Explainer
How new antibiotics are developed, and why so few reach patients
WHO counts 90 antibacterials in clinical development, down from 97 in 2023, and says only 15 are innovative.[1][2] The core problem is economic: new antibiotics are held in reserve to slow resistance, so sales-based returns are small.[3] Governments and funders are testing up-front research grants and subscription-style payments, while AI has found candidates such as halicin and abaucin that have so far been tested only in lab and animal studies.[4][5][6][7]
This page explains research and policy. It is not medical advice and says nothing about which antibiotic anyone should take; talk to a clinician about treatment. For how resistance arises, read antibiotic resistance explained first. New approvals and policy changes are logged in the antimicrobial resistance tracker.
WHO says the world faces an antimicrobial research and development crisis, with few new medicines in the pipeline.[8] The GRAM study estimates that new antibiotics against Gram-negative bacteria could avert 11.08 million AMR-attributable deaths between 2025 and 2050.[9] This page covers why the pipeline is thin, the money tools meant to fix it, and what AI has added.
Why the pipeline is thin
WHO counts 90 antibacterial treatments in clinical trials, down from 97 two years earlier.[1] Only 15 are innovative, and only 5 work against at least one of the bacteria WHO ranks most dangerous.[2] Since July 2017, 17 new antibacterials against priority bacteria have been approved, but only two belong to a new chemical class.[10]
WHO’s Analysis of antibacterial agents in clinical and preclinical development 2025 found the clinical pipeline fell from 97 agents in 2023 to 90, of which 50 are traditional small-molecule antibacterials and 40 non-traditional (bacteriophages, antibodies, microbiome-modulating agents).[1] Only 15 meet WHO’s innovation criteria and only 5 are active against at least one “critical” Bacterial Priority Pathogens List organism.[2] Of 17 agents authorised against priority pathogens since July 2017, two represent a new chemical class.[10] The preclinical pipeline holds 232 programmes across 148 groups.[11]
Most of this work is done by small companies: WHO says 90% of the firms in preclinical antibacterial research have fewer than 50 employees, which makes the field fragile.[11] In March 2026 WHO said few clinical candidates target priority pathogens and even fewer are innovative.[12]
The market failure
Antibiotics are taken for short courses, and new ones are deliberately used sparingly and kept for infections that resist older drugs.[3] Resistance to a new antibiotic can also appear within a few years of launch, so the chance of long-term revenue is slim.[3] The fixes on offer try to separate a developer’s income from the number of doses sold, an idea known as delinkage.[5]
Push incentives: paying for research up front
CARB-X is a global non-profit partnership headquartered at Boston University that backs early antibacterial products from basic research up to first-in-human clinical trials.[4] It is funded by a consortium of governments and foundations.[13] As of November 2026 its website reports $506 million invested in 124 projects.[14] WHO has also published target product profiles that set out what new antibiotics for severe resistant infections and bacterial meningitis should look like.[15]
Pull incentives: paying for a finished drug
In the UK, a pilot launched in July 2019 pays companies for access to antibiotics based on their value to the NHS rather than how much is used.[16] The first two products were cefiderocol and ceftazidime–avibactam.[17] On 8 May 2024 the UK made the model permanent, with initial three-year contracts that can be extended.[18]
In the United States, the proposed PASTEUR Act would let the federal government sign subscription-style contracts with fixed annual payments, regardless of how often the drugs are used.[19] The bill was first introduced in 2020 and has never had a floor vote.[20] Five House members reintroduced an updated version in February 2026.[21]
AI-discovered antibiotics
Machine-learning models can screen huge numbers of molecules quickly.[22] The results so far come from lab and mouse studies, not human trials:[6][23]
| Compound | Year | What the AI did | Evidence so far |
|---|---|---|---|
| Halicin | 2020 | Picked out a former diabetes-drug candidate that killed many resistant bacteria in lab tests | Cleared infections in two mouse models |
| Abaucin | 2023 | Picked out a narrow-spectrum killer of Acinetobacter baumannii from nearly 7,000 compounds | Treated wound infections in mice |
| NG1 and DN1 | 2025 | Designed new molecules from more than 36 million generated compounds | Cleared resistant gonorrhea and MRSA infections in mice |
Sources for the table: halicin,[24][6] abaucin[25][7] and NG1 and DN1.[26][23]
In the halicin experiments, E. coli did not become resistant to halicin over 30 days, while it became about 200 times more resistant to ciprofloxacin.[27] The same model screened more than 100 million molecules in three days.[22] Abaucin appears to work by blocking a protein called LolE, which bacteria use to move lipoproteins.[28] Its narrow spectrum may lower the risk of resistance spreading.[7]
Of 80 AI-designed gonorrhea candidates the MIT team tried to have made, only two could be synthesised.[29] The nonprofit Phare Bio is now modifying NG1 and DN1 for further testing.[23] Across all diseases, few AI-discovered drugs have reached human trials.[30]
For the wider picture of AI in drug design, read how AI designs new proteins, drugs and materials and the evidence on AI-designed drugs. We did not cover antifungal or tuberculosis drug development in detail.
