Inside the hidden supply chain behind the grey-market peptide boom. We contacted the suppliers that sell "factory direct" access. Prices were easy. Warehouses were easy. The factory was not.
In July 2025, a batch of semaglutide changed identity without changing chemistry.
According to an FDA warning letter issued the following May, Harbin Jixianglong Biotech bought semaglutide active pharmaceutical ingredient from a supplier that was not on its approved list. It repackaged the material without documentation, assigned it a fresh batch number — CP-030-20250711 — and printed its own name on the container as the manufacturer. It moved the manufacturing date forward to 25 July 2025 and pushed the retest date out to July 2027, neither supported by data. The API reached the United States on 23 August 2025. Investigators found a second batch handled the same way.
Asked to explain, the company told the FDA that demand for semaglutide in the US was high, and that it had "occasionally purchased the product externally for resale."
Two details make this more than a paperwork violation. The first is where the material went: into US compounding pharmacies, to patients holding prescriptions. It was recalled in February 2026. The FDA's enforcement record cited failures that included process validation and bacterial-endotoxin method validation before distribution.
The second is who Harbin Jixianglong was. In September 2025 the FDA had created a "green list" of foreign facilities cleared to export GLP-1 APIs into the United States, precisely so that unverified material could be detained at the border. Harbin Jixianglong was on it. The FDA later removed the firm from the green list and placed it on import alert.
This happened inside a regulated channel — with inspectors, records, recall authority and a named accountable entity. Hold that in mind, because the grey market has far fewer mechanisms for answering the same basic question: who actually made this batch?
At retail, this market looks enormous. Hundreds of brands, each with a logo, a founder story, a sales representative and a claim to direct sourcing. The word manufacturer appears frequently because it is commercially valuable. Buyers believe that removing the middleman should mean a lower price and a clearer line to the source.
Upstream, the market appears much smaller. Financial Times reporting in 2026 estimated roughly 1,000 China-based sellers targeting overseas peptide buyers while describing manufacturing as concentrated in a far smaller group of factories, with intermediary layers between production and the retail-facing brand. That is an external estimate, not a PeptideAtlas count. But it describes the same structural problem our own outreach kept running into: seller identity is easy to find; production identity is not.
None of this makes distribution illegitimate. A distributor can add useful services: warehousing, customs-risk absorption, replacement policies, independent sampling, lot tracing and customer support. The problem is narrower and more specific: the distribution role is often marketed as if it resolves the manufacturing role. It does not.
A company can also manufacture some products and trade others. Those roles are not mutually exclusive. That is exactly why the label on a website is insufficient. Product-level provenance has to be demonstrated batch by batch.
In April 2026 a GLP-1 Forum member posting as 5byfive attempted to infer common supply chains from laboratory results. He started with roughly 6,200 Finnrick Analytics certificates, removed many apparent US resellers, added JEEP and STG records, and worked with a final dataset of 5,181 records weighted toward recent tests.
The analysis resolved 96 vendor identities into five active statistical groups, with a sixth legacy group that had gone stale. The author was explicit about what those groups meant: common supply chains, not five proven factories. He also acknowledged that a vendor can switch sources, and his recency-weighted update did exactly what a dynamic supply-chain model should do — it moved vendors.
That movement is the most important part of the map. These groups are not surnames for permanent factories. They are fingerprints of sourcing relationships at a point in time.
