10 Million Deaths by 2050: The AMR Crisis That Climate Change Is Accelerating

In 2016, the Review on Antimicrobial Resistance projected 10 million annual deaths from drug-resistant infections by 2050. In 2022, The Lancet published the first comprehensive count of actual AMR deaths: 1.27 million directly attributable, 4.95 million associated, in 2019 alone. Both figures are now central to global health policy. Neither included a variable that research published in 2026 suggests may be significant: climate-driven environmental resistance.

This is an assessment of what those numbers contain, what they exclude, and what that exclusion means for the trajectory ahead.

What the O'Neill Review Actually Said

The UK government and the Wellcome Trust commissioned the Review on Antimicrobial Resistance in 2014. Economist Jim O'Neill chaired the effort. The final report, published in May 2016, projected that antimicrobial resistance would kill 10 million people per year by 2050 if no corrective action was taken, surpassing cancer as a cause of death. The review estimated a cumulative cost of $100 trillion in lost global production over the same period.

These numbers were derived from economic modeling by RAND Europe and KPMG. The models extrapolated clinical resistance trends for six pathogens and diseases, including drug-resistant Klebsiella pneumoniae, E. coli, Staphylococcus aureus, HIV, tuberculosis, and malaria. The extrapolation assumed a continuation of observed resistance trajectories in healthcare settings.

What the models did not include is as important as what they did. The O'Neill methodology tracked clinical resistance - bacteria becoming resistant through exposure to antibiotics in hospitals, communities, and agriculture. It did not model environmental resistance: the selection of resistant bacteria through natural antibiotic exposure in soil, water, and other ecosystems. This was not an oversight at the time. The scale of the environmental resistome's contribution to clinical resistance was not yet established.

The 10-million figure became the most cited number in AMR policy globally. It appeared in WHO resolutions, UN declarations, and national action plans. Its influence is difficult to overstate. Its limitations, however, received less attention.

The Lancet Correction

Murray et al. published the Global Burden of Bacterial Antimicrobial Resistance study in The Lancet in January 2022. This was the first systematic effort to count AMR deaths across 204 countries and territories using a consistent methodology. The study estimated 1.27 million deaths directly attributable to bacterial AMR and 4.95 million deaths associated with bacterial AMR in 2019.

Six pathogens accounted for the majority of deaths: Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, Streptococcus pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa. Lower respiratory infections were the most common infection type associated with AMR deaths. Sub-Saharan Africa and South Asia bore the highest burden per capita.

The Lancet numbers served as both a correction and a confirmation. The 2019 baseline of 1.27 million direct deaths was lower than what a simple interpolation from O'Neill's 2050 projection might have suggested. At the same time, 4.95 million associated deaths exceeded many prior estimates. The study confirmed that AMR was already a leading cause of death globally, exceeding both HIV/AIDS and malaria individually in direct mortality.

What matters for this analysis: the Lancet study, like the O'Neill Review before it, measured clinical outcomes. It counted deaths from resistant infections in healthcare settings. It did not attempt to model the environmental processes that generate and amplify resistance before bacteria reach patients.

The Variable Neither Model Included

In 2026, Shan et al. published findings in Nature Microbiology demonstrating that drought increases the concentration of natural antibiotics in soil. As soil dries, antimicrobial compounds such as phenazine-1-carboxylic acid become more concentrated. This creates stronger selection pressure. Bacteria that survive carry resistance genes. When these resistant bacteria reach human populations through dust, water, or the food chain, they contribute to clinical resistance.

The clinical correlation supports the mechanism. Using antibiotic resistance data from hospitals in 116 countries, the researchers found that the average frequency of drug-resistant clinical isolates was strongly correlated with local aridity.

Climate projections make this finding forward-looking. Under all major emissions scenarios, arid zones expand. The population living in drylands, currently around 2.3 billion, is projected to reach approximately 4 billion by 2050 according to UNCCD data. This expansion does not merely add geography to the resistance map. It adds selection pressure to ecosystems that were previously humid enough to dilute natural soil antibiotics.

