In June 2018, the US Food and Drug Administration approved plazomicin \u2014 brand name Zemdri \u2014 for complicated urinary tract infections caused by carbapenem-resistant Enterobacteriaceae. Carbapenem-resistant organisms are the kind of infections that kill people in hospitals when nothing else works. The FDA reviewers knew this. So did the clinicians. Plazomicin was effective against strains that would otherwise have been nearly untreatable, and it had completed the full clinical trial gauntlet.<\/p>\n\n\n\n
In the twelve months after approval, it generated approximately $800,000 in sales.<\/p>\n\n\n\n
Achaogen, the company that developed it, filed for Chapter 11 bankruptcy in April 2019. Eleven months after FDA approval. The company was destroyed not by clinical failure, not by regulatory rejection, but by the commercial logic of having made exactly the right drug for exactly the right indication and then having hospitals hold it in reserve, as they should, for patients with no other options.<\/p>\n\n\n\n
Achaogen was not unlucky. Its commercial failure after clinical approval was the predictable output of three interlocking systems, each with its own logic, each pointing in the wrong direction.<\/p>\n\n\n\n
The first thing to understand about antibiotic resistance is that it isn’t a problem. It’s a condition.<\/p>\n\n\n\n
Bacteria reproduce in approximately twenty minutes. Escherichia coli<\/em>, the bacterium found in every human gut and on the fingers of food workers and in the blood of patients with sepsis, generates a new population roughly seventy times per day. In any population of a trillion organisms \u2014 a normal figure for a bacterial colony \u2014 if resistance mutations arise once per hundred million replications, that mutation appears approximately ten thousand times in a single generation. Each resistant organism is then exposed to an antibiotic that kills its susceptible neighbors. The resistant variants reproduce unopposed.<\/p>\n\n\n\n This is not misbehavior. It’s arithmetic.<\/p>\n\n\n\n The mechanism accelerates through horizontal gene transfer: bacteria can pass resistance genes between organisms without reproduction, across species lines, across genera. A resistance gene that evolved in livestock bacteria can turn up in a human pathogen in a patient who has never been near a farm. Conjugation, via direct cell-to-cell contact and plasmid transfer, is the primary route. But transformation \u2014 uptake of free DNA from the environment \u2014 and transduction, mediated by bacteriophages, add further pathways. The result is a global genetic commons where resistance information flows freely among bacterial populations with no regard for taxonomic boundaries.<\/p>\n\n\n\n What makes resistance historically anomalous is the pace at which human antibiotic use has compressed what would otherwise be geological timescales of selection. D’Costa et al., writing in Nature<\/em> in 2011, recovered resistance genes from 30,000-year-old Beringian permafrost \u2014 genes encoding resistance to beta-lactam, tetracycline, and glycopeptide antibiotics, including a vancomycin resistance element (VanA) functionally similar to modern variants. Resistance is ancient. The novelty is that we have spent eighty years applying selection pressure at industrial scale.<\/p>\n\n\n\n Alexander Fleming understood the mechanism before it became a clinical problem. In his 1945 Nobel Prize lecture, he said: “The danger that the ignorant man may easily underdose himself and by exposing his microbes to non-lethal quantities of the drug make them resistant.” He was describing exactly the condition under which resistant mutants are selected: enough antibiotic to kill susceptible bacteria, not enough to kill the resistant ones. The resistant variants inherit the world.<\/p>\n\n\n\n Penicillin was introduced for mass clinical use in 1943-1944. Penicillin-resistant Staphylococcus aureus<\/em> strains were documented in hospitals by 1947 \u2014 Mary Barber at Hammersmith Hospital in London published on hospital-wide outbreaks of resistant staph that year. By the late 1940s, penicillin resistance in S. aureus<\/em> had become widespread, and nosocomial epidemics accelerated through the 1950s and into the 1960s. The interval between mass production and documented hospital resistance: roughly three years.<\/p>\n\n\n\n The question, then, is not how to prevent resistance. Selection pressure is the cost of using antibiotics at all. The question is how to slow the accumulation of resistance while maintaining a pipeline of new treatments sufficient to stay ahead of it. On both counts, the current situation is failing. But for reasons that are structural, not moral.<\/p>\n\n\n\n Agriculture is where the volume is.