| Group | Named vendors or anchors recoverable from the published analysis | What distinguishes the group | Status / caution |
|---|---|---|---|
| Group 1 | SRY Labs, Uther, ZLZ are named as the anchor cluster in the latest public update. The author also says seven new vendors entered Group 1, but their names are only in the attached PDF, which is not exposed in the public HTML. | GHK-Cu 50mg fill difference around +17.0%, far above the other groups in the recency-weighted analysis. | Active statistical cluster. In the latest update, six vendors that had previously shared this group moved to Group 2. |
| Group 2 | Latest public update explicitly moves HK Peptides, Qing Li Peptide, Reta-Peptide, Haikou Mingheng Technology, Xingtai Jiachuang Technology and Shanghai Wibson (WBS) into Group 2. Earlier PeptideAtlas research notes also place JEEP, QST (Qingdao Saber Technology Pharma) and Huaian Hanyou in Group 2. | Highest reported Tesamorelin 10mg purity among the groups in the latest public comparison (98.8%); aggregate purity cited by the analyst at 99.52%. | Active, but membership is time-sensitive. Treat older and newer maps as versions, not contradictions. |
| Group 3 | The latest public update identifies SPB as a new Group 3 entrant. The full roster is contained in the forum attachment rather than the HTML post. | Tesamorelin 10mg purity around 93.0% and Tirzepatide 60mg around -3.8% fill difference / 98.52% purity in the author's recency-weighted comparison. | Active statistical cluster; full current roster not independently recoverable from the public page. |
| Group 4 | An earlier map placed HK Peptides, Shanghai Wibson (WBS) and Shanghai ERP among a 16-vendor Group 4. The later recency-weighted update explicitly moved HK Peptides and WBS to Group 2. | Lowest fill-difference standard deviation in the analyst's comparison (+/-15.7%), average fill difference +3.6%, aggregate purity 99.43%. | This group is the best illustration of why the map must be versioned. Do not freeze an older roster and call it current. |
| Group 5 | Earlier PeptideAtlas research notes identify Yabang Peptide in Group 5; the older map described six vendors in the group. | Very wide fill variability in the later aggregate comparison (+/-51.1%). | Active in the analysis, but the complete current roster is not available in the public HTML. |
| Group 6 — stale | Alpha-Gen, Loti Labs, Suaway Lab Research. | No current signature: the author reports no group data after September 2025. | Preserved as a legacy pattern, not treated as a current supply relationship. |
Two additional vendors, ACR and PTB, were reported as having insufficient cross-product data to assign confidently. The public post also says one new vendor, SPB, entered Group 3.
The most revealing update is the coordinated movement of six names — HK Peptides, Qing Li Peptide, Reta-Peptide, Haikou Mingheng Technology, Xingtai Jiachuang Technology and WBS — from the old Group 1 into the newer Group 2 after recent tests were weighted more heavily. The author described that as coherent drift rather than noise.
*Figure 1. A supply-chain map that moves. Six vendors moved together from the older Group 1 into the recency-weighted Group 2 after newer laboratory results were given greater weight. That coordinated movement is evidence that the groups are time-sensitive sourcing fingerprints, not permanent factory identities.*
This is precisely why PeptideAtlas is not publishing the sentence "96 brands are five factories." The cluster count is useful as a map of similarity. It is not a corporate registry, a customs manifest or a factory audit.
The statistical map became more interesting when buyers started comparing what was physically arriving.
One forum user reported that a WBS T30 order and an ERP T30 order carried the same label and the same batch number. Another documented visually matching KLOW kits arriving through WBS, Alimo Peptide and Peptide ABC Bio: same label format, same abbreviations and the same research-use boilerplate.
That does not tell us whether the common node was an API synthesizer, a fill-and-finish operation, a kit packer, a wholesaler or a combination of them. But it is a different class of evidence from statistical similarity. Two nominal competitors presenting the same batch identifier are no longer merely correlated datasets.
The cluster work raised a practical question: if the vendor layer is partly a trading layer, could a buyer simply skip it and go to the actual factory?
PeptideAtlas contacted China- and Hong Kong-based suppliers during August 2026 asking for pricing, inventory, warehouse locations, customization and, critically, the identity of the upstream facility responsible for production.
The pattern in the correspondence was lopsided. Prices were easy to obtain. Catalogues were easy. Stock counts were easy. Warehouse locations were easy. Custom labels and logos were easy. A verifiable upstream synthesis or fill-finish identity was much harder.
That is commercially rational. A trading company that discloses the factory behind its product risks being bypassed by its own customers and undercut by competitors. The information buyers most want — the name of the production source — is therefore the information the intermediary has the strongest incentive to protect.
The evidence does not allow us to say that every company calling itself a manufacturer is lying. Some may synthesize certain products themselves and trade others. What it does show is that the marketing word manufacturer cannot be treated as proof that the seller synthesized or filled the particular product being quoted.
