Fentanyl — 3,600 Dollars of Chemicals, 3 Million in Pills

The core fact of the fentanyl economy is that the physical quantity in dispute is trivially small. RAND estimates that synthetic fentanyl reduces the cost of producing opioids at higher market levels by about 98%, and that perhaps only a few pickup-truck loads of pure fentanyl would meet total United States demand for an entire year [8]. One estimate cited by Undark put all the fentanyl entering the country in 2016 at roughly ten industrial barrels [37]. Interdiction is a volumetric instrument aimed at a problem that has almost no volume. Heroin needed poppy fields, harvest labour, refining sites and bulk freight, all of which can be found, watched and destroyed. Fentanyl needs a handful of listed chemicals, a rented building and a chemist. That single change in the physics of the commodity explains more about the last decade of failed drug policy than any account of demand or despair.

By weight, fentanyl runs up to 50 times the potency of heroin, which is the whole of its commercial logic [8]. A single kilogram can be cut into hundreds of thousands of retail doses, so the ratio of retail revenue to smuggled mass is orders of magnitude better than anything the heroin trade could offer [37]. That ratio is what traffickers optimise, and it is also what kills people. The difference between a tolerable dose and a fatal one is measured in single milligrams, and the mixing is done by hand in unregulated premises. DEA laboratory analysis of counterfeit tablets found an average of 2.3 milligrams of fentanyl per tablet across a range from 0.1 to 7.0 milligrams, with 67% of tablet exhibits containing at least 2 milligrams [14].

The global frame is expanding rather than contracting. UNODC counted 331 million people who used a drug in 2024, or 6.2% of the population aged 15 to 64, against 5.2% a decade earlier, including 63 million opioid users [20]. Authorities identified 755 new psychoactive substances in circulation in 2024, of which 118 appeared for the first time, and seizures now contain roughly five times as many distinct drug types as before 2000 [20]. UNODC Executive Director Monica Juma described an unprecedented spike in new drug types, some more potent or more dangerous than what came before [20]. The lead researcher on the report framed the same finding as a market becoming more diverse and less knowable [20].

Enforcement doctrine assumes that pressure on supply raises price and that higher prices suppress consumption. Fentanyl breaks the first link. Wholesale prices in Culiacan fell from roughly 7,000 dollars a kilogram in 2022 to about 3,000 dollars in 2023, with border prices falling from 15,000 to 7,000 over the same period [12]. When precursor restrictions bit in August 2024 the Culiacan price doubled back to around 6,000 dollars a kilogram, which sounds like a victory until the arithmetic is done [12]. At several hundred thousand retail doses per kilogram, a 3,000 dollar increase in wholesale cost is a rounding error at street level. Retail prices stayed flat throughout the shortage [39].

The same arithmetic governs seizures. RAND has argued that interceptions high in the distribution chain are worth less to enforcement precisely because replacement is cheap for traffickers [9]. A seized kilogram of heroin destroyed months of agricultural investment. A seized kilogram of fentanyl destroys a few thousand dollars of chemicals and a weekend of work. Customs and Border Protection reported more than 100 million lethal doses seized along the south-west border in the first half of fiscal 2026, intercepted by officers at 54 ports of entry across four states [15]. The number is designed to sound decisive. Set against a market whose annual national requirement is measured in truckloads, it describes throughput rather than attrition.

None of this makes enforcement pointless, and that is not the argument here. It makes enforcement a poor instrument for controlling quantity, which is the job it has been assigned, funded and advertised to do. The interesting supply-side questions sit elsewhere: in the precursor market upstream of manufacture, in the purity and adulterant decisions taken by mid-level producers, and in the composition of the retail supply that determines whether a given transaction kills someone. Those three places are where the last three years of falling mortality were actually decided [3][40].

The supply chain has three legs and only one of them is criminal at the point of action. Chemical manufacturers in China and India produce precursors and pill presses that are, in most cases, lawful goods in the jurisdiction where they are made [6]. Brokers move them to Mexico, where clandestine laboratories run by the Sinaloa Cartel and the Jalisco New Generation Cartel convert them into finished fentanyl and press it into counterfeit tablets [13][7]. Distribution into the United States runs mostly through legal ports of entry [17]. Each leg is optimised separately and each carries its own substitution options, which is why disrupting a single leg reroutes the chain rather than breaking it.