Read next
Next in the crash course
Antimicrobial resistance tracker 2025-26: data, drugs and policy
A dated log of antimicrobial resistance news in 2025-26: WHO surveillance data, new antibiotic approvals, the 2026-2036 action plan and the Abuja meeting.
Questions readers ask
How many new antibiotics are in development?
WHO's 2025 analysis counted 90 antibacterials in clinical development, down from 97 in 2023. Fifty are traditional antibacterial agents and 40 are non-traditional approaches such as bacteriophages and antibodies.[1]
Why don't drug companies make more antibiotics?
New antibiotics are used sparingly and kept in reserve for resistant infections, and resistance can appear within a few years of launch, so the prospect of long-term sales revenue is slim.[3]
What is the UK antibiotic subscription model?
Since a 2019 pilot, the NHS has paid companies for access to selected antibiotics based on their value to the NHS rather than how much is used. In May 2024 the UK made the scheme permanent.[16][18]
What is the PASTEUR Act?
A proposed US law that would let the federal government pay developers of critically needed antibiotics and antifungals fixed annual amounts regardless of use. It was first introduced in 2020 and reintroduced in February 2026, but has never had a floor vote.[19][20][21]
Sources
· 30 claims from 11 publishers
Each numbered claim is a statement we checked against the sources listed with it. Tap a number in the text to see its sources in place.
- [1]
WHO's 2025 pipeline analysis found the number of antibacterials in clinical development fell from 97 in 2023 to 90 in 2025, of which 50 are traditional antibacterial agents and 40 are non-traditional approaches such as bacteriophages and antibodies. confirmedas of 2025-10-02
- WHO releases new reports on new tests and treatments in development for bacterial infections · World Health Organization · Primary source · 2025-10-02 (retrieved 2026-11-07)
- [2]
Of the 90 antibacterials in clinical development, WHO counts only 15 as innovative and only 5 as effective against at least one of its "critical" priority bacteria. confirmedas of 2025-10-02
- WHO releases new reports on new tests and treatments in development for bacterial infections · World Health Organization · Primary source · 2025-10-02 (retrieved 2026-11-07)
- WHO releases new reports on new tests and treatments in development for bacterial infections · World Health Organization · Primary source · 2025-10-02 (retrieved 2026-11-07)
- [3]
New antibiotics are used sparingly and held in reserve for infections resistant to older drugs to slow resistance, so developers have little prospect of long-term sales revenue under payment based on volume. confirmedas of 2026-11-07
- Updated PASTEUR Act reintroduced in Congress to boost antibiotic development · CIDRAP (University of Minnesota) · Secondary source · 2026-02-09 (retrieved 2026-11-07)
- UK's antibiotic subscription funding model - a route to market sustainability? · Pharmaceutical Technology · Secondary source · 2024-07-01 (retrieved 2026-11-07)
- [4]
CARB-X is a global non-profit partnership, headquartered at Boston University, that funds early development of new antibiotics, preventatives and rapid diagnostics, from basic research up to first-in-human clinical trials. confirmedas of 2026-11-07
- CARB-X overview · CARB-X · Primary source (retrieved 2026-11-07)
- CARB-X overview · CARB-X · Primary source (retrieved 2026-11-07)
- CARB-X overview · CARB-X · Primary source (retrieved 2026-11-07)
- CARB-X overview · CARB-X · Primary source · Overview (retrieved 2026-11-07)
- [5]
Subscription-style payment models aim to delink developers' revenue from the volume of antibiotics sold. confirmedas of 2026-11-07
- Updated PASTEUR Act reintroduced in Congress to boost antibiotic development · CIDRAP (University of Minnesota) · Secondary source · 2026-02-09 (retrieved 2026-11-07)
- UK's antibiotic subscription funding model - a route to market sustainability? · Pharmaceutical Technology · Secondary source · 2024-07-01 (retrieved 2026-11-07)
