.. The 2030 feed target appears to cover livestock feed only, so it now looks out of reach6,11Evidence: evidence about Vietnam (VN-direct); medium confidence: a reputable secondary source, or a primary source read in part. Forward-looking: estimate, our own calculation from stated assumptions..
The four scenarios.
| Scenario | Diets | Feed efficiency | Alternative protein |
|---|---|---|---|
| S-BASE (trend) | Meat per person rises from 66.5 kg to 87.6 kg by 2050 | Slow gains | None |
| S-HIGH (official growth path) | 101 kg per person by 2050 | As S-BASE | None |
| S-EFF (efficiency) | As S-BASE | Faster gains; less soybean meal per tonne of feed | None |
| S-ALT (alternative protein) | As S-BASE, with 1% (2030), 5% (2040) and 10% (2050) of meat replaced by plant-based and fermented foods | As S-BASE | Microbial protein also replaces part of soybean meal and fishmeal in feed |
S-ALT's shares are illustrative what-if levels, not targetsEvidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs.. Its feed side is context only.
22.3 The imported feed behind it
On trend, the soybean meal behind the meat people eat rises 44% by 2050, almost all imported.
Soybean meal import need, MtEvidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions.
| Scenario | 2025 | 2030 | 2035 | 2040 | 2050 | 2050 against 2025 |
|---|---|---|---|---|---|---|
| S-BASE | 7.20 | 8.46 | 9.47 | 9.95 | 10.40 | +44% |
| S-HIGH | 7.20 | 8.98 | 10.11 | 10.92 | 11.72 | +63% |
| S-EFF | 7.20 | 7.81 | 8.08 | 7.85 | 7.00 | minus 3% |
| S-ALT | 7.20 | 8.32 | 9.01 | 9.15 | 8.78 | +22% |
10.4 Mt (+44% on 2025)
7.0 to 11.7 Mt
Context and sources
S-EFF 7.0, S-ALT 8.8, S-BASE 10.4, S-HIGH 11.7 Mt. Only the efficiency scenario keeps soybean meal import need at or below the 2025 level of 7.2 Mt. Our estimate from the balance model; QNT-01 is used only to calibrate it to 2035.2
Other feed resultsEvidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions.
| Indicator | 2025 | S-BASE 2050 | Range in 2050 |
|---|---|---|---|
| Factory-made feed, Mt | 28.5 | 40.9 | 35.4 to 46.2 |
| Maize import need, Mt | 11.9 | 13.8 | 11.6 to 16.3 |
| Farmland abroad for imported soy and maize, million ha | 4.0 | 4.7 | 3.4 to 5.4 |
about 4.0 million ha (2025); 4.7 million ha (2050, trend)
Readings.
- The near term checks out. The model's 2027 values match US Department of Agriculture forecasts within 2% to 5%3,13,12,2Evidence: evidence about Vietnam (VN-direct); medium confidence: a reputable secondary source, or a primary source read in part..
- Vietnam already farms about 4 million hectares abroad for feed, about five times its own maize area (our calculation)Evidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions..
- Domestic soybean is no lever for feed, but could be one for soy foods. Even a large soybean crop would cover only about 6% of feed needs (our calculation)14,15Evidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions.. But Vietnam's food soybeans, the beans behind tofu and soy milk, are only about 540,000 t a year12Evidence: evidence about Vietnam (VN-direct); medium confidence: a reputable secondary source, or a primary source read in part.. Returning to the 2010 soybean area would cover 76% to 95% of that, at 1.4 to 2.3 times the price of imported beans (our calculation)14,16,17Evidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs.. This is candidate play T12, not ranked until tested (Plays, section 26.8).
Feed efficiency moves the result most. Better feed conversion and formulation (S-EFF) would cut 2050 soybean meal need by about 3.4 Mt, twice what S-ALT wouldEvidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions.. That lever sits outside this study's focus on protein for people.
22.4 What protein for people would change
Protein for people saves the most imports per tonne. In 2050, replacing 1 t of meat protein avoids about 4.5 t of soybean meal and 9.5 t of maize importsEvidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions..
A plant food made from imported soy cuts the tonnage but keeps the dependence. A wholly plant food made from imported soybeans uses about 3.3 t of beans per tonne of protein. That is about a fifth of the imported soy and maize behind the meat it replaces (our calculation)Evidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions.. So it cuts imported tonnage. But it stays import-dependent unless the beans or the protein are Vietnamese: 97.5% of the soybean supply is imported5Evidence: evidence about Vietnam (VN-direct); high confidence: a primary source was read..