The most complete public account of the first leg is the Reuters investigation Fentanyl Express, awarded the 2025 Pulitzer Prize for Investigative Reporting [10]. Reporters bought 11 of the 12 chemicals required to synthesise fentanyl from seven separate Chinese suppliers, paying in bitcoin; the twelfth came from a United States chemical company through Amazon [11]. The total outlay was a little over 3,600 dollars, and the resulting material would have supported up to 3 million dollars in finished tablets [10][11]. No specialist access was required. The series went on to map the vendors, the shipping routes and the customs gaps in both Mexico and the United States [11].

Precursor pricing makes the substitution logic explicit. Before the Chinese restrictions of August 2024, suppliers quoted 1-BOC-4-AP at 700 dollars a kilogram on a 15-kilogram order and 650 dollars on a 20-kilogram order, with 1-BOC-4-Piperidone at 400 and 350 dollars respectively [12]. After the restrictions the same vendors offered unscheduled alternatives: 1-BOC-4-hydroxypiperidine at 190 dollars a kilogram on a 25-kilogram minimum, and a benzene derivative at 65 dollars a litre on a 100-litre minimum, advertised with assurances of no customs problems and fast delivery to Mexico [12]. Control of a named molecule produced a cheaper unnamed one.

Chinese regulatory action has been real and sequential rather than absent. Beijing controlled fentanyl substances as a class in 2019 [6]. In June 2025 it completed the scheduling of every fentanyl precursor listed by the International Narcotics Control Board, and in July 2025 it designated the nitazene class for domestic control [6]. In November 2025 the Ministry of Commerce placed export controls on 13 precursor chemicals bound for North America, and in May 2026 three further chemicals were added [6]. The DEA has since reported that some China-based companies appear reluctant to supply controlled precursors overseas and that some Mexican producers are struggling to obtain key inputs [6].

The binding constraint is that chemical space is far larger than any schedule. UNODC recorded 755 new psychoactive substances in circulation in 2024, with 118 appearing for the first time, which is the same adaptive capacity applied to end products rather than inputs [20]. A scheduling system works by naming things. A synthesis route works by finding things that have not been named. The asymmetry is structural, and it is why the intelligence community still lists China and India as primary precursor sources in March 2026, seven years after the first class-wide Chinese control [6].

The manufacturing leg has consolidated rather than dispersed. The Sinaloa Cartel and the Jalisco New Generation Cartel, designated foreign terrorist organisations in 2025, run the clandestine laboratories and the pill presses, sourcing both from Chinese suppliers [13]. Their associates operate in almost every United States state and sell through social media and messaging applications rather than street corners [13]. A nineteen-month internal war inside the Sinaloa Cartel disrupted operations materially, and it is one of the leading candidate explanations for the supply disturbance visible through 2024 [39]. That disruption was not a policy achievement.

The geography of seizure does not match the geography of the political argument. Most fentanyl is found in vehicles driven by United States citizens through official ports of entry rather than carried between them [17]. Customs and Border Protection announced more than 100 million lethal doses seized along the south-west border in the first six months of fiscal 2026, work performed by officers at 54 ports of entry across California, Arizona, New Mexico and Texas [15]. The enforcement task is in practice a customs inspection problem at legal crossings, which is a very different operational proposition from the one that gets funded and described.

Seizure volumes then did something that should have ended the debate. United States fentanyl seizures fell 53% between January and June 2025 against the same months of 2024, and by March 2025 monthly volumes had dropped to about 760 pounds [16]. Between January and July 2026 authorities seized 6,200 pounds, 2% above the same period of 2025 [16]. Falling seizures are read as either an enforcement failure or a supply contraction, and in this case both readings were advanced simultaneously by people with an interest in each. Neither can be settled from the seizure series alone, which is precisely the problem.

The tariff experiment tested the leverage theory in public. A 10% tariff on Chinese goods framed as a fentanyl measure was imposed in February 2025 and raised to 20% in March [18]. In November 2025 it was cut back to 10% after China agreed to tighten controls on 13 precursor chemicals, and in February 2026 the Supreme Court eliminated the tariffs imposed under emergency economic powers [18][6]. Four rate changes in twelve months amounted to a natural experiment of sorts, and the mortality series did not respond to any of them on the timescale the policy implied.