The WBS material is a useful example. Its August wholesale sheet is headed "Factory R&D and manufacturing" and "WBS Wholesale Price (Shanghai Wibson Biotechnology Co., Ltd)." The same document offers quantity tiers from one kit through 100 kits and explicitly says the company supports product customization, including custom labels and logos. That is a private-label capability sitting directly beside a factory-manufacturing identity. It may reflect vertically integrated manufacturing, trading capability, or both. The document alone cannot tell the buyer which role WBS played for a specific vial.
That ambiguity is the story.
The most concrete new evidence from our outreach came not from a certificate. It came from inventory spreadsheets.
PeptideAtlas received warehouse-stock materials from multiple supplier contacts and compared them line by line. In two pairs, nominally separate supplier identities pointed to the same fulfillment infrastructure:
Shanghai Wibson Biotechnology (WBS) and ERP Biotechnology (ERP): PeptideAtlas's direct-contact records show the same overseas warehouse network and matching stock information. The materials list US warehouses in Portland, Delaware and Atlanta alongside Canada, Australia and Europe, with Seattle inventory called out separately. The item-level inventory counts supplied to us matched across the independently received stock materials.
Hong Kong Lejian Biotechnology / LJ2 and Homopeptide: PeptideAtlas separately received warehouse-stock materials that matched at the warehouse and SKU-inventory level, again indicating shared fulfillment infrastructure rather than two wholly independent logistics systems.
The screenshots are reproduced in the evidence appendix. The original high-resolution files and correspondence are retained by PeptideAtlas.
What does an exact warehouse-stock match prove? It proves less than "same factory" and more than "coincidence." Plausible explanations include a common third-party logistics provider, a shared master distributor, one seller reselling another's stock, or multiple sellers drawing from the same inventory pool. Any of those explanations is relevant to a market sold on the idea that each brand is an independent manufacturer.
It also changes how the lab clusters should be read. WBS and ERP appeared together in an earlier Group 4 map. In the newer recency-weighted analysis, WBS moved into Group 2 while ERP was not named in the movement note. The warehouse evidence shows how two companies can remain linked at the logistics layer even while their recent laboratory signatures stop clustering the same way. Supply chains have layers; one pair of vendors can share one layer and diverge at another.
*Figure 2. Warehouse evidence supplied during PeptideAtlas direct outreach. Vendor-to-image mapping is based on PeptideAtlas's correspondence records; the screenshots themselves are presented as documentary exhibits rather than independent proof of a particular physical factory. The relevant finding is the repeated warehouse network and matching SKU-level inventory across separately contacted supplier identities. Personal photographs in the sender avatars have been redacted.*
Map the chain and it has at least four roles, which real transactions can combine or skip.
The point is not that every vial follows an identical path. It is that the company taking the payment can be several commercial relationships removed from the facilities that synthesized and finished the material.
The attached quote sheets make the economics visible in a way that abstract discussion does not.
PeptideAtlas's August 2026 retail pricing workbook captured 4,544 live priced listings from 44 storefronts, including 4,118 peptide listings and 856 comparable product-and-dose combinations. The workbook normalised currencies to USD and calculated same-dose market medians. We compared those retail medians with the supplier quotes supplied directly to us from WBS, Homopeptide and JEEP.
The result is not a net-profit calculation. It is a gross price-layer comparison before freight, payment processing, testing, packaging, customs loss, advertising, labour, tax, refunds and inventory risk.
| Product / dose | Attached China/HK supplier quote per vial* | Median retail price per vial | Retail / supplier multiple | Implied gross spread before operating costs |
|---|---|---|---|---|
| Semaglutide 5mg | $3.00–$4.00 | $35.48 | 8.9x–11.8x | 89%–92% |
| Tirzepatide 10mg | $5.00–$7.80 | $42.21 | 5.4x–8.4x | 82%–88% |
| Tirzepatide 30mg | $8.50–$13.80 | $93.67 | 6.8x–11.0x | 85%–91% |
| Retatrutide 10mg | $7.00–$12.80 | $72.78 | 5.7x–10.4x | 82%–90% |
| BPC-157 10mg | $4.90–$8.00 | $47.30 | 5.9x–9.7x | 83%–90% |
| GHK-Cu 50mg | $2.50–$4.80 | $35.66 | 7.4x–14.3x | 87%–93% |
*\*Supplier ranges include the attached one-kit, high-volume and/or custom tiers where available. They are point-in-time quoted prices for 10-vial kits, divided by ten. They should not be read as a universal China price or as proof of any seller's actual cost of goods.*
The shape is remarkably consistent. A 10mg tirzepatide vial quoted at roughly five to eight dollars in the supplier layer appears at a median above forty dollars in the retail workbook. A 50mg GHK-Cu vial quoted at roughly $2.50 to $4.80 appears at a median around $35.66. A 10mg BPC-157 vial quoted at about $4.90 to $8 sits at a $47.30 retail median.