But the tonnes that demand supports are small. No market has shown modern plant-based or fermented products displacing a measurable share of meat (Demand sizing).
Meat protein displaced: S-ALT against the demand pathsEvidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions. Demand evidence: inferred, our inference or calculation.
| Meat protein displaced | 2030 | 2035 | 2050 |
|---|---|---|---|
| S-ALT assumption | 11.8 kt (1%) | 39.3 kt (3%) | 144.6 kt (10%) |
| Benchmark demand path | 0.4 kt (0.04%) | 2.4 kt (0.18%) | 6.0 kt (0.41%) |
| Stretch demand path | 3.1 kt (0.27%) | 9.2 kt (0.71%) | 21.9 kt (1.52%) |
| Existing chay days (already in the baseline) | about 29 kt (2.5%) | about 32 kt (2.5%) | about 36 kt (2.5%) |
Read S-ALT's food shares as a what-if, not a demand path. Existing chay (vegetarian) days already avoid a similar share (Chay baseline).
So new protein adds a second source; it does not replace meat. On the benchmark path, new routes deliver about 19,000 t of protein a year by 2035 and 67,000 t by 2050, mostly by replacing imported ingredients (Demand sizing)Evidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions. Demand evidence: inferred, our inference or calculation.. On its own, that does not change national exposure to imported feed.
The import saving is about 0.1% in 2035. Displacing 2.4 kt of meat protein avoids about 11 kt of soybean meal (our calculation)Evidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions.. Tofu and soy milk already carry nearly ten times more protein than the benchmark path delivers in 2035, on the same imported beans5Evidence: evidence about Vietnam (VN-direct); medium confidence: a reputable secondary source, or a primary source read in part.. For security, what matters is the share of food protein made from domestic inputs (Why Vietnam).
S-ALT's feed side is context only. Most of its cut in soybean meal comes from microbial protein in feed. Together, its food and feed sides would need sugar equal to about half of Vietnam's cassava harvest, competing with starch exports and fuel ethanol18,19,20Evidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions..
22.5 Reality check: what other countries have achieved
S-ALT sits above every real-world comparison we found; S-EFF sits at the low end of what China has doneEvidence: global evidence, not specific to Vietnam (general); medium confidence: a reputable secondary source, or a primary source read in part..
- Feed recipes change faster than land use. China cut the soybean meal share of feed by up to 0.45 points a year from 2017 to 2025. Norway cut fish ingredients in salmon feed from 90% to about 30% over two decades21,22,23Evidence: global evidence, not specific to Vietnam (general); medium confidence: a reputable secondary source, or a primary source read in part. Forward-looking: trend, observed change expected to continue..
- Self-sufficiency targets are usually missed, then cut. Japan's feed self-sufficiency was 26% in 2000 and 26% in 2024, and its 2030 target was cut24,25Evidence: global evidence, not specific to Vietnam (general); high confidence: a primary source was read. Forward-looking: trend, observed change expected to continue..
- Novel feed ingredients stay tiny. They were 0.4% of Norwegian salmon feed in 202026,27Evidence: global evidence, not specific to Vietnam (general); high confidence: a primary source was read.. S-ALT's 8% microbial share by 2050 is unprecedentedEvidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions..
- Even the EU struggles. To cut soy imports 20% by 2035 with flat demand, an EU analysis needs four levers at their limits, with diet change carrying 40% of the cut28,29Evidence: global evidence, not specific to Vietnam (general); medium confidence: a reputable secondary source, or a primary source read in part. Forward-looking: projection, a published model result or an official target.. On Vietnam's growing numbers, the same target is far out of reach (our calculation)29,12Evidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions..
On this evidence, a credible ambitious range for Vietnam's soybean meal need in 2050 is about 7 to 9 Mt, against 10.4 Mt on trendEvidence: evidence about Vietnam (VN-direct); low confidence: a single claim, press, company marketing or our derivation from weak inputs. Forward-looking: estimate, our own calculation from stated assumptions.. The study sets no goal; the range is context.
22.6 What this means
- Policy makers: publish a national protein balance each year, showing the protein people eat by food, source and origin, with the imported feed behind it (Moves for any 2050). The 2030 feed target needs a clear scope and a statistic to be judged.
- Investors: plan on the demand that exists: about 67,000 t of protein a year by 2050 on the benchmark path, mostly ingredients that replace imports.
- Manufacturers: demand for meat keeps rising, and new protein grows beside it as ingredients, extenders and plant drinks (Demand sizing).
