Why the world is running out of antibiotics, and what comes next
In January 2022, The Lancet published what remains the most comprehensive study of antimicrobial resistance (AMR) ever conducted. The Global Research on Antimicrobial Resistance (GRAM) project analyzed data from 204 countries, 23 pathogens, and 88 pathogen-drug combinations. The finding: in 2019, drug-resistant bacteria directly killed 1.27 million people and were associated with 4.95 million deaths.
To put that in perspective: HIV/AIDS killed 860,000 people that year. Malaria killed 640,000. Antibiotic resistance killed more than either, and more than both put together.
In September 2024, the GRAM team published their follow-up, extending the analysis from 1990 to 2021 with projections to 2050. The updated picture is worse:
More than 1 million people have died from AMR every single year since 1990. Over 36 million cumulative deaths in three decades. By 2050, AMR will kill nearly 2 million people directly every year — roughly 3 deaths per minute.
Six pathogens are responsible for the vast majority of AMR deaths. Their names are clinical, but their toll is staggering:
| Pathogen | Key Resistance | Deaths (2021) |
|---|---|---|
| Staphylococcus aureus | MRSA | 130,000 |
| Mycobacterium tuberculosis | MDR/XDR-TB | ~150,000 |
| Escherichia coli | 3GC-R, Carb-R | >100,000 |
| Klebsiella pneumoniae | Carb-R | >100,000 |
| Streptococcus pneumoniae | Multi-R | 120,000 |
| Acinetobacter baumannii | Carb-R | >100,000 |
MRSA alone — Staphylococcus aureus resistant to methicillin — killed 130,000 people in 2021, more than double the 57,200 it killed in 1990. Carbapenem-resistant gram-negative bacteria collectively caused 216,000 attributable deaths, up 70% from 1990.
AMR is everywhere, but it is not equally distributed. The WHO's 2025 GLASS report, drawing on 23 million confirmed cases from 104 countries, found:
South-East Asia: 1 in 3 infections resistant
Eastern Mediterranean: 1 in 3 infections resistant
Sub-Saharan Africa: ~1 in 5, but severe data gaps
Europe: 1 in 10 infections resistant
Western Pacific: 1 in 10 infections resistant
In parts of Africa, over 70% of E. coli infections are resistant to third-generation cephalosporins. In Southeast Asia, 41% of Klebsiella infections resist carbapenems — the antibiotics of last resort. The GRAM study projects 11.8 million direct AMR deaths in South Asia alone between 2025 and 2050.
In September 2025, the CDC reported that infections caused by NDM-producing carbapenem-resistant Enterobacterales (NDM-CRE) — bacteria the agency calls "nightmare bacteria" — increased by more than 460% from 2019 to 2023 across 29 US states.
Only two antibiotics remain effective against these organisms. Susceptibility testing for both is not widely available.
Candida auris, a drug-resistant fungus, tells a parallel story. US clinical cases rose from 476 in 2019 to 6,304 in 2024 — a thirteen-fold increase. Pan-resistant strains that defeat all three antifungal drug classes have been documented. Parts of southern Europe now report regional endemicity.
These are not theoretical scenarios. They are current surveillance data from the world's best-resourced healthcare systems.
The golden age of antibiotic discovery ran from the 1940s through the 1960s. Most of the antibiotic classes we use today — penicillins, cephalosporins, tetracyclines, aminoglycosides, macrolides, fluoroquinolones — were discovered in that window. Since then, discovery has slowed to a trickle, and then effectively stopped.
No truly novel class of antibiotic has been discovered and brought to market since 1987.
