Framing
Most carrying capacity estimates are wrong not because the arithmetic is bad but because the question is underspecified. “How many people can Earth sustain” requires four parameters before it’s answerable: at what consumption level, with what technology, measured by what constraint, and for how long. Existing estimates vary so wildly — from under one billion to over ten billion — that the variance is almost entirely in the assumptions, not the biology. That spread is itself information: it tells you the answer is not a fact waiting to be discovered but a function of inputs that have to be declared.
This analysis fixes those parameters explicitly: we are asking how many people Earth can support indefinitely — meaning sustainably on a timescale of centuries or longer, without drawing down non-renewable natural capital — at a level of material welfare consistent with decent living, using technology that is itself sustainable within the same envelope. Fossil fuels are excluded by definition: an indefinite scenario cannot be premised on finite extraction. High-tech renewable infrastructure is included but treated with appropriate skepticism about its own resource dependencies.
The method is to inventory the binding constraints separately, then model their interactions, because the interaction effects are where most analyses fail. Constraints that appear manageable in isolation frequently compound into something harder when they operate simultaneously on the same land, water, and biological systems. The final range is produced by varying assumptions across three scenarios: optimistic, central, and conservative. As will become clear, the direction of travel as you add constraints and extend the timescale is monotonically downward. The range is not symmetric around a comfortable middle.
The Constraint Inventory
1. Nitrogen
Nitrogen is the most immediate hard limit on food production. Plants cannot synthesize protein without it, it depletes from soil through cropping, and prior to the Haber-Bosch process there was no industrial mechanism to replace it. Biological nitrogen fixation — through legumes, free-living soil bacteria, and blue-green algae — is real but limited. The pre-Haber-Bosch ceiling is empirically established: in 1900, with roughly 1.9 billion people, the global food system was already near the biological nitrogen fixation limit. The Green Revolution that allowed population to triple past that point ran on synthetic nitrogen derived from fossil fuel feedstocks (Smil, 2001).
Removing fossil fuels from the nitrogen budget doesn’t return us to 1900 conditions exactly. Biological fixation can be optimized beyond historical practice through careful crop rotation, polyculture design, and management of soil microbial communities. Conservative estimates suggest this could support perhaps 20–30% more food production than pre-industrial biological fixation alone. Diet composition matters significantly: plant-based diets are roughly eight to ten times more nitrogen-efficient per calorie than diets centered on ruminant animals, because the conversion losses through livestock are enormous (Poore & Nemecek, 2018).
Nitrogen constraint ceiling, optimistic assumptions (heavy plant-based diet, optimized biological fixation): approximately 3.5–4 billion.
Nitrogen constraint ceiling, conservative assumptions (mixed diet, realistic biological fixation): approximately 2–2.5 billion.
Uncertainty is moderate. The biological fixation numbers are reasonably well-characterized. The diet-dependency of the ceiling is the main variable.
2. Phosphorus
Phosphorus receives less attention than nitrogen but may ultimately be the harder constraint. It is an irreplaceable plant nutrient — there is no biological analogue to nitrogen fixation for phosphorus, no atmospheric reservoir to draw on. Plants need it, it depletes from soil, and the only sources are geological deposits or recycling from existing biological systems.
Current phosphate rock reserves, at current extraction rates, are estimated to last 50–300 years depending on the methodology and whose reserves are counted (Cordell et al., 2009). But this is the wrong frame for an indefinite analysis. The question is not whether we run out of mineable phosphorus but whether we can close the phosphorus cycle — recover and return essentially all the phosphorus that enters the food system back to agricultural land, indefinitely.
This is technically possible. Phosphorus can be recovered from human waste, animal manure, food processing residues, and waterways. Pre-industrial agriculture in many cultures did this through systematic return of human and animal waste to fields. The problem is efficiency: modern sanitation systems are designed to remove phosphorus from water, not recover it. Closing the loop requires restructuring wastewater infrastructure globally and eliminating the phosphorus losses to oceans that currently run at roughly 10–15 million tonnes per year (Cordell & White, 2011).
A sustainable phosphorus cycle at current population levels would require near-complete recovery and return of all biological phosphorus — essentially no losses to waterways or landfills. This is physically achievable but represents an enormous infrastructural transformation, and it constrains where people can live relative to where food is grown, because long-distance phosphorus transport becomes an energy cost that itself has limits.