*Figure 3. Supplier quote versus retail median. Retail medians come from PeptideAtlas's 23 August 2026 pricing workbook. Supplier quotes come from the attached WBS, Homopeptide and JEEP materials. This is a price-layer comparison, not a net-margin estimate.*
*Figure 4. Gross price multiple from attached supplier quotes to retail median. The range reflects the supplier quote range in the attached materials. It does not deduct the substantial costs and risks between wholesale acquisition and retail sale.*
This helps explain why the upstream factory identity is valuable commercial information. If an intermediary can purchase at single-digit dollars per vial and sell into a retail market several multiples higher, the right to stand between factory and customer is economically meaningful. Disclosing the source can destroy the intermediary's moat.
This is the number everyone wants and the number the grey market does not disclose.
We could not establish the true manufacturing cost of WBS, ERP, Homopeptide, LJ2, JEEP or any unnamed upstream facility from the materials provided. A wholesale quote is an upper bound on the combined cost of manufacturing, fill-finish, packaging and the supplier's own margin — not a factory cost sheet.
There is, however, a useful industrial benchmark for semaglutide.
In 2025, engineers at Intelligen modeled a conceptual commercial semaglutide API facility using SuperPro Designer. The plant was designed for 500 kilograms of purified semaglutide per year. The model estimated a unit production cost of $105,500 per kilogram, or about $0.1055 per milligram of API.
That works out to roughly $0.53 of modeled API production cost in 5mg of semaglutide, and $1.06 in 10mg.
The benchmark is not a claim about any grey-market supplier's cost. It assumes a large industrial facility and covers API production, not a finished research vial delivered to a warehouse. Fill-and-finish, analytical testing, vial/stopper/label, quality systems, packaging, freight and commercial overhead all sit on top of it. Grey-market producers may also use different synthesis routes and operate at very different scales.
Still, it puts the layers in perspective. Against that modeled $0.53 of API cost, the attached supplier materials quote a 5mg semaglutide finished vial around $3–$4 at the wholesale layer, while PeptideAtlas's retail workbook puts the same-dose median around $35.48.
*Figure 5. Semaglutide 5mg across three economic layers. The first bar is a modeled industrial API-production benchmark, not a measured grey-market COGS. The second is an attached finished-vial supplier quote. The third is the observed retail median.*
Peptide manufacturing is not simply the cost of amino acids. Industrial reviews of large-scale peptide synthesis emphasize the heavy use of solvents, resins, coupling reagents, protected amino-acid building blocks, purification media, equipment and waste handling. Labour, QC, facility overhead and yield losses add further layers. That is why we treat the $0.53 figure as a process-economics benchmark rather than a universal "true cost."
The important finding is not that a vial "costs fifty cents." It is that the visible retail price contains several stacked economic layers, while the identity and economics of the first two layers are the least transparent.
In June 2026 Chainalysis added a more unsettling dimension, and the wording of what follows matters.
Its blockchain investigation reported that some China-based chemical suppliers in the grey-market peptide ecosystem had previously been involved in fentanyl or amphetamine precursor sales. Chainalysis was explicit that the market is vast and that most overseas laboratories have no such links.
One case study was Shanghai Sigma Audley New Material Technology, which Chainalysis had identified in 2023 as a fentanyl-precursor supplier, tracing cryptocurrency flows associated with its operations and exposure to darknet vendors and a cartel-related money-laundering operation. In early 2025 the company began appearing on forums selling cosmetic and weight-loss peptides using the same Chinese phone number it had used in precursor listings.
A second involved Bigreat Technology, linked by Chainalysis to precursor sales and Russian darknet markets, which created a separate corporate identity — Zhengzhou DEPU Technology — for peptide sales.
