Dairy is the worst of the five on greenhouse gases and on land, and it is not close. Below dairy, the rankings stop agreeing with each other. Almond has the lowest emissions of any common plant milk and the second-highest water use. Soy has the best protein and the worst reputation. Oat sits in the middle of every environmental column and near the bottom on protein.
Which impact you mean is most of the work in this comparison. A milk can win on carbon and lose on water by a factor of 13, and a single word covering both decides nothing.
The numbers everyone is arguing over
Our World in Data published a five-milk comparison in January 2022, built on Joseph Poore and Thomas Nemecek’s 2018 meta-analysis in Science of more than 38,000 farms in 119 countries. Its figures describe 2013 production, per liter of milk:
- Greenhouse gases, kg CO2e: dairy 3.15, rice 1.18, soy 0.98, oat 0.90, almond 0.70
- Land use, m2: dairy 8.95, oat 0.76, soy 0.66, almond 0.50, rice 0.34
- Freshwater use, liters: dairy 628, almond 371, rice 270, oat 48, soy 28
- Eutrophication, g PO4e: dairy 10.65, rice 4.69, oat 1.62, almond 1.50, soy 1.06
Read down the columns rather than across them. Dairy leads all four, by between three and 23 times, and that is the one finding the literature does not fight about. Below dairy the order rearranges in every column. Almond is first on emissions and fourth on water. Rice is last on land and second-to-last on emissions. Soy is first on water and eutrophication and third on land.
The World Resources Institute reached the same shape on February 13, 2025, with plant milks averaging “roughly one-third or less of the greenhouse gas emissions of cow’s milk.” It also put almond in the awkward position the per-liter table implies: good for the climate, and using about as much water as dairy.
Almond buys carbon savings with irrigation
Almond’s 371 liters of water per liter of milk is the number that generated the backlash, and it is real. It is also the only column almond loses. On emissions, land, and eutrophication it is at or near the front of the plant milks.
Whether that trade matters depends on where the orchard is, which is a question the global average cannot answer. A per-liter figure averaged across 119 countries treats irrigation drawn from a stressed aquifer the same as irrigation drawn from a river in flood. Water use is the impact category where the location of the farm changes the meaning of the number most, and it is the category almond is judged on.
Soy is the land-efficient one, and the one with the reputation
Per liter, soy uses 0.66 m2 of land. Per gram of protein it is further ahead than that, because it is the only plant milk in the group that carries protein comparable to dairy. Sai Kranthi Vanga and Vijaya Raghavan, writing in the Journal of Food Science and Technology in November 2017, measured 8.71 g of protein in a 240 ml serving of soy milk against 8.11 g for cow’s milk, 1.67 g for almond, and 0.85 g for rice.
The deforestation association is the counterweight, and it is worth being exact about it. Our World in Data’s account of the soy trade, published February 4, 2021, breaks 2018 production down as 77% animal feed, 7% direct human food, and about 4% industrial uses including biodiesel. Poultry alone takes 37% of world soy. Tofu, tempeh, edamame, and soy milk together account for the 7%.
The same analysis finds pasture, not soy, drives most direct Amazon clearance, that soy expansion onto recently deforested Brazilian land fell from 30% to 1% between 2006 and 2014 after the Soy Moratorium, and that beef accounts for 41% of tropical deforestation. Soy is implicated in forest loss. The soy in a soy latte is a small and mostly food-grade fraction of the crop that does the implicating, and blaming the drink for the feed sector is a category error.
Oat is unremarkable in both directions
Oat’s environmental case is that it has no bad column. It is second on land, third on emissions, third on eutrophication, and second-lowest on water at 48 liters per liter. Nothing about it is exceptional and nothing about it is a liability, which is a defensible reason to pick it.
Its nutrition is the weak half. WRI describes oat and coconut as having lower emissions and water use than cow’s milk while being “relatively low in protein.” Romdhane Karoui and Ines Bouaicha, in a review of milk-alternative protein quality published in Frontiers in Nutrition on July 5, 2024, report Digestible Indispensable Amino Acid Scores for oat beverages spanning 0.73 to 1.83 depending on the amino acid assessed, against 1.17 to 1.98 for cow’s milk. Oat protein is not absent. It is less of it, and less complete.
Rice is the one that loses on its own terms
Rice milk has the lowest land use of the five and the worst emissions and eutrophication figures of the four plant milks, which is what growing a flooded cereal does to a footprint. Nutritionally it is the thinnest of the group: 0.85 g of protein and 25.28 g of carbohydrate per 240 ml serving in Vanga and Raghavan’s figures, against 11.5 g of carbohydrate for cow’s milk.
There is also a contaminant issue no other milk in this comparison carries. Rice takes up inorganic arsenic from soil and water more readily than other cereals, and UK guidance published January 9, 2018 advises that children under five should not be given rice milk as a substitute for breast milk, infant formula, or cow’s milk, on the reasoning that young children drink more milk relative to body weight. For adults it is treated as fine within a varied diet.
What dairy is still carrying
The nutritional gap is not only protein. Hallie Lundquist and colleagues, writing in JDS Communications in February 2024, report 85 micrograms of iodine in a 237 ml cup of nonfat cow’s milk — 57% of a pregnant woman’s daily requirement — against an average of 3.1 micrograms per 250 ml across 30 plant-based products. The reason given is not cost but flavor: iodine fortification carries a metallic aftertaste, so most manufacturers skip it.
That is a specific, checkable gap, and it is the sort of thing a per-liter environmental table is structurally unable to show. It also cuts against the tidy version of this argument in both directions: the environmental case against dairy is strong and the nutritional substitution is not automatic.
Where the comparison breaks down
Every figure above is per liter, and a liter of one of these is not a liter of another. Comparing them by volume assumes the drinker wants volume, which is roughly true in a latte and much less true across a diet.
For a coffee drinker the practical finding is narrower and firmer than the general one. In a milk drink, the dairy dominates the footprint, so the choice of milk is the largest lever a person holds over a cup they did not grow. Any of the four plant milks moves that lever a long way. Which one you pick after that is a choice between a water number, a protein number, and an arsenic advisory, and no label on the carton will resolve it for you — the same problem, in a different aisle, as reading a certification logo and assuming it settled the question.
Sources
- Our World in Data — Dairy vs. plant-based milk: what are the environmental impacts?
- World Resources Institute — Environmental Impacts of Dairy and Plant-based Milks
- Our World in Data — Soy
- Journal of Food Science and Technology — How well do plant based alternatives fare nutritionally compared to cow’s milk?
- Frontiers in Nutrition — A review on nutritional quality of animal and plant-based milk alternatives: a focus on protein
- JDS Communications — Cow milk is an important source of iodine for prenatal health, and switching to plant-based milk can lead to iodine insufficiencies
- GOV.UK — Arsenic in rice