Phosphorus constraint: does not produce a clean population ceiling on its own, but significantly constrains the geographic distribution of sustainable population and imposes a hard requirement for circular nutrient management that most carrying capacity estimates ignore. Any estimate that doesn’t specify a phosphorus recycling regime is incomplete.
3. Topsoil
Topsoil is the constraint most people understand least, probably because its depletion is slow enough to be invisible within a human lifetime. It is being lost globally at an estimated 24 billion tonnes per year under industrial agricultural practice — removed by wind and water erosion faster than it forms. Soil formation from bedrock and organic matter accumulation proceeds at roughly 1 centimeter per 500 years under natural conditions, faster under optimized regenerative management but still on century timescales (Montgomery, 2007).
This is the definitional non-renewable resource for an indefinite analysis. An agricultural system that mines topsoil is not sustainable by definition, regardless of what else it does right. Regenerative practices — no-till, cover cropping, perennial polycultures, managed grazing — can halt erosion and gradually rebuild soil depth, but they require significantly lower yield intensity than industrial monoculture during the transition period, and they require land to be managed with a patience that market agriculture does not incentivize (Montgomery, 2017).
The productivity implications are substantial. Industrial monoculture yields are high partly because they’re subsidized by soil capital accumulated over millennia. Regenerative yields are lower in the short run and stabilize at levels that can be sustained — but “stabilize at” is probably 30–60% of peak industrial yields for most staple crops, with significant regional variation (Montgomery, 2017).
Topsoil constraint: does not set a hard population ceiling independently but acts as a downward multiplier on food production estimates from other constraints. Any scenario premised on current agricultural yields is not an indefinite scenario — it is a temporary one.
4. Freshwater
Approximately 70% of global freshwater withdrawal is agricultural (Postel et al., 1996). A large fraction of this — supporting food production for hundreds of millions of people in South Asia, the Middle East, Northern Africa, and the American West — comes from non-renewable aquifer drawdown. The Ogallala Aquifer under the American Great Plains, the Arabian Aquifer, the Northwest Sahara Aquifer System, and major aquifers under the Indo-Gangetic Plain are all being depleted at rates that exceed recharge by orders of magnitude. They are, in the relevant sense, fossil water.
An indefinite food system can only use renewable freshwater: rainfall, sustainable river withdrawals, and groundwater extracted at or below recharge rates. Mapping current agricultural land against renewable freshwater availability shows that a significant fraction of current food production — estimates range from 15–35% of global irrigated crop production — is drawing on water that will not be there in a sustainable scenario (Gleick, 2003; Postel et al., 1996).
This removes substantial productive capacity from the system, concentrated in regions that currently produce large food surpluses. The American Midwest, the Middle East, and parts of South Asia look very different in a sustainable water scenario than they do today.
Freshwater constraint, sustainable withdrawal only: reduces effective agricultural productive capacity by roughly 20–30% from current levels, concentrated in currently high-producing regions. The interaction with climate change — which is already shifting rainfall patterns and glacier-fed river flows — makes this estimate worse over time.
5. Land and the Ecosystem Function Requirement
This is where the analysis diverges most sharply from standard carrying capacity models, and where the stakes are highest.
E.O. Wilson’s Half Earth proposal — that roughly half of Earth’s land surface needs to remain in natural state to maintain biodiversity and ecosystem function — is not merely a conservationist preference. It reflects empirical research on minimum viable habitat for species survival, on the area requirements of intact ecological networks, and on the spatial scale needed to maintain the regulatory services that ecosystems provide to agriculture and human welfare (Wilson, 2016).
Currently, approximately 50% of Earth’s ice-free land has been converted to human use — agriculture, cities, infrastructure (Vitousek et al., 1997; Rockström et al., 2009). Another 25–30% is degraded but not fully converted. This leaves perhaps 20–25% in reasonably intact natural state, which is approximately half of what the ecological evidence suggests is necessary for stable ecosystem function.
The implication is that a sustainable scenario requires not just halting further habitat conversion but actively rewilding significant areas — removing them from food production and allowing ecosystem recovery. This directly reduces the agricultural land base. The area that needs to come out of production to restore ecosystem function is contested, but credible estimates suggest it is substantial: 15–30% of current agricultural land globally, with the most ecologically sensitive and already-degraded areas being the priority.