<\/p>\n\n\n\n Approximately 73% of global antibiotic consumption by mass is in food animals. The Van Boeckel et al. study in PNAS<\/em> in 2015 estimated total veterinary antibiotic consumption in 2010 at 63,151 tonnes globally, against approximately 40,000 tonnes for human use. In the United States, according to FDA sales data, roughly 70% of medically important antibiotics sold are for animal use.<\/p>\n\n\n\n The dominant uses in agriculture are growth promotion and mass prophylaxis. Growth promotion \u2014 administering sub-therapeutic doses to healthy animals \u2014 increases feed efficiency and weight gain through mechanisms that remain incompletely understood, probably involving changes to gut microbiota. Mass prophylaxis means treating entire crowded herds or flocks with antibiotics to prevent infection in conditions where individual disease spread is near-certain. Neither use is therapeutic in any meaningful sense; neither targets a specific pathogen in a specific animal. Both involve administering antibiotics at doses chosen for economic or husbandry purposes rather than microbiological ones.<\/p>\n\n\n\n Denmark ran the experiment. In 1998, Denmark banned antibiotic growth promotion in chickens; pigs followed in 1999. Total antibiotic use in livestock fell from approximately 206 tonnes in the early 1990s to 81 tonnes after the ban \u2014 a reduction of more than 60%. Over the same period, Danish swine production increased by nearly 50% relative to 1992 levels. The cost increase per pig from birth to slaughter was approximately one euro. The World Health Organization assessed the Danish ban and found no significant harm to animal health or farmer income. What Denmark showed is that the growth promotion use is not biologically essential; it’s economically convenient.<\/p>\n\n\n\n The United States enacted its Veterinary Feed Directive rule on January 1, 2017. Eighteen years after Denmark. The VFD rule eliminated over-the-counter sale of medically important antibiotics for use in animal feed without veterinary oversight, but it leaves substantial room for “therapeutic” use that functions in practice as prophylaxis for chronically overcrowded conditions. The gap between the Danish evidence and the American regulatory response is forty-eight years from first concern. The Swann Report \u2014 formally, the Report of the Joint Committee on the Use of Antibiotics in Animal Husbandry and Veterinary Medicine<\/em> \u2014 was published in November 1969 and recommended restricting penicillin and tetracyclines in animal feed. Its core logic was accepted in 1969 and acted on in the US in 2017.<\/p>\n\n\n\n Global veterinary antibiotic use is projected to continue growing. A 2025 Nature Communications<\/em> study by Van Boeckel and colleagues, extending projections through 2040, estimated a 29.5% increase in global antibiotic use from a 2019 baseline of roughly 110,777 tonnes, driven primarily by expanding meat production in middle-income countries.<\/p>\n\n\n\n The political arithmetic is straightforward. Agriculture constitutes a large, organized, economically and politically significant industry in most countries. Cross-border coordination is required \u2014 unilateral restriction creates economic disadvantage for complying producers, as Denmark’s producers understood when they accepted the one-euro premium. And in lower-income agricultural economies, where livestock represent a primary protein source and farm incomes are thin, the economic stakes are more complex than in Denmark. None of this is irrational. It’s just slow.<\/p>\n\n\n\n Fleming named the mechanism in 1945. The dominant global antibiotic use pattern \u2014 sub-therapeutic doses, mass medication, measured in tens of thousands of tonnes \u2014 still largely conforms to it. Bacterial generations are twenty minutes. The regulatory pace has been measured in decades.<\/p>\n\n\n\n Developing a new antibiotic costs approximately $1.3 billion all-in, including the cost of capital, failed trials, and regulatory requirements \u2014 a figure derived from analysis published by AMR.Solutions in January 2021. The timeline runs ten to fifteen years from early-stage research to approval. These figures are comparable to oncology drug development costs.<\/p>\n\n\n\n The commercial profiles are not comparable.<\/p>\n\n\n\n A cancer drug approved for, say, metastatic non-small-cell lung cancer will be used as first- or second-line treatment, prescribed chronically for years, priced at $50,000 to $200,000 per patient per year, and have a potential market of every patient diagnosed with that cancer globally. A new antibiotic against carbapenem-resistant organisms will be held in reserve \u2014 correctly \u2014 until all other options have failed. Clinicians are trained to steward it, meaning to not use it until they must. A course of treatment is seven to fourteen days. Payer expectations, shaped by decades of cheap generics, keep antibiotic prices at commodity levels. The population of patients who will receive it in any given year is small by design.