The carbon impact of plant-based drinks compared to cow's milk

The short version

A litre of cow’s milk emits roughly 3.15 kg CO₂ equivalent according to the largest meta-analysis of food impacts available. Plant drinks land between 0.70 and 1.18 kg. That is a factor of about three, not the factor of five or ten that gets repeated online.

No plant drink wins on every measure. Almond has the lowest greenhouse gas emissions and the highest freshwater use of the alternatives. Rice emits the most of the five plant options and still sits well below dairy. Soy is the best all-round compromise and the only one whose protein content comes close to milk.

Two things matter far more than where the product was made: which product you buy, and whether it has been fortified. Transport accounts for roughly 5% of global food system emissions. In a German market survey, around 80% of plant drinks were not fortified at all, and fewer than 10% contained iodine.

Two safety rules that have nothing to do with climate: no plant drink is suitable as a main drink before age 1, and rice drinks should not be given to children under 5 because of inorganic arsenic.

What a carbon footprint actually measures

A food’s carbon footprint sums the greenhouse gases emitted across its life cycle, expressed as CO₂ equivalent. Three gases dominate in food: carbon dioxide, methane from ruminant digestion and flooded rice paddies, and nitrous oxide released from fertilised soils.

CO₂ equivalent is not a neutral yardstick. Each gas is weighted by its global warming potential, usually calculated over a hundred years. On that horizon methane counts roughly thirty times more than CO₂. Over twenty years it would count far more. The choice of horizon therefore shifts dairy’s position, because so much of its footprint is methane. This is one reason numbers differ between sources, and a good reason to compare values drawn from the same study rather than mixing figures from different ones.

Every number in this article comes from a single source: the synthesis by Joseph Poore and Thomas Nemecek published in Science in 2018, charted by Our World in Data. The study pooled data from more than 38,000 commercial farms across 119 countries and covers the full chain, from land-use change through to packaging and retail. Comparisons are per litre of ready-to-drink product.

Cow’s milk: around 3.15 kg CO₂eq per litre

A litre of cow’s milk emits 3.15 kg CO₂ equivalent on average. That is a global mean, and the spread between farms is wide: the highest-emitting systems can be several times the lowest for the same litre.

Three sources account for most of it.

Enteric fermentation. Cattle are ruminants. Microorganisms in the rumen break down cellulose and produce methane, most of which is belched out. This methane makes up a large share of milk’s footprint and cannot be removed, only trimmed at the margins through feed composition, breeding or certain feed additives.

Feed production. Growing forage and concentrates consumes energy, nitrogen fertiliser (a direct source of nitrous oxide) and land. In Switzerland a substantial share of that land is permanent grassland that could not produce human food any other way, which is a genuine agronomic argument. It does not make the methane disappear.

Manure management and processing. Slurry storage emits methane and nitrous oxide. Pasteurisation, continuous refrigeration and packaging add their share, usually modest next to the first two.

Dairy also loses heavily on the other environmental indicators: roughly ten times more agricultural land per litre, and a much higher level of eutrophication, meaning nitrogen and phosphorus pollution of waterways.

The real ranking of plant drinks

Here are the per-litre figures from the same meta-analysis. Some of them cut against the received wisdom.

DrinkGHG (kg CO₂eq/L)Land (m²/L)Freshwater (L/L)Eutrophication (g PO₄eq/L)
Cow’s milk3.158.9562810.65
Almond0.700.503711.50
Soy0.980.66281.06
Oat0.900.76481.62
Rice1.180.342704.69

Three readings stand out.

Almond emits least and drinks nearly as much water as dairy. Its reputation as a water hog is deserved, but it concerns water, not carbon. A litre of almond drink takes around 371 litres of freshwater against 28 for soy. Since most of the world’s almonds come from California, a region with recurring drought, that water is drawn where it is scarce. Where the water comes from matters as much as how much.

Oat is not automatically the best. It is very good and very consistent, with no weak point, which makes it a sensible default. It does not come first on any of the four measures. Its reputation as “the most sustainable” owes more to marketing and to the fact that oats grow in temperate climates than to any measured advantage.

Rice is the weakest of the five and still beats dairy. Flooded paddies produce methane through anaerobic fermentation in waterlogged soil, which explains the 1.18 kg. Even so, it emits far less than a litre of cow’s milk. Worst among the alternatives is not the same as bad.

Soy is the best compromise. Strong everywhere, first on water and on eutrophication, weak nowhere. It is also the only alternative whose protein profile approaches milk, which matters when the drink genuinely replaces dairy in a diet.

Soy and deforestation: an objection worth reframing

The argument comes up constantly: drinking soy funds the destruction of the Amazon. The numbers say something else.

About 95% of Brazilian soy goes to animal feed, and more than three-quarters of global soy by mass is used as fodder. The largest single driver of Amazonian deforestation remains pasture for beef. Land pressure from soy is therefore overwhelmingly a consequence of livestock farming rather than of anyone drinking soy milk or eating tofu.

