Can regenerative farming feed the world?

On the most careful assessment anyone has made of it, all the grassland on Earth, grazed as well as the evidence allows, would supply about 7 grams of animal protein per person per day. Assume substantial gains in how well animals convert grass and the estimate reaches 18. World supply today is 37 grams a day. In the United States it is 84.

That gap is the argument, and the arithmetic on it is division rather than opinion. On the conservative end of the estimate, keeping the world's current diet on grass alone would take five Earths' worth of grassland, and feeding everyone the way Americans eat would take twelve. Grant the optimistic end instead and it is two Earths, and five. There is one.

Regenerative grazing means moving herds so pasture recovers between visits, the way wild herds moved. Done well it builds soil carbon, brings degraded land back, carries more species than a monoculture, and gives animals a life a feedlot cannot. Allan Savory took the idea global by arguing that grazing half the world's grasslands could return atmospheric carbon to pre-industrial levels, and a genre of books and documentaries has carried it since. That particular claim has not survived review. The gains below are the ones that have.

From the POV of a welfarist it wins against the feedlot. The question is what it does at eight billion people, and that has some well-researched answers.

The ceiling is set by grass

Grazed and Confused? is the most thorough assessment: two years of work by researchers at Oxford, Aberdeen, Cambridge, Wageningen, SLU, FiBL and CSIRO, written to steelman the grazing case rather than attack it.

If all animal protein came from grasslands alone, the report finds per capita daily supply “could be increased to between 7–18g depending upon, among other things, assumptions about animal productivity.” Feed pigs and poultry on food waste, crop residues and by-products as well, and the ceiling reaches 21g.

Those two ends are not equally safe to quote. 7g is what grassland yields on animal productivity as it stands. Reaching 18g means assuming animals convert grass into protein substantially better than they do now, everywhere, at once. The 21g figure is a different scenario again, because most of what it adds is not grass.

Set that against what people are supplied with. The United States supplies 84.3g of animal protein per person per day, France 78.0, the United Kingdom 64.7, China 47.2, and the world as a whole 37.0, on FAO food balance sheets for 2020.

The division is unforgiving. An American landing at the bottom of the grass range keeps 8% of current supply, at the top 21%, and 25% with waste streams included. The world average would have to fall by half at best and four fifths at worst.

Expressed as land rather than as diet, the same sum reads: two to five Earths' worth of grassland to keep the world eating as it does now, and five to twelve to feed everyone the way the United States eats. That is a ratio of land demand rather than a survey of real acres, and it assumes the missing grassland would be as productive as the grassland already counted, which land at the margin rarely is. There is one Earth, and on most land-use datasets somewhere around a quarter of its ice-free surface is already grazed.

Two caveats keep it honest. Food balance sheets measure supply rather than intake, so the 37 and the 84 include what gets thrown away; intake estimates for the same period run a few grams lower. And FAO counts fish and seafood in animal protein while the grasslands scenario covers land animals only, so fisheries would sit on top of the 7 to 18g band. Neither moves the direction. Both mean the comparison is approximate at the edges.

Does rotation raise the ceiling?

That ceiling is for grazing in general. The claim specific to regenerative grazing is that resting pasture between visits grows more grass, so the question is whether managing land that way lifts the number.

The controlled evidence says it does not. Briske and colleagues assembled the experimental comparisons of rotational against continuous grazing and found continuous grazing equal or better in 87% of experiments on plant production (20 of 23), 92% on animal production per head (35 of 38), and 84% on animal production per unit area (27 of 32). The paper opens: “In spite of overwhelming experimental evidence to the contrary, rotational grazing continues to be promoted and implemented as the only viable grazing strategy.”

Advocates of adaptive multi-paddock grazing answer that those trials did not test the version they practise, and Richard Teague, a co-author of that review, has since argued exactly that. Grazed and Confused weighed the case and found the claims “rest on extrapolation from a small number of case-studies.”

Which leaves the regenerative version of the sum looking like the same sum. Controlled comparisons show no general productivity advantage for rotation, and in most of those trials it came out slightly behind, so on that evidence matching a given output would take more land rather than less. That does not rule out a well-run system beating a badly run one on a given site, or rotation earning its keep on erosion, sward composition and drought resilience, which a yield figure does not capture. It does rule out the large general gain the argument needs. Five Earths is not the figure for careless grazing that good management would fix. It is the figure with good management already assumed.

The appetite is growing, and growing where the ceiling bites

The 7 to 18g band is not an abstraction. It brackets what two countries eat right now: Nigeria at 7.2g, India at 17.8g. Those are among the few places the ceiling already fits, and they are among the places consumption is rising fastest. Why? Because the more prosperous countries are, the more animal products they can afford.