Two things must be said precisely. These companies were not shown to have manufactured fentanyl; Chainalysis describes precursor supply and blockchain exposure. And none of this implicates Chinese peptide manufacturers as a class. China is also home to sophisticated, regulated peptide manufacturing serving pharmaceutical customers worldwide.
The finding is about infrastructure: cross-border sales, payments, logistics and corporate identities can migrate between markets faster than a buyer can reconstruct the underlying history.
PeptideAtlas has seen claims that particular Chainalysis-listed entities appear within the 5byfive clusters. We are not elevating those placements to fact without the underlying cluster attachment and records. They are leads for the next phase of the investigation.
Opacity is a structural problem until it becomes a clinical one.
JEEP — Guangzhou Jeep Biotechnology — was repeatedly recommended in the English-speaking buyer community. In January 2026 it shipped a batch of 30mg tirzepatide identifiable by an orange cap.
In April a buyer-organised testing group sent a vial for endotoxin analysis. The estimate came back at 1,350 EU per vial. Assuming an artefact, testers redirected two more vials from the same batch, donated by different buyers in different countries. One estimated 1,360. The third, at higher dilution, returned a confirmed 2,306.
For context, the commonly used USP bacterial-endotoxin limit for many parenteral drugs is expressed as 5 EU per kilogram of body weight per dose — about 350 EU for a 70kg patient. The reported JEEP values were several times that dose benchmark. This is community testing, not a regulator finding, and the comparison does not by itself establish the amount any individual user received.
The incident does not establish where the endotoxin entered. Raw materials, water systems, equipment, cleaning, handling and fill-and-finish can all matter. What it demonstrates is simpler: a high HPLC purity result cannot rule out a failure class that was never tested, and a consumer-facing brand does not necessarily identify the facility responsible for the relevant process step.
JEEP's public response evolved over several days, eventually including replacement product. The sequence is worth reporting because it shows what accountability looks like when the seller is visible and the production node behind the seller is not.
The evidence question this raises — what a certificate can and cannot establish — is the subject of our companion investigation, The COA Mirage.
The Harbin letter documents provenance failure inside a regulated API supply chain. A US prosecution documents the retail version.
Matthew Kawa and Jennifer Stechkober pleaded guilty in December 2025 to federal charges in the Northern District of Indiana. The Department of Justice said Paradigm Peptides had represented products as safe, pure, tested, US-manufactured and FDA-approved when those claims were false; material was imported from countries including China and India; and products marketed as SARMs contained undisclosed testosterone.
On 30 July 2026 Kawa was sentenced to 70 months in federal prison and ordered to forfeit $5 million. Stechkober received 16 months. Among the conduct reported in the case: forged laboratory certificates.
Neither case licenses calling every vendor fraudulent. Their value here is narrower: manufacturer identity and testing claims are commercially valuable, and the public record contains documented cases where those claims were false.
On 23 and 24 July 2026, the FDA's Pharmacy Compounding Advisory Committee considered seven peptide substances for possible inclusion on the 503A bulk drug substances list. Contemporary reporting described six favorable recommendations. Those votes are advisory; they do not themselves approve a peptide or establish safety.
The relevance to this investigation is structural. A regulated compounding pathway, where lawful, introduces named pharmacies, supplier qualification, documented lots, complaint systems and recall authority. Harbin Jixianglong shows that formal safeguards can fail. It also shows what happens when there is a traceable system: investigators can identify batches, remove a supplier from an import safeguard and initiate a recall.
That is the institutional capability the grey market often lacks. The question is not whether every regulated actor is flawless. It is whether a failed batch leaves a trail that can be followed.
Return to the groups.
The most memorable sentence available here — 96 brands are really five factories — is also too certain to publish. The better finding is more useful: a large set of ostensibly independent vendors compressed into a handful of recurring laboratory signatures; some nominal competitors have also been documented with matching physical batch identifiers; PeptideAtlas's direct outreach found shared warehouse inventory across additional seller pairs; and the identities of the upstream facilities remained far harder to verify than prices, products or stock.
That is a much more interesting picture than a fixed "five factories" claim because it explains the market as a network: one API source can feed multiple finishers; one finisher can serve multiple traders; one warehouse can serve multiple brands; one trader can switch sources; and one brand can therefore look independent at the storefront while sharing several upstream nodes with a supposed competitor.