Marcus Noland of the Peterson Institute for International Economics set out the sequencing problem directly. Overdose deaths began falling in 2023, before the tariffs existed [18]. He finds no evidence that the tariffs reduced overdoses, while assessing that earlier diplomatic engagement, particularly the 2019 agreement between the United States and China, was associated with a reduction of roughly 1,000 deaths [18]. The distinction matters for policy design: the instrument that appears to have moved the series was a negotiated scheduling change rather than a trade penalty, and the two are routinely conflated in the same speeches.

The deeper problem is that interdiction has no observable target. Policy success is reported in kilograms intercepted, pills seized and laboratories dismantled, none of which is the variable that kills people. The variable that kills people is the dose distribution in the retail supply, and until recently nobody measured it systematically. The DEA seized approximately 47 million counterfeit pills in 2025 [13]. The number is real and the enforcement is real, but no published federal estimate gives the counterfactual quantity that reached users, which means the ratio between the two — the only figure that would make the seizure number interpretable — does not exist.

There is one supply-side intervention with a reasonably clean evidence trail, and it is not a seizure programme. Chinese class-wide scheduling in 2019 measurably curtailed direct shipments of finished fentanyl into the United States and pushed synthesis to Mexico [6]. That is a real effect, and it is also a warning. The intervention did not reduce the drug reaching users; it relocated the synthesis step and lengthened the chain. Supply-side measures at this point in the market reliably change the shape of supply. Whether they change its lethality is a separate question, and the answer turns out to be yes, in ways nobody intended.

Richard Cowan named the governing mechanism in 1986: the more intense the law enforcement, the more potent the drugs will become [37]. Leo Beletsky and Corey Davis revived the argument in the International Journal of Drug Policy in 2017, applying it to the transition from heroin to illicitly manufactured fentanyl [38]. The logic is an application of the Alchian and Allen effect: when a fixed cost is imposed on each unit of transport, consumption shifts towards the more concentrated variant. Prohibition raises the cost of moving mass, so mass falls and concentration rises. Alcohol potency rose an estimated 150% during United States Prohibition [37].

The academic reception is genuinely divided, and the disagreement is not a technicality ⚖ Contested. Keith Humphreys of Stanford called the iron law an opinion with no empirical foundation, noting that cannabis potency rose after legalisation as well [37]. Jonathan Caulkins of Carnegie Mellon dismissed it outright, and Peter Reuter of the University of Maryland suggested that copper law would be a more honest name [37]. Beletsky himself has conceded that it is not an ironclad law [37]. Sarah Mars of the University of California San Francisco argues that fentanyl spread because synthesis is easy, not because prohibition forced the substitution [37].

What happened next is the part no version of the theory predicted. From 2023 the composition of the United States fentanyl supply moved in the opposite direction. DEA laboratories recorded average powder purity falling from 19.2% in 2022 to 11.36% in 2024, with individual samples ranging from 82% down to 0.07% [13]. Counterfeit tablets containing a potentially lethal dose fell from 76% of those tested in fiscal 2023 to 29% in fiscal 2025 [13]. InsightCrime recorded the same turn from the production side, citing DEA analyses showing five in ten seized pills carrying a lethal dose in 2024 against seven in ten in 2023 [12].

Adulteration compounded the effect in a direction nobody chose. Medetomidine displaced xylazine as the most common adulterant in the Philadelphia opioid supply during 2024, appearing in 72% of tested samples in the final four months of the year while xylazine detection fell from 98% to 31% [41]. A New Haven study using liquid chromatography and mass spectrometry found medetomidine in 64.8% of samples and xylazine in 42.2%, with fentanyl nearly ubiquitous at 95.3% but ranging in concentration from under 1% to over 20% [43]. Sedatives of this class are not opioids and do not produce the respiratory arrest that kills.

The dividend arrives with a bill. Clinicians in Philadelphia began reporting a severe withdrawal syndrome distinct from both fentanyl and xylazine withdrawal, marked by profound autonomic dysfunction, between September 2024 and January 2025 [42]. Medetomidine is estimated to be more than a hundred times as potent and selective at the alpha-2 receptor as xylazine [41]. Fewer people are dying at the moment of use and more are arriving in hospital with a syndrome that has no settled treatment protocol. Sampling also shows that retail branding carries no information about content: fentanyl concentration varies widely within products sold under the same stamp [43][51].