- [6]
Halicin cleared infections in two mouse models; it has been tested in animals, not in human trials. confirmedas of 2020-02-20
- Artificial intelligence yields new antibiotic · MIT News · Primary source · 2020-02-20 (retrieved 2026-11-07)
- Artificial intelligence yields new antibiotic · MIT News · Primary source · 2020-02-20 (retrieved 2026-11-07)
- [7]
Abaucin has a narrow spectrum, which the researchers say lowers the risk of resistance spreading and would likely spare beneficial gut bacteria; it treated wound infections in mice. confirmedas of 2023-05-25
- Using AI, scientists find a drug that could combat drug-resistant infections · MIT News · Primary source · 2023-05-25 (retrieved 2026-11-07)
- Using AI, scientists find a drug that could combat drug-resistant infections · MIT News · Primary source · 2023-05-25 (retrieved 2026-11-07)
- [8]
WHO says the world faces an antimicrobial research and development crisis, with few new medicines in the pipeline. confirmedas of 2026-11-07
- Antimicrobial resistance (fact sheet) · World Health Organization · Primary source · 2026-07-16 (retrieved 2026-11-07)
- [9]
In GRAM's scenarios, better infection care and access to antibiotics could avert 92 million deaths between 2025 and 2050, and new antibiotics against Gram-negative bacteria could avert 11.08 million AMR-attributable deaths. confirmedas of 2024-09-16· forecast
- The Lancet: More than 39 million deaths from antibiotic-resistant infections estimated between now and 2050, suggests first global analysis · Institute for Health Metrics and Evaluation (GRAM study authors) · Primary source · 2024-09-16 (retrieved 2026-11-07)
- The Lancet: More than 39 million deaths from antibiotic-resistant infections estimated between now and 2050, suggests first global analysis · Institute for Health Metrics and Evaluation (GRAM study authors) · Primary source · 2024-09-16 (retrieved 2026-11-07)
- [10]
WHO counts 17 new antibacterial agents against priority pathogens authorised since July 2017, of which only two represent a new chemical class. confirmedas of 2025-10-02
- WHO releases new reports on new tests and treatments in development for bacterial infections · World Health Organization · Primary source · 2025-10-02 (retrieved 2026-11-07)
- [11]
WHO counts 232 preclinical antibacterial programmes across 148 groups, and says 90% of the companies involved are small firms with fewer than 50 employees, which makes the research ecosystem fragile. confirmedas of 2025-10-02
- WHO releases new reports on new tests and treatments in development for bacterial infections · World Health Organization · Primary source · 2025-10-02 (retrieved 2026-11-07)
- [12]
WHO said in March 2026 that, despite 90 new antibacterial agents in development, few clinical candidates target priority pathogens and even fewer are innovative. confirmedas of 2026-03-11
- WHO releases new target product profiles for urgently needed antibiotics · World Health Organization · Primary source · 2026-03-11 (retrieved 2026-11-07)
- [13]
CARB-X says it is funded by a global consortium of governments and foundations. confirmedas of 2026-11-07
- CARB-X overview · CARB-X · Primary source (retrieved 2026-11-07)
- [14]
As of November 2026, CARB-X's website reports $506 million invested in 124 projects. confirmedas of 2026-11-07
- CARB-X home page · CARB-X · Primary source · Statistics panel on the home page (retrieved 2026-11-07)
- [15]
In March 2026 WHO published target product profiles for new antibiotics against severe multidrug-resistant Gram-negative infections, resistant Gram-positive infections in immunosuppressed and critically ill patients, and bacterial meningitis. confirmedas of 2026-03-11
- WHO releases new target product profiles for urgently needed antibiotics · World Health Organization · Primary source · 2026-03-11 (retrieved 2026-11-07)
- [16]