The next PeptideAtlas analysis is designed to convert more of those relationships from inference to exact-match evidence: duplicate Janoshik task numbers across brands, identical COA image hashes, common manufacturer fields, matching three-decimal purity/date fingerprints, shared warehouse inventories, corporate registrations and customs records.
A cluster says look here. A matching task number, matching batch label or matching inventory ledger tells us much more.
In July 2025, at a longevity conference at a Las Vegas casino resort, at least two women left in ambulances and were hospitalised in critical condition, intubated to help them breathe after receiving peptide injections at a booth. Police indicated as many as seven attendees may have been affected; investigators could not confirm it.
ProPublica documented the regulatory aftermath. Both women recovered. The cause was never determined. Regulators said they could not establish whether the reactions came from contamination, the peptides themselves or something else, and investigators were unable to obtain the serums for testing.
The question is not whether approved peptide medicines can work. Tirzepatide and semaglutide have strong clinical evidence under controlled manufacturing and clinical use.
The question is whether this particular vial can be traced from the label back through the entities that synthesized, filled, tested, warehoused and distributed it — and whether the evidence attached to it actually covers the risks being claimed.
That is the illusion of choice at the centre of the grey market. The storefronts look independent. The underlying relationships are fewer, layered, fluid and frequently hidden. A brand name tells you who took the order. It may tell you little about who synthesized the material, who filled the vial or whose warehouse actually shipped it.
The direct-contact evidence makes that less abstract. Two seller identities can send the same warehouse inventory. A price sheet can promise "factory R&D and manufacturing" while offering private-label logos. A vendor can move from one statistical sourcing group to another as its recent test signature changes. The same product can move from a modeled API production cost measured in cents per milligram to a wholesale finished vial in single-digit dollars and then to a retail price several multiples higher.
The hard question in this market is not which logo looks trustworthy. It is where the chain of custody actually begins.
This article combines five evidence classes that should not be blurred:
The 5byfive clustering analysis is community work. It assumes relatively stable sourcing relationships and uses laboratory data with mixed sample provenance. Similarity does not establish common ownership, a common legal entity or a single physical production site. A group may reflect a common API source, finisher, production practice, laboratory routing, vendor switching or some combination. Membership is time-sensitive; the author explicitly moved six vendors after applying greater weight to recent tests.
The named group table is deliberately not presented as the full 96-vendor roster. The current public forum HTML exposes only part of the membership; the complete roster resides in an attached PDF that was not publicly retrievable during this review. PeptideAtlas lists names only where they are present in the public post or in the retained research materials and labels older versus newer map evidence.
The warehouse evidence establishes shared fulfillment infrastructure or inventory access, not a shared synthesis facility. It should not be used to claim common factory ownership without additional evidence.
The pricing comparison is point-in-time. Supplier quote sheets and retail prices can change rapidly. "Implied gross spread" is retail price less quoted supplier price divided by retail price; it is not net margin and does not deduct shipping, losses, testing, labor, tax, advertising, merchant fees or other costs.
The Intelligen semaglutide manufacturing figure is a conceptual techno-economic model for a 500kg/year industrial API facility. It is used only as a benchmark for API-production economics; it is not a measurement of the cost structure of any grey-market supplier.
This report provides no sourcing, purchasing, administration or dosing guidance for unapproved products.
The documentary exhibits behind sections IV, VI and VII — the warehouse ledgers and the wholesale price sheets — are reproduced in full on the evidence appendix page.
Tier 1 — primary records
Tier 2 — published reporting and research
Tier 3 — community analysis, attributed as such
Tier 4 — PeptideAtlas original research retained on file
PeptideAtlas is not affiliated with Janoshik Analytical, Finnrick Analytics, JEEP, WBS, ERP, Homopeptide, LJ2, Eli Lilly or any vendor named in this article. We do not sell peptides, testing, referrals or vendor placements, and we accept no vendor sponsorship. Nothing here is medical, legal or purchasing advice. Vendor names are included for documentary and research purposes, with evidence levels stated in context.
Related: The COA Mirage, Supply-Chain Evidence Appendix.