Human milk oligosaccharides are the closest thing precision fermentation has to a finished commercial story. Seven distinct molecules authorised in the European Union across 28 regulatory instruments in seven years. Ten separate notifiers cleared in the United States across the two waves below. A rise from 4% to 20% of global infant formula launches between 2018 and 2023.
Every operator in alternative protein should look closely at how that happened. But the most useful thing in the record is not the growth curve. It is what has happened to the supply base in the last eighteen months.
Of the ten HMO notices that have completed at the US Food and Drug Administration in its current inventory listing, eight came from Chinese companies. One of those eight was rejected outright. Meanwhile not one of the established Western producers has ever published a capacity figure for this category, while the Chinese entrants are publishing tonnages, build costs, and — in one case, in answer to an investor question on a stock exchange platform — a price.
The price has roughly halved.
The US record, in two waves
The founding record is European. The recent record is not.
The foundational notices, 2014–2018
| GRN | Substance | Notifier | Closed |
|---|---|---|---|
| 546 | 2’-O-fucosyllactose | Glycom A/S (Denmark) | 16 September 2015 |
| 547 | Lacto-N-neotetraose | Glycom A/S | 2 October 2015 |
| 650 | 2’-O-fucosyllactose | Glycom A/S | 23 November 2016 |
| 659 | Lacto-N-neotetraose | Glycom A/S | 23 November 2016 |
| 735 | 2’-Fucosyllactose | Glycosyn LLC and FrieslandCampina Domo B.V. | 6 April 2018 |
| 749 | 2’-O-fucosyllactose | DuPont Nutrition & Health | 23 April 2018 |
The 2025–26 wave
| GRN | Substance | Notifier | Closed | Outcome |
|---|---|---|---|---|
| 1238 | 2’-fucosyllactose | Cataya Bio (Shanghai) | 17 June 2025 | No questions |
| 1274 | 2’-Fucosyllactose | Tianjin Hesheng Biotechnology | 8 December 2025 | No questions |
| 1261 | 2’-Fucosyllactose | Suzhou Yixi Biotech | 26 January 2026 | No questions |
| 1275 | 6’-Sialyllactose sodium salt | Cataya Bio (Shanghai) | 20 March 2026 | No questions |
| 1262 | 2’-Fucosyllactose | Zhuhai Long Health Biotechnology | 8 April 2026 | No questions |
| 1257 | 2’-Fucosyllactose | Chr. Hansen A/S (Denmark) | 16 April 2026 | No questions |
| 1258 | Lacto-N-tetraose | Chr. Hansen A/S (Denmark) | 16 April 2026 | No questions |
| 1299 | 3’-sialyllactose sodium salt | Synaura Biotechnology (Shanghai) | 13 May 2026 | No questions |
| 1281 | 2’-Fucosyllactose | CABIO Biotech (Wuhan) | 8 May 2026 | Not a basis for a GRAS determination |
| 1298 | 2’-fucosyllactose | Beijing Zeno Technology Development | 23 June 2026 | No questions |
Two details deserve attention.
GRN 1281 is a rejection. FDA told CABIO Biotech that its notice did not provide a basis for a GRAS conclusion — seven weeks before clearing two other Chinese notifiers for the same molecule. Rejections in this inventory are rare enough to be worth tracking; this one establishes that the agency is not simply rubber-stamping the wave.
The US inclusion ceiling has not moved. 2.4 g/L of 2’-FL in non-exempt term infant formula runs consistently from GRN 546 in 2014 through GRN 1298 in 2026. Twelve years, a dozen notifiers, one number. Which means every new clearance is a supply-side entry into a market whose per-unit demand is fixed by regulation.
Practical consequence: in a category with a regulated maximum inclusion rate, additional approved suppliers cannot expand volume. They can only compete on price. That is the whole mechanism behind what follows.
Europe: 28 instruments, and a correction worth making
The EU record is the most complete regulatory dataset in precision fermentation, and it is routinely misdescribed.
Seven distinct HMO molecules have been authorised or amended across roughly 28 Commission Implementing Regulations between 2019 and February 2026 — 2’-FL, LNnT, the 2’-FL/DFL mixture, LNT, 3-FL, 3’-SL and 6’-SL, plus an LNFP I mixture in 2024. A large share of those instruments are not new authorisations but additions of new production strains — E. coli K-12, BL21(DE3), W (ATCC 9637), MG1655, and Corynebacterium glutamicum ATCC 13032 — each one opening the molecule to another manufacturer.
That is the regulatory machinery of commoditisation, visible in the Official Journal.
One correction to a widely repeated claim: EU novel food authorisations are generic by default, not applicant-specific. The Commission’s own page states that the Union list “serves as a reference for economic operators who wish to place in the market an authorised novel food unless data protection is requested by the applicant.” Exclusivity under Article 26 of Regulation (EU) 2015/2283 is an opt-in, not the standing rule.