That year, the lipopeptide class (daptomycin) was identified. It didn't reach patients until 2003. The few "new" antibiotics approved since then — linezolid (2000), fidaxomicin (2011), bedaquiline (2012) — all belong to classes originally discovered decades earlier. They are variations on old themes, not genuinely new weapons.
| Drug (Approval) | Marketed As | Class Actually Discovered |
|---|---|---|
| Linezolid (2000) | Novel mechanism | Oxazolidinones, 1978 |
| Daptomycin (2003) | New class | Lipopeptides, 1987 |
| Retapamulin (2007) | First-in-class | Pleuromutilins, 1952 |
| Fidaxomicin (2011) | Novel narrow-spectrum | Tiacumicins, 1970s |
| Bedaquiline (2012) | New TB drug | Diarylquinolines, 1990s |
Why did discovery stop? Partly because the easy targets were found first. Bacteria have a limited number of essential metabolic pathways, and most have already been exploited. Partly because the natural sources — soil bacteria, marine organisms — have been extensively mined. But mostly because the economics turned catastrophic.
The antibiotic market is not merely struggling. It is a textbook case of multiple, compounding market failures so severe that they have produced a paradox: society desperately needs new antibiotics, but it is economically irrational to develop them.
A new oncology drug generates, on average, 25 times more revenue in its first two years than a new antibiotic. The top-performing antibiotics peak around $100–200 million in annual sales. Keytruda, Merck's cancer immunotherapy, generates over $25 billion per year.
The failure has five interlocking components:
Short treatment courses. Antibiotics cure infections in 3–21 days. Cancer drugs, diabetes drugs, and biologics are taken for months, years, or a lifetime. Each antibiotic patient is a one-time sale.
The conservation paradox. The most valuable new antibiotics — those effective against resistant organisms — are deliberately held in reserve. Physicians prescribe them only when everything else has failed. This is medically essential but economically devastating: the best new antibiotics are used the least.
Generic competition. Older antibiotics are available as cheap generics for $5–20 per course. Hospitals are under intense pressure to use these first. New antibiotics must compete against drugs that cost pennies per dose.
Low reimbursement. Unlike oncology drugs ($100,000+ per course), antibiotics are priced low — typically $1,000–3,000 per course for novel agents in the US. Hospital reimbursement bundles drug costs into fixed payments, creating a perverse incentive to use the cheapest available option.
Inevitable obsolescence. Unlike other drugs, antibiotics lose effectiveness over time as resistance spreads. Companies invest billions developing a product whose utility will erode with every use.
The fundamental paradox: society needs new antibiotics but does not use them. The public health value is enormous; the commercial value approaches zero. This is not a problem the free market can solve.
The market failure is not theoretical. It has produced a trail of corporate wreckage.
Of roughly 18 major pharmaceutical companies that once had active antibiotic programs, at least 14 have left. The exits accelerated in 2016–2018:
Today, only about four large pharmaceutical companies — GSK, Pfizer, Merck, and Roche — maintain any active antibiotic R&D programs.
With big pharma gone, the hope was that small biotech companies would fill the gap. They tried. They failed. The pattern is devastating: every small or mid-size company that brought an FDA-approved antibiotic to market since 2010 has gone bankrupt or been sold at a loss.
| Company | Drug (Approved) | Fate | Key Detail |
|---|---|---|---|
| Achaogen | Plazomicin (2018) | Bankrupt 2019 | $560M invested; $800K first-year sales |
| Melinta | Baxdela, Vabomere, Orbactiv | Bankrupt 2019 | Four antibiotics; $1B in debt |
| Tetraphase | Xerava (2018) | Fire-sale ~2020 | Sought emergency loans; sold at distressed valuation |
| Nabriva | Xenleta (2019) | Wound down 2023 | First novel-mechanism pneumonia drug in ~20 years |
| Paratek | Nuzyra (2018) | Acquired below cost | Sold to Novo Holdings at a loss |
The Achaogen story is the most instructive. The company spent $560 million developing plazomicin, an antibiotic effective against carbapenem-resistant bacteria — exactly the kind of drug the world needs. The FDA approved it in June 2018. In its first year on the market, plazomicin generated $800,000 in revenue. Ten months after FDA approval, Achaogen filed for bankruptcy. Its assets were auctioned for $16 million.