For a European consumer a second point matters: almost all Brazilian soy is genetically modified, and the use of GM soy in food is tightly regulated in the European Union. Most soy consumed directly in Europe is grown in Europe, at yields comparable to Brazil’s (about 3 tonnes per hectare in France against 2.9 in Brazil, and higher in Spain and Italy). Swiss retailers’ soy drinks generally state a European or Swiss origin, which you can check on the carton.

Buying local barely moves the needle

This is the most repeated environmental advice and, for most foods, the least effective. Globally, transport accounts for around 5% of food system emissions. The bulk comes from land-use change and on-farm production.

The reason is mechanical: nearly everything that crosses an ocean travels by ship, and air freight makes up just 0.16% of food tonne-kilometres. A container ship emits very little per tonne per kilometre. Shipping a kilo of avocados from Mexico to the UK generates about 0.21 kg CO₂eq in transport, roughly 8% of the product’s total footprint.

The direct consequence for our subject: a litre of soy drink made from European soy and hauled a thousand kilometres still sits far below a litre of milk from the farm down the road. The product matters much more than the distance it travelled. The same logic applies to other categories, as we discuss for local legumes and carbon footprint and for butter versus plant margarine.

Buying local keeps excellent justifications: supporting a regional economy, knowing the farming practices, preserving varieties, cutting packaging. Those are valid reasons. Carbon simply is not the main one.

What the per-litre comparison leaves out

Comparing a litre of milk to a litre of plant drink has a serious limit: they are not nutritionally interchangeable. Whole cow’s milk contains roughly 3.2 g protein and 120 mg calcium per 100 ml according to USDA FoodData Central. Unsweetened almond drink contains around 0.6 g protein. Oat drink, around 0.8 g. Only soy, at roughly 3.5 g per 100 ml in measured products, plays in the same league.

Calcium deserves particular attention. When a plant drink matches milk’s calcium, it is because calcium was added. A German study published in 2025 in the Journal of Health, Population and Nutrition modelled replacing milk with various plant drinks in a reference meal plan for children aged 1 to 3. With non-fortified products, which made up about 80% of the German market, daily intakes of calcium, iodine, vitamin B2 and vitamin B12 fell to roughly half the reference. With fortified products, only iodine remained short, because it is rarely added.

The detail that matters for everyone: in that market survey at most 20% of products were fortified with calcium and fewer than 10% with iodine. Which plant the drink is made from matters less than whether the manufacturer chose to fortify. One factor makes this worse in Europe: organic regulation restricts the addition of vitamins and minerals, so an unfortified organic drink often delivers fewer micronutrients than a fortified conventional one.

Iodine is the nutrient most often forgotten in this debate. Milk and dairy are a major source in many countries, Switzerland included. An adult needs around 140 micrograms a day according to the NHS. If you drop dairy and your plant drink is not fortified, you need to cover it elsewhere, iodised salt being the usual route. We go into detail in our article on balancing iodine on a vegetarian diet.

Two other nutrients to watch if dairy disappears completely: calcium without dairy, with an adult requirement of around 700 mg a day per the NHS, and vitamin B12, which occurs naturally in no unfortified plant food.

Safety: children, arsenic and rice drinks

This section has nothing to do with climate, and for the families concerned it matters more than everything above.

No plant drink before age 1 as a main drink. The NHS is explicit that drinks marketed as rice, oat or almond should not be given to babies under one year old in place of breast milk or infant formula. Cow’s milk is not suitable as a main drink before age 1 either.

No rice drink before age 5. The UK Food Standards Agency advises against giving rice drinks to children under five as a substitute for breast milk, infant formula or cow’s milk. Rice takes up more arsenic from its environment than other cereals, and in a more toxic form. Young children drink proportionally far more liquid relative to their body weight, which concentrates exposure. Over five and in adults, rice drink fits into a balanced diet without concern.

For a child who must avoid dairy, the substitution should not be improvised from a blog post. Cow’s milk protein allergy and lactose intolerance are managed with a doctor or dietitian, not least because the choice of substitute and its fortification shape the balance of the entire diet.

Allergies to watch for everyone: nut-based drinks (almond, hazelnut, cashew) fall under tree nut allergy, and soy is a declarable allergen. These are always stated on the label.

The Swiss framework: why it is never “oat milk”

In Switzerland, the FDHA ordinance on foodstuffs of animal origin (SR 817.022.108) defines milk in article 32 as the product of normal mammary secretion from one or more mammals. The name is reserved. That is why cartons read “oat drink”, “soy drink” or “Hafer Drink” rather than “oat milk”, whatever everyday speech does.

The same ordinance sets the fat categories for drinking milk: at least 35 g per kilogram for whole milk, which is the 3.5% on the label. It also requires the heat treatment to be stated, hence the “past.”, “UHT” or high-pasteurisation markings.

To compare two plant drinks on the shelf, a few reflexes are enough, expanded in our guide to reading a nutrition label:

  • Start with the ingredients list. The percentage of the base ingredient (oats, soy, almonds) typically runs from 5% to 15%; the rest is water.
  • Check for calcium and, if you avoid dairy, for B12 and iodine in the nutrition table. Absent from the table means not added.
  • Watch the sugars. Oat drinks naturally contain sugars released by enzymatic hydrolysis of starch, even with no sugar added. Compare the unsweetened and standard versions.
  • Texture additives and added oils are common; we cover them in our article on additives in commercial plant milks.