This variable was absent from the O'Neill models. It was absent from the Lancet's burden estimates. It was absent from every major AMR projection published through 2025. Its inclusion would not merely add a percentage to existing numbers. It introduces an independent driver that accelerates resistance through a mechanism entirely separate from clinical antibiotic use.

Assessment: the presence of this unmodeled driver suggests the O'Neill projection is more likely an underestimate than an overestimate. By how much remains unknown. No published model has yet integrated climate-driven environmental resistance into AMR death projections.

The WHO Policy Framework and Its Gaps

The World Health Organization adopted its Global Action Plan on Antimicrobial Resistance at the World Health Assembly in 2015, one year before the O'Neill Review's final report. The plan established five strategic objectives: improve awareness and understanding; strengthen surveillance and research; reduce infection incidence; optimize antimicrobial use; and ensure sustainable investment in countering AMR.

WHO set a target for all member states to develop National Action Plans on AMR by 2017. By 2024, 178 countries had developed such plans. Implementation, however, lagged far behind policy language. Only 11 percent of those countries had dedicated funding for implementation. In many low- and middle-income countries, the plans existed on paper but lacked the resources to translate into action.

The UN Interagency Coordination Group on Antimicrobial Resistance published its final report in April 2019, calling for a One Health approach that would integrate human health, animal health, and environmental dimensions. The UN General Assembly held a High-Level Meeting on AMR in September 2024, the second such meeting after 2016. The declaration acknowledged links between climate change and AMR but set no specific targets for addressing the climate-AMR nexus.

The WHO AWaRe classification system, which categorizes antibiotics into Access, Watch, and Reserve groups, has become a stewardship tool used globally. Its adoption has helped rationalize antibiotic prescribing in some contexts. It does not, by design, address environmental resistance generation.

Assessment: the policy framework was built on the premise that AMR is primarily a problem of antibiotic misuse in human medicine, animal husbandry, and agriculture. This premise is not wrong, but it is incomplete. The framework has no mechanism for addressing climate-driven environmental resistance.

The Funding Gap

Global public and philanthropic spending on AMR research and development runs at approximately $1.6 to $1.8 billion annually, according to WHO data. The O'Neill Review recommended a $2 billion Global Innovation Fund over five years for early-stage AMR research, alongside much larger market entry rewards for successful new drugs. The broader investment targets remain unmet.

For comparison, the U.S. National Institutes of Health alone allocates approximately $3.3 billion annually to HIV/AIDS research. Public and philanthropic cancer research funding averaged $5.5 to $6.6 billion per year between 2016 and 2019, with private industry spending adding substantially to that total. These comparisons are imperfect because the disease categories differ in scope, but they illustrate a scale mismatch. AMR is projected to kill more people than cancer by 2050, yet receives a fraction of cancer research funding.

The most significant dedicated AMR funding mechanisms are CARB-X, which has received over $500 million in funding since 2016 and directed $361 million into preclinical antibiotic research projects, and GARDP, the Global Antibiotic Research and Development Partnership, which operates with an annual expenditure of approximately 26 million EUR. Both focus on developing new therapeutics rather than understanding environmental resistance drivers.

Climate adaptation budgets present a parallel gap. The hundreds of billions allocated globally for climate adaptation do not account for AMR as a climate-driven health outcome. Health chapters in national climate adaptation plans rarely mention antimicrobial resistance. AMR national action plans rarely mention climate change. The two policy streams run in parallel, addressing overlapping problems in isolation.

What Actually Works

Evidence from Northern Europe suggests that targeted interventions can reduce resistance rates. Sweden, Norway, and Denmark maintain among the lowest antibiotic consumption rates in Europe, according to ECDC surveillance data. Their AMR rates are correspondingly lower than the European average. The combination of strict prescribing guidelines, comprehensive surveillance, strong infection prevention programs, and public awareness has produced measurable results over decades.