<\/p>\n\n\n\n Kevin Outterson, Executive Director of CARB-X, said it at the 2020 World Economic Forum: “If you invent a mediocre antibiotic, your sales will be low; if you develop a great antibiotic, your sales will be lower.” It is not a paradox. It is a description of how the market works. Clinical excellence in an antibiotic means it’s held in reserve. Reserve means low volume. Low volume means low revenue. Low revenue means bankruptcy.<\/p>\n\n\n\n Achaogen’s plazomicin development, analyzed in a 2024 paper in Humanities and Social Sciences Communications<\/em>, involved raising approximately $770 million over fifteen years, with an estimated $560 million invested directly into plazomicin development. BARDA \u2014 the Biomedical Advanced Research and Development Authority \u2014 contributed $124.4 million of that total across multiple contract options beginning in 2010. Plazomicin received BARDA support throughout its development pipeline and was still commercially destroyed eleven months after FDA approval.<\/p>\n\n\n\n Public funding into development. Commercial destruction at market.<\/p>\n\n\n\n Achaogen was not the exception. Melinta Therapeutics brought four FDA-approved antibiotics to market: Baxdela (delafloxacin), Orbactiv (oritavancin), Vabomere (meropenem-vaborbactam), and Minocin (minocycline). Four approved drugs. $41 million in revenue for the first nine months of 2019. Approximately $1 billion in debt. Chapter 11 in December 2019. The Melinta case matters specifically because it eliminates the single-drug explanation. It wasn’t one flawed product or one botched launch \u2014 it was four clinically effective antibiotics, all approved, none generating revenue anywhere near sufficient to sustain the company that held them.<\/p>\n\n\n\n Tetraphase Pharmaceuticals developed Xerava (eravacycline), approved in 2018 for complicated intra-abdominal infections. Xerava generated $3.6 million in net sales for all of 2019. Tetraphase was acquired by La Jolla Pharmaceutical Company in 2020 for $43 million in upfront cash \u2014 against a peak market capitalization that had once been close to $1.8 billion. The asset hadn’t failed clinically. Its commercial profile had been dictated by the same structural logic: effective for drug-resistant infections, used only when necessary, generating revenue roughly proportional to “when necessary.”<\/p>\n\n\n\n The pipeline reflects these outcomes. The share of FDA-approved new drugs that were antibiotics fell from approximately 20% in 1980 to roughly 6% today. The large pharmaceutical companies that once sustained antibiotic R&D have largely left: Pfizer exited preclinical antibiotic research in 2011, AstraZeneca spun out its antibiotic assets in 2015, Novartis halted its antibiotic research in 2018, Sanofi followed the same year. The WHO’s 2024 analysis of the antibacterial pipeline described the preclinical stage as “dynamic and innovative, but also fragile” due to small team sizes and high company and product turnover \u2014 WHO’s own language, not a journalist’s characterization.<\/p>\n\n\n\n Which brings us to zoliflodacin. The FDA approved it on December 12, 2025, for uncomplicated urogenital gonorrhea. It’s a spiropyrimidinetrione \u2014 a genuinely new mechanism of action, the first bacterial topoisomerase inhibitor of its class. It was developed not by a pharmaceutical company but by GARDP, the Global Antibiotic Research and Development Partnership, a not-for-profit founded under the auspices of the Drugs for Neglected Diseases initiative.<\/p>\n\n\n\n One interpretation of zoliflodacin is that it proves the system eventually works \u2014 that patience and persistence can produce a genuinely new mechanism even without structural change. That reading is wrong. For nearly four decades, commercial pharmaceutical R&D produced no genuinely new antibiotic mechanism. When a new mechanism finally arrived, it came through a charitable organization specifically created because commercial incentives were insufficient. That’s not a workaround. It’s a data point about what the market produces when left to its own logic.<\/p>\n\n\n\n The mechanisms of market failure are precisely understood. So are the proposed corrections.