The OECD-FAO Agricultural Outlook 2025-2034 projects global meat output up 13% over the decade, with 45% of consumption growth in upper-middle-income countries, driven by population and income growth. Africa's population rises from 1.5 to 1.8 billion and its meat consumption rises 33%. Per person, global meat consumption rises only 3%, to 29.3 kg a year, because high-income countries have largely stopped growing: they are 17% of the world's population and 35% of its meat consumption.

Put together, that makes the arithmetic harder rather than easier. The ceiling is fixed per person and the number of people is rising. The countries sitting inside the band today are the ones moving out of it, and they are moving because they are getting richer, which is not a trend anyone proposes to reverse.

The American version of the same sum

Hayek and Garrett modelled the entire US beef herd switching to grass finishing. Everything in this section is that model's output, under its assumptions, for one country.

Start with what it provides now. Beef finished entirely on forage is under 1% of US supply. The ceiling on existing pastureland, meaning land already grazed, is 27% of current beef output. That is the number that matters, because it is the only one that takes nothing away from anybody.

Above 27%, the model only reaches further by taking cropland. Allowing cropland-grown hay and silage reaches 61%; converting land now growing feed grain over to alfalfa reaches 71%. Both take land that grows human food, and that trade does not get made. Food supply is a security question and animal products are a preference. No government prices one against the other and picks the preference.

Holding output constant makes it worse. On the study's production parameters, grass-finished cattle gain 1.4 lbs a day against 3.0 in a feedlot and are slaughtered at 1115 lbs rather than 1386 - figures that move with breed, climate, forage and supplementation, though not in a direction that closes the gap. Matching current output needs the national herd to grow from 77 million to 100 million cattle. Methane per unit of beef rises 43%, and total US methane rises about 8%. Avoiding both native vegetation clearance and competition with human food would need pasture productivity to rise 40 to 370%.

The paper's conclusion: “only reductions in beef consumption can guarantee reductions in the environmental impact of US food systems.”

Soil carbon does not close the gap

Sequestration from managed grazing is real in places. Grazed and Confused puts it at 20 to 60% of emissions from grazing systems, 4 to 11% of total livestock emissions, and 0.6 to 1.6% of total annual human greenhouse gas emissions.

Two limits sit underneath that. Accumulation slows as soils approach equilibrium, over decades rather than indefinitely, so the benefit has a horizon rather than a run rate. And it reverses: a change of management, a fire or a drought releases what was stored.

On methane, the report tests the argument that a short atmospheric lifetime makes it unimportant and answers directly: “This argument is false.” The short lifetime changes the timing and the accounting, not the fact of the warming: a single pulse fades, while a continuing source holds the atmospheric burden up for as long as it runs. Keep the ruminants and you keep the methane.

Claims made for holistic and adaptive grazing management, the report notes, “rest on extrapolation from a small number of case-studies,” with mixed peer-reviewed results and small benefits where any appear at all.

It answers one objection out of two

Grazed and Confused covers greenhouse gases only and says so, which leaves this question standing on its own.

A grass-fed steer is trucked to the same slaughterhouse as a feedlot steer, at around 30 months rather than 18. The same structural facts about ordinary farming apply to him: dairy still runs on pregnancy and separation whatever the cow grazes. Hayek and Garrett note that welfare gains from pasture may be offset by thinner oversight of larger herds, and that cattle on US farms have no federal or state legal protection at all.

So the model moves one thing and not the other. If what you object to is how animals live, pasture is a real improvement, bought by keeping 30% more of them. If what you object to is killing them, nothing here touches it.

Where that leaves it

Regenerative grazing beats a feedlot on soil, on biodiversity and on how the animals live, and it is not a solution to the problem it gets invoked to solve. On land per unit of protein, and on methane per unit of beef, it does worse. Advocates on both sides keep one half and drop the other.

The gains are there: better soil in places, more species, animals outdoors, degraded land recovered. Anyone waving that away is arguing against the evidence.

The ceiling is there too, and it is division. Grass grows at a given rate over a given area, feeds a given number of animals, and on the best estimate available yields 7 to 18 grams of animal protein per person per day, 21 once non-grass waste streams are added. World supply is 37, the United States is at 84, and there is one Earth.

One thing the sum does not say is that grazing has no place. Animals eating crop residues, by-products and land that grows nothing else are a real part of a food system, and in low-income pastoral regions they are a livelihood and a micronutrient source rather than a preference. What the arithmetic rules out is 37 grams a day for everyone, and 84.

Followed to the end, the regenerative case arrives where the argument it was built to answer arrives: eat far fewer animal products. Its own most careful advocates produced the numbers that say so.