Nabriva's story is perhaps more poignant. Xenleta (lefamulin) was the first antibiotic with a novel mechanism of action for community-acquired pneumonia in nearly 20 years. It worked. It was approved. And it was commercially unviable. Nabriva terminated all employees in 2023.
The WHO's most recent pipeline analysis (February 2025) counts 90 antibacterials or combinations in clinical development. This is down from 97 in the previous analysis. Of these:
For comparison, the oncology pipeline contains over 2,500 clinical-stage drugs. The ratio is roughly 28 to 1. Oncology R&D investment exceeds $80 billion annually. The number of AMR researchers globally is estimated at approximately 3,000 — roughly ten times fewer than oncology researchers.
In 2024, four new antibiotics were approved: Exblifep (Allecra), Emblaveo (Pfizer), ORLYNVAH (Iterum), and Zevtera (Basilea). In early 2025, GSK's gepotidacin (Blujepa) became the first truly first-in-class antibiotic approved in years, for uncomplicated UTIs. These are meaningful — but they are drops against a tide.
The picture is not entirely bleak. Several genuinely novel approaches have emerged in the past decade, some of them now reaching clinical trials. Here is an honest assessment of each.
Phages — viruses that infect and kill bacteria — were discovered a century ago but sidelined when chemical antibiotics proved easier to manufacture. With antibiotics failing, phages are back.
The field's modern catalyst was the Tom Patterson case (2016). Patterson, a UC San Diego psychiatry professor, contracted a pan-resistant Acinetobacter baumannii infection in Egypt. When every antibiotic failed and he lapsed into a coma, his wife Steffanie Strathdee — an infectious disease epidemiologist — orchestrated emergency phage therapy using viruses sourced from the US Navy and Texas A&M. Patterson emerged from his coma within three days of treatment and eventually recovered fully.
The case led to the creation of the Center for Innovative Phage Applications and Therapeutics (IPATH) at UC San Diego. Today, over 60 interventional phage studies are listed on ClinicalTrials.gov.
BiomX reported positive Phase 2 results (March 2025) for BX211 in diabetic foot infections — statistically significant ulcer reduction with no serious adverse events. Locus Biosciences published positive Phase 2 results in The Lancet Infectious Diseases for a CRISPR-enhanced phage cocktail targeting E. coli UTIs.
Belgium's Queen Astrid Military Hospital operates under a unique framework allowing personalized phage therapy. Their review of 100 cases showed 77% clinical improvement. The critical finding: phage therapy works best alongside antibiotics, not as a replacement.
Honest assessment: Phage therapy is real and advancing. Its personalized nature — matching specific phages to specific bacterial strains — is both its strength and its regulatory obstacle. First FDA approval is plausible within 3–5 years.
In 2020, James Collins' group at MIT used a deep learning model to screen 107 million chemical compounds and identified halicin — the first antibiotic discovered by artificial intelligence. Halicin has broad-spectrum activity including against carbapenem-resistant bacteria, and works through a mechanism unlike any existing antibiotic.
In 2023, the same group found abaucin, a narrow-spectrum agent targeting Acinetobacter baumannii specifically. Later that year, they identified an entirely new structural class active against MRSA. In August 2025, they moved from screening existing molecules to designing new ones from scratch using generative AI — computationally screening over 36 million hypothetical compounds.
Critically, the 2023 MRSA work also used explainable AI to identify which chemical substructures drove antibacterial activity, addressing the "black box" criticism that has dogged AI drug discovery.
Honest assessment: AI has genuinely transformed the discovery phase, compressing years of screening into days. But discovery is not development. None of these AI-discovered compounds has entered human clinical trials. The bottleneck remains the same: preclinical toxicology, formulation, pharmacokinetics, Phase 1–3 trials, manufacturing. Timeline for the first AI-discovered antibiotic to reach patients: 7–10 years at minimum. The real value of AI may be in making the discovery pipeline continuously productive rather than in any single compound.