Dairy in Switzerland: some orders of magnitude

To place the debate in Swiss context, a few figures from the federal agricultural report for 2024. Swiss milk production is stable at around 3.7 million tonnes a year. Most of it goes to cheese: 204,315 tonnes produced in 2024, including 30,311 tonnes of Gruyère AOP and 12,395 tonnes of Emmentaler AOP.

On the consumption side, drinking milk per capita was 44.7 kg in 2024 and butter 4.9 kg, both stable. Cheese rose to 23.1 kg per person, 400 grams more than in 2023. Comparing the 2000-2002 average with 2022-2024 shows declines for milk, yoghurt, butter and cream, and increases for cheese, processed cheese and fondue.

That shift is worth noting because it changes the reach of the individual gesture: swapping the milk in your morning coffee does nothing about the cheese share, which is precisely the one that is growing and which takes roughly ten litres of milk per kilo of hard cheese.

Making your own: what it changes

Making a plant drink at home removes the individual carton and part of the finished product’s transport. The carbon gain is real but modest, since those are minor items. Two other effects are more tangible.

Cost. A litre of homemade oat drink comes down to a few tens of centimes of rolled oats, far below carton prices. The full calculation is in our comparison of the real cost of homemade oat milk.

Contents. You control the proportion of the base ingredient and the absence of added sugar and oil. The method is simple and needs no machine.

The trade-off is important and rarely mentioned: a homemade drink is fortified with nothing. No added calcium, no B12, no vitamin D, no iodine. If it occasionally replaces a carton in a porridge, this makes no difference. If it genuinely replaces milk in a dairy-free diet, those nutrients have to be covered elsewhere, deliberately.

One more thing: the leftover pulp (okara from soy, oat residue) is food, not waste. It works in breads, patties and crumble toppings.

Seven common mistakes

  1. Quoting “1 kg CO₂ per litre of milk”. This figure circulates widely. The Poore & Nemecek reference value is 3.15 kg. A lower value may reflect a particular system or an incomplete boundary, but it is not comparable with the 0.7 to 1.18 kg of the plant drinks from the same study.
  2. Assuming almond is bad across the board. It is the best of the five on greenhouse gases. Its problem is water, not carbon.
  3. Making “local” the main criterion. Transport is around 5% of food system emissions. Which product you buy matters more than where it came from.
  4. Treating a plant drink as a nutritional equivalent of milk. On protein, only soy comes close. On micronutrients, only fortification counts.
  5. Assuming organic means better fortified. The reverse: organic regulation restricts added vitamins and minerals, and most unfortified drinks are organic.
  6. Giving rice drink to a young child. Not advised under five as a milk substitute, because of inorganic arsenic.
  7. Looking only at carbon. Water, land and eutrophication tell different and sometimes contradictory stories about the same product.

How to actually decide

If climate is your criterion, any of the five plant drinks does the job: all emit at least two and a half times less than a litre of cow’s milk. Choosing between them comes down to other axes.

If you live where water is a pressure point, or the issue matters to you, skip almond and go for soy or oat, both very frugal with water.

If the drink genuinely replaces milk in your diet rather than appearing in the occasional coffee, choose soy fortified with calcium and B12. It is the only profile that holds up long-term without reorganising the rest of your diet.

If taste decides, oat behaves best in hot coffee and foams reasonably, almond suits sweet preparations, soy holds up well in savoury cooking.

And if you still eat dairy, the biggest lever is not the morning carton: it is the total quantity of animal products across the week, cheese included.

Frequently asked questions

What is the actual gap between cow’s milk and plant drinks? Roughly a factor of three on greenhouse gases (3.15 kg against 0.70 to 1.18 kg per litre), about ten on land, and two to twenty on freshwater depending on the drink. The widest gap is on eutrophication.

Is oat the most sustainable drink? Not strictly. It is very consistent with no weak point, but it comes first on none of the four measured indicators. Almond emits less, soy uses less water and causes less eutrophication, rice uses less land.

Should I avoid soy because of deforestation? For soy eaten directly in Europe the objection does not hold: about 95% of Brazilian soy goes to animal feed and most European food soy is grown in Europe. Forest pressure comes first from cattle pasture, then from fodder soy.

Can a plant drink replace milk for a child? Not before age 1 as a main drink, never rice drink before age 5, and in all cases with medical advice if the child must avoid dairy. The critical point is whether the chosen product is fortified with calcium, iodine, B2 and B12.

Is homemade more sustainable? Slightly, mostly on packaging. The main gain is financial. Be aware that a homemade drink contains no added calcium, B12 or iodine.

Why can’t Swiss packaging say “oat milk”? Because article 32 of the ordinance on foodstuffs of animal origin reserves the word “milk” for the mammary secretion of mammals. Hence “drink” on the carton.

Is organic milk much better on carbon than conventional? The spread between farms is wide, but enteric methane is there regardless of the production method. A litre of organic milk still sits well above any plant drink on greenhouse gases.

Sources