The UK's antimicrobial payment model, launched as a pilot in July 2019, pays companies for access to antibiotics based on a product's value to the NHS rather than on how much is used. confirmedas of 2026-11-07
- World-first scheme underway to tackle AMR and protect UK patients · UK Government (GOV.UK) · Primary source (retrieved 2026-11-07)
- UK's antibiotic subscription funding model - a route to market sustainability? · Pharmaceutical Technology · Secondary source · 2024-07-01 (retrieved 2026-11-07)
- [17]
The first two products in the UK subscription scheme were cefiderocol (Fetcroja, Shionogi) and ceftazidime–avibactam (Zavicefta, Pfizer); the Pharmaceutical Journal reports both contracts were awarded in July 2022. confirmedas of 2024-08-15
- Expanded 'subscription model' for new antibiotics to be funded with £100m per year · The Pharmaceutical Journal · Secondary source · 2024-08-15 (retrieved 2026-11-07)
- UK's antibiotic subscription funding model - a route to market sustainability? · Pharmaceutical Technology · Secondary source · 2024-07-01 (retrieved 2026-11-07)
- [18]
On 8 May 2024 the UK government moved the subscription model from a pilot to a permanent scheme open to more products, with initial three-year contracts that can be extended. confirmedas of 2024-08-15
- UK's antibiotic subscription funding model - a route to market sustainability? · Pharmaceutical Technology · Secondary source · 2024-07-01 (retrieved 2026-11-07)
- Expanded 'subscription model' for new antibiotics to be funded with £100m per year · The Pharmaceutical Journal · Secondary source · 2024-08-15 (retrieved 2026-11-07)
- UK's antibiotic subscription funding model - a route to market sustainability? · Pharmaceutical Technology · Secondary source · 2024-07-01 · Article body (retrieved 2026-11-07)
- [19]
The proposed US PASTEUR Act would let the federal government sign subscription-style contracts paying developers of critically needed antibiotics and antifungals fixed annual amounts regardless of how often the drugs are used. confirmedas of 2026-02-09
- Updated PASTEUR Act reintroduced in Congress to boost antibiotic development · CIDRAP (University of Minnesota) · Secondary source · 2026-02-09 (retrieved 2026-11-07)
- PASTEUR Act Reintroduced in Congress · Contagion Live · Secondary source · 2026-02-05 (retrieved 2026-11-07)
- [20]
The PASTEUR Act was first introduced in September 2020 and reintroduced in 2021 and April 2023, but has never received a floor vote. confirmedas of 2026-02-09
- PASTEUR Act Reintroduced in Congress · Contagion Live · Secondary source · 2026-02-05 (retrieved 2026-11-07)
- Updated PASTEUR Act reintroduced in Congress to boost antibiotic development · CIDRAP (University of Minnesota) · Secondary source · 2026-02-09 (retrieved 2026-11-07)
- Reps Peters & Levin, Sens Bennet & Young, and Others Reintroduce Bipartisan PASTEUR Act to Fight Antimicrobial Resistance · Office of Rep. Scott Peters (US House of Representatives) · Primary source · 2023-04-27 · Press release dated April 27, 2023 (retrieved 2026-11-07)
- PASTEUR Act Reintroduced in Congress · Contagion Live · Secondary source · 2026-02-05 · History of the bill (retrieved 2026-11-07)
- [21]
In early February 2026 five members of the US House, led by Buddy Carter with Scott Peters, Nick Langworthy, Mike Levin and Mike Carey, reintroduced an updated PASTEUR Act, its fourth introduction. confirmedas of 2026-02-09
- PASTEUR Act Reintroduced in Congress · Contagion Live · Secondary source · 2026-02-05 (retrieved 2026-11-07)
- Updated PASTEUR Act reintroduced in Congress to boost antibiotic development · CIDRAP (University of Minnesota) · Secondary source · 2026-02-09 (retrieved 2026-11-07)
- [22]
The same model screened more than 100 million molecules in three days and flagged 23 candidates structurally unlike existing antibiotics. confirmedas of 2020-02-20
- Artificial intelligence yields new antibiotic · MIT News · Primary source · 2020-02-20 (retrieved 2026-11-07)
- Artificial intelligence yields new antibiotic · MIT News · Primary source · 2020-02-20 (retrieved 2026-11-07)