- Startups: size plants against the demand for ingredients that replace imports, not against S-ALT (Demand to frontier).
- Researchers and international bodies: the gaps that matter most are measured inputs, such as how many people keep chay days, and Vietnamese data on feed conversion and soybean meal use by species.
Related: Demand sizing, Outlook.
Sources cited on this page
Numbers in the text link here. Every source is also listed in the full source list. Full source list
- GT-12Population, including UN projections
- QNT-01OECD and FAO. OECD-FAO Agricultural Outlook 2026-2035, data via the OECD SDMX API, dataflow OECD.TAD.ATM DSD_AGR@DF_OUTLOOK_2026_2035 version 1.1. June 2026
- MAC-01USDA FAS GAIN. Vietnam: Grain and Feed Annual, report VM2026-0012. 15 Apr 2026
- FS-23USDA FAS. Vietnam: Grain and Feed Annual. April 2026
- DIE-07FAO. FAOSTAT Food Balance Sheets. 2025-10-14
- NTS-23Prime Minister. Quyết định 1625/QĐ-TTg, Đề án Phát triển công nghiệp chế biến thức ăn chăn nuôi đến năm 2030. 15 Dec 2023
- REG-54Prime Minister. Quyết định 1520/QĐ-TTg 2020 chiến lược phát triển chăn nuôi 2021 to 2030. 6 Oct 2020
- MAC-23Prime Minister of Vietnam. Decision 1520/QD-TTg approving the Livestock Development Strategy 2021-2030, vision 2045. 6 Oct 2020
- QNT-02USDA Office of the Chief Economist. USDA Agricultural Projections to 2035, OCE-2026-1. 13 February 2026
- QNT-15ABARES. Food demand to 2050: opportunities for Australian agriculture. 2012
- MAC-28VietDVM. San luong thuc an chan nuoi nam 2024 dat 21,5 trieu tan, tang 3,4% so voi nam 2023. 2024
- MAC-04USDA FAS GAIN. Vietnam: Oilseeds and Products Annual, report VM2026-0006 (Nguyen Linh). 15 Apr 2026
- MAC-03USDA FAS GAIN. Vietnam: Grain and Feed Quarterly, report VM2026-0030. 4 Aug 2026
- AQF-20VnEconomy. Ngành đậu tương có nhiều 'nút thắt' cần tháo gỡ. 7 Nov 2024
- AQF-02OECD and FAO. OECD-FAO Agricultural Outlook 2026-2035, SDMX API, dataflow OECD.TAD.ATM DSD_AGR@DF_OUTLOOK_2026_2035 version 1.1. June 2026
- AIB-06QNS. Báo cáo thường niên 2025. 2026-03
- SEC-05United Nations Statistics Division (UN Comtrade). Viet Nam imports of protein-relevant HS lines by partner, 2023. 2026-10-01
- FS-01Dân Việt. Việt Nam đang là nhà cung cấp lớn nhất một thứ tinh bột cho Trung Quốc, năm 2025 bán được gần 3,7 triệu tấn. January 2026
- FS-31Doanh nghiệp Hội nhập. Ngành mía đường sản xuất gần 1,3 triệu tấn, áp lực tiêu thụ vẫn đè nặng. 16 July 2026
- GT-10MOIT. Thông tư 50/2025/TT-BCT quy định lộ trình áp dụng tỷ lệ phối trộn nhiên liệu sinh học với nhiên liệu truyền thống tại Việt Nam. 7 Nov 2025
- GEO-10farmdoc daily (University of Illinois). Can China Reduce Soybean Import Demand? Evaluating Soybean Meal Reduction Efforts. December 2025
- GEO-11South China Morning Post. China failed to hit 2025 soybean meal dependence target. US deal makes it harder. 9 Feb 2026
- VIS-19Ytrestøyl T., Aas T.S., Åsgård T.. Utilisation of feed resources in production of Atlantic salmon (Salmo salar) in Norway. 2015
- VIS-11MAFF Japan. Summary of the Basic Plan for Food, Agriculture and Rural Areas. March 2020
- VIS-12MAFF Japan. 飼料をめぐる情勢(データ版). August 2026
- VIS-18Aas T.S., Åsgård T., Ytrestøyl T.. Utilization of feed resources in the production of Atlantic salmon (Salmo salar) in Norway: An update for 2020. 2022
- VIS-20Nofima. Salmon feed is slowly changing. 19 Apr 2022
- PPE-03The Protein Project. From dependency to diversity. 2026-06
- PPE-04The Protein Project. From dependency to diversity: Technical Appendix. 2026-06