Combined with the soil and water constraints — which themselves require some land to be managed at lower intensity or rested during recovery — the effective sustainable agricultural land base is significantly smaller than what current food production occupies.
6. Ecosystem Service Dependencies
This constraint is the least quantified and potentially the most important, because it operates through nonlinear thresholds rather than gradual curves. You don’t lose ecosystem services proportionally as ecosystems degrade — you lose them suddenly when functional thresholds are crossed.
Pollination. Approximately 75% of food crops depend on animal pollination to some degree — many staple crops significantly, many fruits and vegetables almost entirely (Foley et al., 2011). Wild pollinator populations are in documented global decline, driven by habitat loss, pesticide exposure, and pathogen pressure. Managed honeybee populations are increasingly fragile. The agricultural system is currently running on a pollination subsidy from wild insect populations that are being depleted faster than they’re being replaced. A sustainable food system requires intact pollinator habitat at landscape scale, which directly conflicts with intensive agricultural land use.
Soil biology. Plant productivity under natural conditions is mediated by extraordinarily complex soil microbial communities — bacteria, fungi, nematodes, arthropods — that fix nutrients, suppress pathogens, regulate water infiltration, and facilitate root function. Industrial agriculture disrupts these communities through tillage, synthetic inputs, and monoculture. Regenerative agriculture works largely by restoring them (Montgomery, 2017). But soil biological communities are incompletely understood, their recovery trajectories after disturbance are poorly characterized, and the functional thresholds below which they fail to support crop production are unknown. This is a large, poorly mapped risk.
Water cycling. Forests and intact wetlands regulate regional rainfall patterns, buffer seasonal flow extremes, and recharge groundwater. Deforestation in one location affects precipitation in agricultural regions that may be far removed — the “flying rivers” of moisture transported from the Amazon basin support rainfall across South America including major agricultural zones. As forest cover declines, these moisture transport systems weaken. The agricultural system depends on rainfall patterns that are themselves partially maintained by the forest cover that agriculture is destroying (Rockström et al., 2009).
Ocean productivity. Marine fisheries currently provide approximately 17% of global animal protein. Global fish stocks are substantially overfished — the FAO estimates that over 35% of stocks are harvested at biologically unsustainable levels, and many more are at maximum sustainable yield with no buffer (FAO, 2022). A sustainable food system cannot depend on fisheries beyond their regenerative capacity, which is significantly below current catch levels. This removes a meaningful protein source from the global food budget.
The critical feature of these dependencies is that they interact. Pollinator decline compounds with habitat loss which compounds with soil biology disruption which compounds with water cycle disruption. The system doesn’t degrade linearly — it tips. And we don’t know where the tipping points are, only that they exist and that we are moving toward them.
Interaction Effects
The constraints above are not independent, and their interactions are where carrying capacity analyses most consistently fail.
Soil loss × water depletion: Degraded soils have lower water retention capacity, requiring more irrigation to support equivalent yields. As topsoil thins and organic matter decreases, the same rainfall or irrigation produces less crop growth — meaning the water constraint and the soil constraint compound rather than add.
Ecosystem loss × agricultural productivity: As pollinator populations decline, as soil biological communities degrade, and as water cycling disrupts regional rainfall, agricultural yields fall even on land that appears physically intact. The ecosystem services that industrial agriculture currently receives for free — and doesn’t account for — are being drawn down simultaneously with the physical inputs. Removing them from the model produces yield estimates that are systematically too high.
Nitrogen × land: Biological nitrogen fixation through legumes requires land — you cannot grow nitrogen-fixing cover crops and cash crops on the same land at the same time, or you do so at reduced yield. The land required for sustainable nitrogen management competes directly with the land required for food production.
Rewilding × food production: Land removed from agriculture for ecosystem restoration is, in the short to medium term, unavailable for food production. The ecosystem services it provides — pollination, water regulation, climate buffering — eventually increase the productivity of surrounding agricultural land, but the transition period involves a real reduction in food system capacity.
Energy × everything: Without fossil fuels, agricultural machinery, transport, refrigeration, food processing, and water pumping all require renewable energy. Renewable energy infrastructure has resource requirements — rare earth minerals, steel, concrete — that are themselves constrained. A fully renewable industrial civilization is possible but operates at lower total energy throughput than the fossil fuel system, with implications for agricultural productivity, food distribution, and the ability to maintain complex global supply chains for inputs and outputs.