- [23]
Two lead compounds from the 2025 MIT work, NG1 and DN1, cleared drug-resistant gonorrhea and MRSA infections respectively in mouse models, and the nonprofit Phare Bio is modifying them for further testing. confirmedas of 2025-08-14
- Using generative AI, researchers design compounds that can kill drug-resistant bacteria · MIT News · Primary source · 2025-08-14 (retrieved 2026-11-07)
- Using generative AI, researchers design compounds that can kill drug-resistant bacteria · MIT News · Primary source · 2025-08-14 (retrieved 2026-11-07)
- Using generative AI, researchers design compounds that can kill drug-resistant bacteria · MIT News · Primary source · 2025-08-14 (retrieved 2026-11-07)
- [24]
In 2020 MIT researchers used a machine-learning model to identify halicin, a compound previously investigated as a possible diabetes drug, which in lab tests killed many problematic bacteria, including strains resistant to all known antibiotics. confirmedas of 2020-02-20
- Artificial intelligence yields new antibiotic · MIT News · Primary source · 2020-02-20 (retrieved 2026-11-07)
- Artificial intelligence yields new antibiotic · MIT News · Primary source · 2020-02-20 (retrieved 2026-11-07)
- [25]
In 2023 researchers at MIT and McMaster University used a machine-learning model to pick out abaucin, from a library of nearly 7,000 compounds, as an antibiotic that kills Acinetobacter baumannii. confirmedas of 2023-05-25
- Using AI, scientists find a drug that could combat drug-resistant infections · MIT News · Primary source · 2023-05-25 (retrieved 2026-11-07)
- Using AI, scientists find a drug that could combat drug-resistant infections · MIT News · Primary source · 2023-05-25 (retrieved 2026-11-07)
- [26]
In 2025 an MIT team used generative AI to design more than 36 million possible compounds; the top candidates were structurally distinct from existing antibiotics and appear to disrupt bacterial cell membranes. confirmedas of 2025-08-14
- Using generative AI, researchers design compounds that can kill drug-resistant bacteria · MIT News · Primary source · 2025-08-14 (retrieved 2026-11-07)
- [27]
In the MIT experiments, E. coli did not develop resistance to halicin over 30 days, while it became about 200 times more resistant to ciprofloxacin. confirmedas of 2020-02-20
- Artificial intelligence yields new antibiotic · MIT News · Primary source · 2020-02-20 (retrieved 2026-11-07)
- Artificial intelligence yields new antibiotic · MIT News · Primary source · 2020-02-20 (retrieved 2026-11-07)
- [28]
Abaucin appears to kill cells by interfering with lipoprotein trafficking, by inhibiting a protein called LolE. confirmedas of 2023-05-25
- Using AI, scientists find a drug that could combat drug-resistant infections · MIT News · Primary source · 2023-05-25 (retrieved 2026-11-07)
- Using AI, scientists find a drug that could combat drug-resistant infections · MIT News · Primary source · 2023-05-25 (retrieved 2026-11-07)
- [29]
Of 80 AI-designed gonorrhea candidates the MIT team tried to have made, only two could be synthesised. confirmedas of 2025-08-14
- Using generative AI, researchers design compounds that can kill drug-resistant bacteria · MIT News · Primary source · 2025-08-14 (retrieved 2026-11-07)
- [30]
The rentosertib paper notes that few novel AI-discovered or AI-designed drugs have reached human clinical trials. confirmedas of 2025-06-03
- A generative AI-discovered TNIK inhibitor for idiopathic pulmonary fibrosis: a randomized phase 2a trial · Nature Medicine · Primary source · 2025-06-03 (retrieved 2026-10-10)
Revision history (1)
- Page created.
Created Nov 7, 2026. Last reviewed by an editor on Nov 8, 2026. Next scheduled review: Mar 7, 2027.
Cite this page
"How new antibiotics are developed, and why so few reach patients." ContentLora, updated Nov 7, 2026. https://contentlora.com/explain/how-new-antibiotics-are-developed
Spotted an error? Suggest a correction or emailcorrections@contentlora.com.