The strategic reading is uncomfortable for every side of the policy argument. Mortality fell because the drug got weaker and got mixed with something that suppresses breathing less, which is a supply-composition accident rather than a public health achievement. Nothing in the current policy mix holds that composition in place. If precursor availability improves, if a producer decides that potency is again a competitive advantage, or if nitazenes take the share fentanyl currently holds, the dose distribution moves back and the mortality series moves with it [21][24].

The scale of the reversal has no precedent in the modern series. Deaths peaked above 106,000 in 2021 after two decades of near-continuous increase [52]. By the twelve months ending January 2026 the CDC provisional model predicted 69,147 deaths, 13.2% below the preceding twelve months [2]. Brandon Marshall, the Brown University epidemiologist, described the speed of the decline as unprecedented [5]. The fall is also uneven: Rhode Island, New York, North Carolina, Alabama and Vermont recorded declines of 25% or more, while New Mexico, Arizona and Colorado recorded increases of 10% or more [1].

The most careful decomposition comes from the Centers for Disease Control division of overdose prevention. Deborah Dowell and colleagues, writing in the Lancet Regional Health Americas in September 2025, modelled deaths as the product of three terms: the population at risk, overdoses per person, and the probability that an overdose proves fatal [3]. Deaths fell 25% between March 2024 and March 2025, from 103,529 to 77,648 [3]. Opioids were involved in 66.2% of deaths in the twelve months to March 2025 and synthetic opioids in 57.9% [3].

Two counterfactuals in that paper carry the argument. Had the population at risk stayed at its earlier level, the model implies 109,783 additional deaths [3]. Had the probability of fentanyl involvement stayed constant, 260,024 fewer deaths would have occurred [3]. The first figure says the epidemic partly exhausted its own pool of people; the second says the composition of the supply is the dominant term by a wide margin. Dowell and colleagues also note that fentanyl had effectively saturated the opioid supply by 2023, leaving limited opportunity for new exposure among people already using opioids [3].

The upstream numbers support the population reading. Opioid prescribing fell from 46.8 prescriptions per 100 people in 2019 to 37.5 in 2023, and estimated opioid use disorder prevalence fell from 2.1 million in 2016 to 1.6 million in 2019 [3]. Mary Beth Sheridan, writing for the Inter-American Dialogue in April 2026, added the demand-side observation that street prices stayed flat through the 2023 shortage, which is what a contracting market looks like rather than a constrained one [39]. Fewer young people took up fentanyl than took up the opioids of the previous wave [39].

The supply-shock account has its own evidence and it is not weak. Researchers tracking online drug forums found complaints about fentanyl shortages rising in July 2023, spiking later that year and remaining high through 2024, which lines up with Mexican producers struggling to obtain Chinese precursors [39]. Seizures dropped 37% by the end of 2024, and the overdose death rate fell by about a third between its mid-2023 peak and the end of that year [39]. The correlation is real. What it cannot establish is whether the constraint came from enforcement, from cartel conflict, or from Chinese regulatory action taken for domestic reasons.

The Commonwealth Fund review published in September 2026 declines to name a single cause, and that is the honest position [4]. Its author, Evan Gumas, reports that no one factor explains the decrease, listing expanded naloxone access, wider availability of medications for opioid use disorder, better treatment in prisons, improved data infrastructure, a depleted at-risk population, declining fentanyl potency and the shift towards less immediately lethal contaminants [4][5]. The policy consequence of that list is specific: only some of those factors can be deliberately sustained, and the ones currently being cut are the ones that can.

The order-of-magnitude difference between North America and Europe is the single most informative fact in the comparative data. The European Union counted about 7,600 deaths involving illicit drugs in 2024, rising above 8,300 when Norway and Turkiye are included, at 25 deaths per million residents aged 15 to 64, with men accounting for four in five [21]. The United States recorded 69,973 in 2025 with a population roughly two-thirds the size of the European Union [1]. Fentanyl reached one market comprehensively and the other marginally, and that difference in supply does most of the explanatory work.