The interaction effects consistently work in the same direction: they reduce the sustainable population ceiling below what single-constraint analyses suggest.
Three Scenarios
1. Optimistic Scenario
Assumptions: Near-complete phosphorus recycling globally. Regenerative agriculture adopted at scale, stabilizing soils and rebuilding soil biology over a 100-year transition. Diet is predominantly plant-based (≥80% of calories from plants). Renewable energy provides adequate industrial base. 35% of current agricultural land rewilded, productivity gains from restored ecosystem services partially offset the reduced land base. Pollinator populations stabilized through habitat corridors. Fisheries managed at maximum sustainable yield.
Food system capacity: Roughly 60–70% of current caloric production, more equitably distributed, more stable on multi-century timescales.
Population range: 2.5 – 4 billion.
Caveats: This scenario requires a level of global coordination, institutional capacity, and voluntary consumption reduction that has no historical precedent. It assumes the transition to regenerative systems proceeds without catastrophic loss of soil or ecosystem function during the changeover — a significant assumption given current trajectory. It also assumes renewable energy infrastructure can be built and maintained without requiring inputs that create their own depletion problems.
2. Central Scenario
Assumptions: Substantial but incomplete phosphorus recycling (70–80% recovery). Mixed agricultural systems — regenerative on most land, some continued conventional production. Diet predominantly plant-based but with meaningful animal product consumption (~60% plant calories). 25% of agricultural land rewilded. Renewable energy with some continued use of mineral resources at rates that are sustainable but require active management. Fisheries at reduced but sustainable harvest levels.
Food system capacity: Roughly 50% of current caloric production.
Population range: 1.5 – 2.5 billion.
Caveats: This scenario is probably more achievable than the optimistic one but still requires transformations far beyond anything currently politically visible. The phosphorus recycling and rewilding requirements are particularly demanding. The population range overlaps with but extends below the optimistic scenario, reflecting both lower assumed productivity and more realistic assumptions about coordination failures.
3. Conservative Scenario
Assumptions: Takes ecosystem function thresholds seriously as hard constraints rather than gradual degradations. Assumes nonlinear losses in pollination, water cycling, and soil biology as habitat falls below functional thresholds — thresholds we are likely already approaching in some regions. Requires 50% of land in natural state (Wilson, 2016). Accounts for the transition period during which degraded soils are recovering and not fully productive. Includes realistic estimates of fishery collapse under continued pressure (FAO, 2022). Does not assume successful global phosphorus recycling coordination.
Food system capacity: Roughly 25–40% of current caloric production, with significant regional variation.
Population range: 600 million – 1.5 billion.
Caveats: This scenario may appear catastrophist but is arguably the most internally consistent with the question as posed. It is the scenario produced by taking all the constraints seriously simultaneously, on an indefinite timescale, without optimistic assumptions about coordination. The lower end of the range — under 1 billion — reflects the possibility that ecosystem function thresholds have already been crossed in ways that will not be apparent until the supporting systems fail.
Key Uncertainties
Ecosystem threshold locations. The largest uncertainty in this analysis is where the nonlinear thresholds in ecosystem function actually sit. If pollinator collapse, soil biology failure, or water cycle disruption happen at less habitat loss than current estimates suggest, the conservative scenario’s lower bound could be too high. This uncertainty is asymmetric: being wrong in this direction is catastrophic and irreversible, which argues for weighting the conservative scenario more heavily.
Regenerative agriculture yield stabilization. The 30–60% yield reduction estimate under regenerative systems is a range with real uncertainty. Some practitioners and researchers report smaller yield gaps, particularly for legumes and diverse polycultures. If regenerative systems can sustain 70–80% of industrial yields at full implementation, the optimistic scenario population ceiling rises. But this has not been demonstrated at scale under the full constraint set described here.
Renewable energy infrastructure sustainability. The optimistic and central scenarios assume renewable energy can support meaningful industrial capacity indefinitely. The mineral requirements for solar, wind, and battery systems — lithium, cobalt, neodymium, silver — are not clearly sustainable at the scale required for a global industrial civilization. If energy throughput in a sustainable scenario is significantly lower than renewable advocates project, agricultural productivity, food distribution, and water management all suffer.