Teixobactin (2015): Discovered from unculturable soil bacteria using a device called the iChip. Kills gram-positive bacteria (MRSA, VRE) by binding two cell wall precursors simultaneously — targets so essential that bacteria have never evolved resistance to them. Still preclinical. No activity against gram-negatives.
Clovibactin (2023): A cousin of teixobactin discovered by the same method. Wraps around bacterial cell wall precursors like a cage. Active against MRSA and tuberculosis. Resistance frequency below 10-10. Still preclinical.
Novltex (September 2025): A synthetic platform from the University of Liverpool that solves the manufacturing problem. Inspired by teixobactin and clovibactin, Novltex replaces expensive building blocks with cheap threonine, achieving 30% synthesis yield with 10-minute coupling cycles. Preclinical.
Zosurabalpin (Roche): The most exciting candidate in the pipeline. A macrocyclic peptide that blocks lipopolysaccharide transport in Acinetobacter baumannii — described as the first new mechanism targeting gram-negative bacteria since 1968. Phase 1 showed good tolerability. Roche is launching a Phase 3 trial of approximately 400 patients with CRAB infections across Europe, the Americas, and Asia.
Honest assessment: Zosurabalpin is the headline. It addresses the WHO's #1 priority pathogen through a genuinely new mechanism, and Roche has the resources to push it through trials. Possible approval: 2028–2030. The teixobactin/clovibactin family is scientifically compelling but years behind.
Instead of poisoning bacteria with chemicals, CRISPR antimicrobials use phages as delivery vehicles for gene-editing systems that cut specific DNA sequences — either resistance genes on plasmids (disarming the bacteria) or essential chromosomal genes (killing them).
Locus Biosciences uses CRISPR-Cas3 (which degrades DNA processively, unlike Cas9's single cut) engineered into phages. Their LBP-EC01, targeting E. coli UTIs, published positive Phase 2 results in The Lancet Infectious Diseases. A Phase 2/3 trial is underway. Backed by $24M from BARDA and $35M from Johnson & Johnson.
SNIPR Biome (Copenhagen) takes a precision approach: their SNIPR001 selectively eliminates drug-resistant E. coli from the gut to prevent bloodstream infections in immunosuppressed cancer patients. Phase 1 showed E. coli reduction in 24 of 36 participants. Now in Phase 1b/2a.
Honest assessment: CRISPR antimicrobials are genuinely novel. Delivery remains the central challenge — getting CRISPR systems to bacterial cells inside complex microbial communities within the body. First approval: possible 2027–2029 for narrow indications like UTIs or gut decolonization.
The unglamorous workhorses of the resistance fight. Beta-lactamase inhibitor combinations pair existing antibiotics with molecules that neutralize the enzymes bacteria use to destroy them. Three important new combinations were approved in 2023–2024:
Honest assessment: These are available now and helping patients today. But they are playing whack-a-mole — each new combination addresses specific resistance enzymes, and bacteria evolve new ones. Efflux pump inhibitors, which could be more broadly useful, have been a graveyard of failed programs despite decades of research.
Two fecal microbiota products have been approved for recurrent Clostridioides difficile infection: REBYOTA (November 2022, rectal administration) and VOWST (April 2023, the first oral product). In the ECOSPOR III trial, 88% of VOWST patients were recurrence-free at 8 weeks versus 60% on placebo.
The 2024 AGA Practice Guideline now recommends fecal microbiota transplantation for the majority of recurrent C. diff patients. Recent 2025 evidence showed FMT is non-inferior to vancomycin even for initial (not just recurrent) infection.
Honest assessment: FMT for C. diff is a genuine success story. Expansion to other indications (resistant organism decolonization, potentially IBD) is 3–5 years out. Engineered probiotics, meanwhile, have been disappointing — Synlogic, the leading company in the space, discontinued its programs after clinical failures.