Europe is nonetheless acquiring the problem, in a different chemical form. Deaths linked to fentanyl and its derivatives in Europe rose to about 210 in 2024 from around 130 in 2023, with Germany reporting the largest national count at 95 [21]. Nitazenes have driven death peaks in Estonia, Finland, France, Germany, Latvia, Norway and Sweden [22]. Estonian drug-induced deaths rose from 82 in 2022 to 119 in 2023, of which protonitazene accounted for 40 and metonitazene for 32, leaving Estonian drug-induced mortality six times the European Union average [22].

The United Kingdom is the clearest European case of a synthetic wave in progress. At least 400 deaths involving nitazenes were recorded between June 2023 and January 2025, in every region [24]. England and Wales recorded 195 nitazene-involved deaths in 2024 against 52 in 2023 [24]. Scotland recorded 76 in 2024 and 38 in the first quarter of 2025 alone, a pace that would double the annual figure [24]. Public Health Scotland issued a formal contamination alert in August 2025, noting that nitazenes there appear mainly as contaminants in material sold as heroin, benzodiazepines and oxycodone [23].

Canada ran the most explicit policy experiment and then reversed it. The British Columbia decriminalisation pilot, permitting adults to possess up to 2.5 grams of specified substances, began on 31 January 2023 [25]. In May 2024 criminal penalties for possession in public places were reinstated, and in January 2026 the province declined to extend the pilot [25]. A population-level time-series analysis in the Canadian Medical Association Journal found that decriminalisation substantially reduced police encounters for possession, and that neither decriminalisation nor partial recriminalisation produced a detectable change in drug toxicity deaths or drug-related hospitalisations [25][26]. Canadian overdose deaths fell about 9% in 2024 [4].

Four further national cases bracket the range. Switzerland decriminalised use, made methadone available around the clock and offered pharmaceutical-grade heroin to treatment-resistant patients; overdose deaths fell by half between 1991 and 2010, new heroin initiation fell about 80%, and its opioid death rate now runs near one-twentieth of the United States rate [44]. Portugal, having decriminalised possession of up to ten days of supply, saw drug-induced deaths fall from about 90 in 2008 to 27 in 2016 and now reports the lowest rate in Western Europe [45][46]. Sweden has cut deaths roughly 25% since 2017 by expanding medication access [4]. Japan reports no widespread fentanyl misuse at all, with customs authorities not intercepting the drug [50].

The comparative lesson is narrower than either camp usually claims. Countries with low synthetic opioid mortality are mostly countries fentanyl has not reached at scale, and Japan belongs in that category rather than in the policy-success category [50]. Among the countries the drug has reached, the ones with falling mortality are the ones with dense treatment coverage and saturated naloxone distribution [4][29]. Enforcement intensity does not sort the outcomes in either group. The United States has the most enforcement and the worst result, which is not a causal finding but does bound what enforcement can be expected to deliver.

Naloxone is the intervention with the cleanest evidence base and the lowest unit cost. The March 2025 systematic review and meta-analysis in BMC Public Health found overdose education and naloxone distribution associated with 27% and 46% reductions in opioid overdose mortality in low- and high-implementing communities respectively [29]. A New York State study published in October 2025 attributed more than 6,500 lives saved over two years to naloxone distribution, equivalent to more than 204,000 years of life, and found the programme economically beneficial [30]. Virginia recorded a 40% fall in deaths between 2023 and 2024 after a saturation plan, and Maryland more than doubled distribution between 2021 and 2025 [4].

Medication for opioid use disorder is the intervention with the largest unmet gap. Of the estimated 9.37 million United States adults who needed treatment for opioid use disorder in 2022, 25.1% received the recommended medications [34]. Put the other way, close to nine in ten people with opioid use disorder are not receiving the medication that would most reduce their risk of dying [35]. Methadone use among Medicaid enrollees rose more than threefold between 2010 and 2020, from 1.9 to 6.2 users per thousand, which is genuine progress against a baseline so low that tripling it still leaves most people untreated [36].

Service capacity is expanding where states chose to expand it. Colorado more than doubled its methadone clinics from 25 to 54 in five years, and New York increased medication access in prisons and jails sevenfold between 2022 and 2025 [4]. Rhode Island opened the first state-sanctioned overdose prevention centre in the United States in January 2025; through April 2026 it recorded approximately 12,040 visits, more than 150 interventions and zero fatal overdoses [32]. Rhode Island accidental overdose deaths fell to 219 in 2025 from 329 in 2024 and 436 in 2022 [31].