<\/p>\n\n\n\n Antibiotic development incentives divide into push and pull. Push incentives fund the upfront cost of research \u2014 NIH grants, BARDA contracts, the CARB-X fund (a public-private partnership involving US BARDA, the UK government, and the Gates Foundation, which has committed over $600 million to early-stage antibiotic development since its 2016 launch). Push incentives exist and have been effective at moving compounds through development pipelines. Achaogen had $124.4 million in BARDA support and still went bankrupt eleven months after approval. Push doesn’t fix the launch problem.<\/p>\n\n\n\n Pull incentives work differently. They promise a downstream reward \u2014 a payment, a prize, an extended revenue stream \u2014 that is deliberately delinked from sales volume. The value of the antibiotic is recognized regardless of how many doses are dispensed.<\/p>\n\n\n\n The United Kingdom piloted the most concrete version. In July 2022, NHS England signed subscription contracts with Shionogi (for cefiderocol, brand name Fetcroja) and Pfizer (for ceftazidime-avibactam, brand name Zavicefta), paying each company \u00a310 million per year for ten years, regardless of how many doses NHS hospitals used. The subscriptions were explicitly modeled on delinked payment theory \u2014 the NHS was paying for the availability of the drug, not for the drug itself. The model was extended under the UK government’s “Confronting Antimicrobial Resistance 2024 to 2029” plan.<\/p>\n\n\n\n Outterson and colleagues have estimated that a pull incentive sufficient to restore commercial viability for antibiotic development would need to deliver approximately $220 to $480 million in average annual revenues over ten years \u2014 a “best estimate” of around $310 million per drug per year. The NHS contracts deliver \u00a310 million annually: roughly $13 million at current exchange rates.<\/p>\n\n\n\n The proof of concept is real. The scale is somewhere between one-twentieth and one-fortieth of what economists estimate is required to actually fix the market.<\/p>\n\n\n\n In the United States, the PASTEUR Act \u2014 Pioneering Antimicrobial Subscriptions to End Upsurging Resistance \u2014 was reintroduced in Congress in February 2026 as its latest iteration. The legislation would establish federal subscription contracts of $75 million to $300 million per drug per year, delinked from sales volume, funded through the Department of Health and Human Services, with a total program cost of approximately $6 billion. The Center for Global Development has estimated a return of $28 for every dollar spent when measured over a 30-year horizon, including reduced healthcare costs and prevented deaths. The PASTEUR Act has been introduced in various forms since 2020 without reaching a floor vote.<\/p>\n\n\n\n The EU has proposed transferable exclusivity vouchers \u2014 a mechanism allowing antibiotic developers to sell patent extensions on other, more profitable drugs. M\u00e9decins Sans Fronti\u00e8res has opposed TEVs on the grounds that they transfer development costs onto patients who need those other drugs. The criticism is structurally coherent: TEVs shift the subsidy burden rather than resolving the market failure that created it.<\/p>\n\n\n\n The reason none of this has been implemented at scale is itself a structural problem. Antibiotics are global public goods. A new antibiotic developed at US government expense, through PASTEUR Act subscriptions, is available \u2014 via import, via generic manufacture, via standard global medical supply chains \u2014 to patients in France, Germany, Japan, Bangladesh, and everywhere else. No single country captures the full value of its investment. The classic collective action logic of public goods applies: each nation has incentive to let others pay for development while benefiting from the result. Outterson and Kesselheim identified this dynamic directly in their 2010 Health Affairs<\/em> paper, “Fighting Antibiotic Resistance: Marrying New Financial Incentives to Meeting Public Health Goals,” and in their 2011 analysis in the Yale Journal of Health Policy Law and Ethics<\/em>. The scholarship describing the problem is over fifteen years old.<\/p>\n\n\n\n The PASTEUR Act would take ten to fifteen years to produce clinical drugs. Decisions made now determine what treatments exist for 2035’s infections. The Lancet GRAM study, published September 2024, projected that antimicrobial resistance will directly cause 39 million deaths between 2025 and 2050, with directly attributable deaths rising from 1.14 million in 2021 to 1.91 million per year by 2050. The study estimated that improved access and new drug development could avert 92 million of the deaths that would otherwise occur. These numbers come from systematic analysis across 22 pathogens, 84 pathogen-drug combinations, and 204 countries. They are not alarming rhetoric. They are arithmetic applied to current trajectories.