Everyone agrees the market is broken. Solutions exist. They have not been implemented.
The most important piece of antibiotic legislation in the United States is the PASTEUR Act (Pioneering Antimicrobial Subscriptions To End Upsurging Resistance). It would create a "subscription model" for antibiotics: the federal government would enter fixed-price contracts with manufacturers, paying $75–300 million per year per qualifying drug regardless of volume used. Total authorization: $6 billion.
The PASTEUR Act was first introduced in 2020. It has been reintroduced in every subsequent Congress, most recently in February 2026 (PASTEUR v2.0) with bipartisan sponsors.
It has never received a floor vote in either chamber.
The UK has gone furthest in actually testing subscription-style payments. In 2022–2023, the NHS piloted the world's first antibiotic subscription model, paying fixed annual fees for ceftazidime-avibactam and cefiderocol regardless of use. In August 2024, the program expanded: the NHS tendered contracts worth an estimated £1.9 billion over 16 years, with an annual budget of £100 million for subscription contracts.
Sweden ran a parallel pilot from 2020 to 2024, guaranteeing revenue for two antibiotics. The payments were much smaller: approximately SEK 4 million (~$380,000) per product per year.
The UK's £10–20 million per drug per year is an important proof of concept. But analyses suggest that a positive return on antibiotic investment requires payments on the order of $1–2.6 billion per drug over 10 years. The UK model is a start; it is not a solution at scale.
In December 2025, the EU agreed to implement transferable exclusivity vouchers (TEVs) for priority antimicrobials: a 12-month extension to data exclusivity on another drug, transferable or sellable. The idea is that the antibiotic subsidizes itself through delayed generic competition on a different product. Critics, including The Lancet, argue this is an indirect, expensive subsidy paid by patients and health systems through higher drug prices elsewhere — estimated at €1–4 billion per voucher.
Government and philanthropic organizations are providing push funding:
These amounts matter, but they fund discovery and development. They do not fix the commercial model. You can push a drug through trials with public money, but if the company that manufactured it goes bankrupt the day after approval, you have not solved the problem.
The GRAM study modeled two scenarios for the period 2025–2050:
39.1 million direct AMR deaths
169 million associated deaths
AMR deaths in over-70s rise 146%
Healthcare costs: $159 billion/year by 2050
92 million lives saved
New gram-negative antibiotics alone save 11 million
Improved healthcare access saves most lives in South Asia and Africa
Returns $7–13 per $1 invested
The intervention scenario requires three things:
The honest timeline for the most promising new approaches:
| Approach | Stage | Timeline to Impact | Confidence |
|---|---|---|---|
| Resistance breaker combinations | Approved | Now | High |
| Microbiome/FMT (C. diff) | Approved | Now | High |
| Phage therapy | Phase 2 | 3–5 years | Moderate-High |
| Novel classes (zosurabalpin) | Phase 1–3 | 3–5 years | Moderate |
| CRISPR antimicrobials | Phase 1–2 | 3–5 years | Moderate |
| AI-discovered antibiotics | Preclinical | 7–10 years | Moderate |
| AMR vaccines | Phase 1 | 5–10+ years | Low |
Antibiotic resistance is not a future threat. It is a current emergency that kills more people than HIV/AIDS and malaria combined. The science to fight it exists — phages, CRISPR, AI, novel classes, combination therapies — but the economics to develop it does not. Every company that successfully brought a new antibiotic to patients went bankrupt for the trouble.
The paradox at the heart of this crisis is simple: the better an antibiotic works, the less it is used; the less it is used, the less money it makes; the less money it makes, the less likely anyone is to develop the next one. Breaking this cycle requires a policy intervention that treats antibiotics as public goods — like national defense or clean water — rather than as commercial products.
The science is advancing. The policy is not. And every year that the policy lags behind the science, another 1.27 million people die from infections that Alexander Fleming's discovery was supposed to have conquered.