Climate change trajectory. This analysis has largely set aside the effects of already-committed climate change on agricultural productivity, rainfall patterns, and ecosystem stability. Those effects are real and mostly negative for food production in the regions where the remaining agricultural land base is concentrated (Steffen et al., 2015). Including them pushes all three scenario ceilings downward.
Synthesis
The three scenarios produce a total range of approximately 600 million to 4 billion. This looks wide, but the structure of the range matters more than its width.
The direction of travel as you add constraints and take the timescale seriously is consistently downward. The optimistic scenario requires coordination, technology deployment, and behavioral change at scales without historical precedent. The conservative scenario requires only that the physical and biological constraints behave as the evidence suggests they do. This asymmetry means the conservative scenario is the more defensible prior, with the optimistic scenario representing an upper bound that is achievable in principle but not the default trajectory.
The central estimate — 1.5 to 2.5 billion — represents what a well-governed, globally coordinated, ecologically serious civilization might achieve on an indefinite timescale. It requires genuinely radical transformation of food systems, land use, energy infrastructure, and consumption patterns, but does not require everything to go right simultaneously.
Current world population is approximately 8 billion. We are already above the optimistic scenario ceiling. This means the current global population is not sustainable on an indefinite timescale under any internally consistent set of assumptions — we are sustaining it by drawing down natural capital that is not being replenished. This is not a prediction about when collapse occurs. Natural capital depletion can continue for a long time before systems fail visibly. But it does mean that population stabilization and gradual reduction — through non-coercive means: education, women’s autonomy, economic security, access to family planning — is not just a concern but a physical requirement on a long enough timescale.
It also means that the standard debate between “overpopulation” and “overconsumption” is a false dichotomy. Both are true simultaneously on an indefinite timescale. Reducing consumption buys time and significantly raises the sustainable ceiling, but it does not make any current population level indefinitely sustainable. And reducing population without reducing consumption per capita does not solve the problem either. The constraints are on total throughput — people multiplied by consumption — and both terms matter.
The number most carrying capacity analyses avoid stating directly is the central estimate: roughly 2 billion, on a planet currently supporting 8. That gap is not an argument for coercion. It is an argument for taking seriously, on a civilizational timescale, what the physical constraints actually imply — rather than continuing to produce analyses scoped to 2050 that end before the slow-moving depletions become dominant, which is precisely what happens when you choose a timeframe short enough to avoid them.
References
Soil & agriculture
Montgomery, D.R. (2007). Dirt: The Erosion of Civilizations. University of California Press.
Montgomery, D.R. (2017). Growing a Revolution: Bringing Our Soil Back to Life. W.W. Norton & Company.
Nitrogen & the Haber-Bosch limit
Smil, V. (2001). Enriching the Earth: Fritz Haber, Carl Bosch, and the Transformation of World Food Production. MIT Press.
Phosphorus cycling
Cordell, D., Drangert, J-O., & White, S. (2009). The story of phosphorus: Global food security and food for thought. Global Environmental Change, 19(2), 292–305.
Cordell, D., & White, S. (2011). Peak phosphorus: Clarifying the key issues of a vigorous debate about long-term phosphorus security. Sustainability, 3(10), 2027–2049.
Food systems & diet
Foley, J.A., et al. (2011). Solutions for a cultivated planet. Nature, 478, 337–342.
Poore, J., & Nemecek, T. (2018). Reducing food’s environmental impacts through producers and consumers. Science, 360(6392), 987–992.
Planetary boundaries & ecological economics
Rockström, J., et al. (2009). A safe operating space for humanity. Nature, 461, 472–475.
Steffen, W., et al. (2015). Planetary boundaries: Guiding human development on a changing planet. Science, 347(6223), 736 / Article 1259855.
Biodiversity & land
Wilson, E.O. (2016). Half-Earth: Our Planet’s Fight for Life. Liveright Publishing.
Vitousek, P.M., et al. (1997). Human domination of Earth’s ecosystems. Science, 277(5325), 494–499.
Freshwater
Postel, S.L., Daily, G.C., & Ehrlich, P.R. (1996). Human appropriation of renewable fresh water. Science, 271(5250), 785–788.
Gleick, P.H. (2003). Global freshwater resources: Soft-path solutions for the 21st century. Science, 302(5650), 1524–1528.
Fisheries
FAO. (2022). The State of World Fisheries and Aquaculture 2022. Food and Agriculture Organization of the United Nations, Rome.
Mind Stuffing 