The Netherlands achieved a 69 percent reduction in livestock antibiotic use between 2009 and 2020 through regulatory pressure, industry cooperation, and veterinary oversight reform. Agricultural AMR rates declined in parallel.

The United Kingdom published a 5-Year National Action Plan on AMR covering 2019 to 2024 with measurable targets for reducing antibiotic use and resistant infections. The UK also piloted a subscription-based antibiotic procurement model, sometimes called the Netflix model, in which the government pays pharmaceutical companies a fixed annual fee for access to new antibiotics regardless of the volume used. This addresses the market failure that has driven companies out of antibiotic development: the more effective a new antibiotic is kept in reserve, the less revenue it generates.

These interventions share a critical limitation. All of them target antibiotic use in clinical and agricultural settings. None addresses the environmental generation of resistance through natural processes like drought-driven selection pressure. Stewardship reduces the clinical contribution to resistance. It does not reduce the environmental contribution that the Shan et al. finding identifies.

Assessment: the proven interventions work for the problem they were designed to solve. They are necessary. They are also insufficient if climate-driven environmental resistance constitutes a significant and growing portion of total resistance pressure.

Climate-Driven AMR and the Developing World

The geographic overlap between projected climate impacts and existing AMR burden is substantial. Sub-Saharan Africa and South Asia already carry the highest per-capita AMR death rates according to the Lancet 2022 data. The same regions face the greatest projected aridification under IPCC scenarios. Health systems in these regions typically spend between $10 and $50 per capita on healthcare, compared to $5,000 or more in high-income countries.

Many low- and middle-income countries lack basic microbiology laboratory capacity. Without the ability to identify which bacteria are causing infections and whether they are resistant, surveillance is blind. AMR deaths go uncounted. The true burden in these regions is almost certainly higher than the Lancet estimates, which relied heavily on hospital data from countries with functioning laboratory systems.

Over-the-counter antibiotic access in many LMICs compounds the problem. When patients can purchase antibiotics without prescription, subtherapeutic doses become common, driving clinical resistance. When this clinical pressure is layered on top of climate-driven environmental resistance, the compound effect exceeds what either driver would produce alone.

Climate adaptation planning in these countries focuses on food security, water access, and extreme weather preparedness. AMR rarely appears in adaptation frameworks. Conversely, the AMR national action plans that exist in these countries rarely account for how expanding drought will alter their resistance landscape over the coming decades.

What the Revised Calculus Looks Like

The O'Neill Review's 10 million annual deaths by 2050 was always a projection built on assumptions that its authors were transparent about. The Lancet 2022 study provided the first reliable baseline: 1.27 million direct AMR deaths in 2019. The trajectory from that baseline toward the 2050 projection requires acceleration.

Climate-driven environmental resistance is a candidate mechanism for that acceleration. The Shan et al. finding does not prove that AMR deaths will exceed 10 million by 2050. It establishes that a significant driver of resistance was excluded from the models that produced that number.

What we know: AMR killed at least 1.27 million people directly in 2019 and was associated with 4.95 million deaths. Resistance rates are increasing in most regions. Arid zones are expanding under climate change. Drought increases natural antibiotic concentrations in soil, selecting for resistant bacteria. The policy framework and funding architecture were built without accounting for this environmental driver.

What we do not know: the magnitude of the climate contribution to total AMR burden. No model has quantified it. No study has isolated its effect from clinical and agricultural drivers at a global scale.

What is clear: the most widely cited AMR projection is based on an incomplete set of inputs. The variable it excluded works in the direction of higher, not lower, resistance. The policy frameworks built around that projection are correspondingly incomplete. Until climate-driven environmental resistance is integrated into AMR models, surveillance systems, and policy responses, the global approach to antimicrobial resistance is addressing a partial picture of the problem.