The money exists and its stewardship is the open question. More than 50 billion dollars in opioid settlement funds has been committed across United States jurisdictions since 2022, with at least 70% required to be spent on opioid remediation [47][48]. Nearly all jurisdictions have now established formal governance structures for allocation [47]. States sit at very different stages of accountability: South Carolina runs a public portal for grant awards and expenditures, Missouri publishes annual reports on state and local spending, and Michigan added annual local reporting from fiscal 2026 [47]. Comprehensive independent tracking has been left largely to journalists [49].

Cost-effectiveness is not the binding constraint on any of these interventions. The New York naloxone evaluation found distribution economically beneficial on its own terms [30]. A systematic review of economic evaluations of overdose prevention centres across twelve North American cities reached broadly favourable conclusions, and a 2026 study found that opening a centre in New York City was not associated with significant changes in local commercial activity, which is the specific harm most often cited in opposition [33]. Sarah Wakeman of Mass General Brigham frames the underlying point as a matter of failure rather than tragedy: nobody should die of an opioid overdose when an antidote exists [5].

The demand-side case is not that supply does not matter. It is that supply-side effects arrive through composition rather than quantity, and composition is not something any government currently controls. Treatment coverage, naloxone density and low-threshold services are instruments that governments do control, and their effect sizes can be estimated in advance [29][30]. The 2025 mortality figures were produced by an accidental improvement in the drug supply layered on top of a deliberate expansion of these services [4]. Only one of those two inputs can be ordered again next year.

Four things in this record are not seriously disputed. The physical quantity required to supply the United States opioid market is small enough to defeat volumetric interdiction [8]. Precursor scheduling reliably produces substitutes [12]. United States overdose deaths have fallen for three consecutive years [1]. And roughly three quarters of Americans who need medication for opioid use disorder are not receiving it [34]. Every serious position in this debate has to be built on those four facts, and most public argument is built on none of them.

What is disputed is causal attribution for the decline, and the dispute has a clear structure ⚖ Contested. One camp reads the Chinese precursor controls, the cartel disruption and the seizure record as evidence that supply pressure worked [6][39]. The other reads the potency and adulterant data, the timing, and the CDC decomposition as evidence that the market changed its own product for reasons largely independent of policy [40][3][18]. Both camps point at the same 2024 turning point. Neither has produced an identification strategy that separates the two.

Two further disputes sit underneath that one. The iron law of prohibition is contested on the evidence rather than on emphasis: its authors concede it is not ironclad, and senior researchers including Keith Humphreys and Jonathan Caulkins reject it [37][38]. Prescribed safer supply is contested on the magnitude of diversion: a Vancouver observational evaluation found 62% of participants taking hydromorphone as prescribed, 14% smoking or injecting some of it, and 20% selling or trading a proportion [27]. British Columbia responded by requiring witnessed dosing from February 2025 [28].

The decriminalisation evidence is the most widely misread ⚖ Contested. The Canadian Medical Association Journal time-series found that the British Columbia pilot reduced police encounters for possession substantially and changed neither drug toxicity deaths nor drug-related hospitalisations [25][26]. Read as a mortality intervention it failed. Read as what it was, a change in how possession is policed, it did what it was designed to do, and the province ended it in January 2026 on grounds of public order rather than mortality [25].

That suggests a measurement agenda before a policy agenda. Drug-checking and forensic sampling produce the only series that tracks the variable actually driving deaths, and the existing data — DEA purity profiles, Philadelphia and New Haven adulterant sampling, stamp-level variability studies — already show how much can be learned [13][41][43][51]. None of it is currently assembled into a national indicator published on the cadence that seizure statistics enjoy. Making supply composition a reported figure would change both what enforcement agencies are held to and what treatment systems can anticipate.

The honest summary is that the fentanyl economy is a supply-side problem that supply-side policy cannot solve. Its inputs are lawful goods made in jurisdictions with strong reasons to keep making them, its manufacture needs no land and little capital, its transport requirement is measured in truckloads a year, and its lethality is set downstream of every intervention currently deployed [8][6][12]. The measures with estimable effects all sit on the other side of the transaction: naloxone, medication coverage, low-threshold services, supervised consumption [29][30][32]. Three consecutive years of falling deaths bought the time to build them, and that supply-side gift is the one thing in this system nobody can renew [1][3].