<\/p>\n\n\n\n Return to Achaogen. In the opening: a drug that worked, a company that didn’t survive, and no obvious villain.<\/p>\n\n\n\n Achaogen’s failure was not bad luck. It was the predictable output of three structural systems, each well-understood, none fixed, each obstructing the fix for the others. The name for that is not a market failure or a regulatory failure or a political failure. It’s all three, operating simultaneously, in a configuration designed to self-reinforce.<\/p>\n\n\n\n Each of these failures is well-understood. Each has well-understood fixes. The same structural logic that creates the problem also blocks the solution.<\/p>\n\n\n\n The drugs needed to treat 2035’s resistant infections need to enter development now. There is no commercial incentive to develop them. The PASTEUR Act, which would create one, has not passed. Global agricultural antibiotic reform, which would slow the resistance clock, moves on political timescales that bear no relationship to biological ones.<\/p>\n\n\n\n Zoliflodacin \u2014 approved December 12, 2025, new mechanism, new class, developed by a not-for-profit \u2014 treats one bacterial infection. It took a workaround to produce it. The pipeline for everything else runs on the same market logic that destroyed Achaogen.<\/p>\n\n\n\nHow resistance jumps between species<\/strong>\n\nThe three mechanisms of horizontal gene transfer are conjugation, transformation, and transduction. In conjugation, two bacterial cells make direct contact and a plasmid \u2014 a small, circular piece of DNA \u2014 is copied from one cell to the other. That plasmid can carry multiple resistance genes simultaneously. Transformation occurs when bacteria take up free DNA from their environment, including DNA from dead organisms. Transduction occurs when bacteriophages (viruses that infect bacteria) accidentally incorporate bacterial DNA and transfer it to a new host. None of these mechanisms requires the recipient to be the same species as the donor. A resistance gene that evolved in Salmonella in a chicken barn can, through this pathway, appear in Klebsiella pneumoniae in an intensive care unit. The patient in that ICU may have never touched a farm animal in their life.<\/code><\/pre>\n\n\n\nThe largest reservoir<\/h3>\n\n\n\n
Why sub-therapeutic doses are the worst case<\/strong>\n\nThe relevant microbiological concept is the minimum inhibitory concentration (MIC): the lowest concentration of an antibiotic that prevents visible bacterial growth. Doses above MIC kill or halt susceptible bacteria. Doses below MIC apply selection pressure \u2014 they kill some susceptible bacteria but not all, creating conditions under which resistant mutants proliferate more rapidly than susceptible ones. Growth promotion and prophylactic mass medication routinely deliver sub-MIC concentrations to entire animal populations over extended periods. This is precisely the condition Fleming named in 1945: \"exposing microbes to non-lethal quantities of the drug.\" An antibiotic administered therapeutically to a sick patient for seven to ten days at doses above MIC is a far smaller evolutionary opportunity than months of sub-therapeutic exposure across hundreds of millions of animals. The agricultural use pattern is, from the bacteria's perspective, optimal for developing resistance.<\/code><\/pre>\n\n\n\nThe bankruptcy of success<\/h3>\n\n\n\n
What \"new antibiotic class\" actually means<\/strong>\n\nAntibiotic classes are defined by mechanism of action \u2014 specifically, what the drug does to the bacterial cell. Beta-lactams disrupt cell wall synthesis. Fluoroquinolones inhibit DNA gyrase and topoisomerase IV. Macrolides target the bacterial ribosome. A new compound within an existing class is a new drug but not a new weapon; resistant bacteria with existing resistance mechanisms may remain unaffected. A genuinely new class attacks a target bacteria hasn't encountered before, offering a fresh start against organisms resistant to everything else. The last genuinely novel antibiotic mechanism to emerge from commercial pharmaceutical research was oxazolidinones, first reported at scientific conferences in 1987 and approved as linezolid in 2000. Post-2000 approvals \u2014 lipopeptides (daptomycin, 2003), pleuromutilins (retapamulin, 2007), lipiarmycins (fidaxomicin, 2011) \u2014 derived from scaffolds identified decades earlier. The FDA approved zoliflodacin (Nuzolvence) on December 12, 2025 as a spiropyrimidinetrione \u2014 a genuinely new mechanism targeting bacterial type II topoisomerase through a distinct binding mode. It was developed by GARDP, the Global Antibiotic Research and Development Partnership, a not-for-profit organization.<\/code><\/pre>\n\n\n\nThe architecture of a fix<\/h3>\n\n\n\n
The shape of the failure<\/h3>\n\n\n\n