The contrast that makes the point
Now set the HMO record beside precision-fermented animal proteins — the same regulation, the same fermentation technology, largely the same regulators.
In the US, the proteins do fine. Perfect Day’s beta-lactoglobulin cleared as GRN 863 in March 2020; Remilk (GRN 1056), Imagindairy (1145), Vivici (1200) and Shanghai Changing Biotech (1247) followed. Clara Foods cleared ovomucoid (967) and pepsin A (1025), The EVERY Company ovalbumin (1104), Onego Bio ovalbumin (1249), TurtleTree (1219) and All G (1284) bovine lactoferrin. Fermify withdrew a beta-casein notice (1228); a Formo/NewMilkBuzz αS1-casein notice (1312) is pending.
In the European Union, not one precision-fermented animal protein has been authorised. There are two applications on the Commission’s register: NF 2023/19996 (Remilk’s beta-lactoglobulin) and NF 2024/32870 (a human-equivalent lactoferrin, applicant anonymised). Neither is authorised. This is consistent with what we found when we counted Europe’s precision fermentation approvals.
A trap worth flagging. Commission Implementing Regulation (EU) 2022/2534, applicant Arla Foods Ingredients, authorises “bovine milk beta-lactoglobulin” and is frequently miscited as the EU’s first approval of a precision-fermented dairy protein. Read the specification: the substance is produced “from bovine whey by a series of steps involving filtration, concentration, crystallisation.” It is a whey fraction. It is not fermentation-derived.
In the UK, Perfect Day’s beta-lactoglobulin dossier RP1571 was submitted in April 2022 and was still on the ACNFP agenda in April 2026. Four years, unresolved.
So the asymmetry is not about the regulatory route — HMOs used the identical pathways. It is about the molecule and the market.
Why the sugars cleared and the proteins did not
Three structural reasons, and only the third is really about regulation.
They are used in milligrams. FDA’s own record in GRN 924 notes that oligosaccharide content in US infant formula is “100 to 1000 times lower” than in human milk. Commercial inclusion runs around 1.6% by weight of powder. An HMO manufacturer never had to solve the bulk scale-up problem, because the category never posed it.
The application was singular and pre-identified. One product category, a small number of multinational buyers, a specification written by the customer. Compare that to a precision-fermented casein looking for a route to market across cheese, beverage and bakery applications with different functional requirements in each.
They were nutrients, not replacements. An HMO is added to formula. A recombinant casein displaces an incumbent dairy ingredient — and the incumbent has a supply chain, a lobby, and in the EU a set of compositional standards written around it.
Practical consequence: the HMO playbook transfers to specialty ingredients used at low inclusion in a defined application. It does not transfer to bulk protein, and citing HMO success in a bulk-protein investment case is a category error.
The counter-example that breaks the “first to scale” story
It is tempting to call HMOs the first precision-fermented ingredient to reach genuine scale. That is wrong, and the correction is useful.
Fermentation-produced chymosin has been GRAS in the United States since 23 March 1990. 21 CFR 184.1685 affirms chymosin preparation from recombinant E. coli K-12, extended to Kluyveromyces marxianus var. lactis in February 1992 and Aspergillus niger var. awamori in May 1993. Its conditions of use are as permissive as the code gets: “used in food with no limitation other than current good manufacturing practice.”
A recombinant protein, cleared for direct human food use with no numerical limit, across the entire cheese category, thirty-six years ago — a quarter-century before the first HMO authorisation. Made, then as now, by the same two companies that dominate HMOs.
Chymosin is also the better analogy for precision-fermented dairy protein, because it displaced an animal-derived input in a bulk dairy application and won on cost and consistency. Anyone arguing that recombinant dairy proteins face an insurmountable path should explain why chymosin does not falsify the claim. Anyone arguing HMOs prove the path is easy should explain why it took thirty-six years to get from chymosin to a second such case.
The supply base inverted, and the price moved
Here is the part that is not being reported.
What the Chinese entrants disclose
| Company | Disclosed capacity | Investment | Date |
|---|---|---|---|
| Suzhou Yixi Biotech (一兮生物) | Phase 2 line 3,000 t/yr; 3,500 t/yr total; seven HMOs industrialised | RMB 200m | Start-up April 2026 |
| Baolingbao Biology (保龄宝, SZSE 002286) | 2,500 t/yr — “basically built,” small-batch production | Undisclosed | 14 November 2025 |
| Zhuhai Long Health (珠海朗健) | Phase 1 260 t/yr; Phase 2 740 t/yr | Undisclosed | Reported May 2026 |
What the Western incumbents disclose
| Company | Capacity in tonnes | Most specific location disclosure found |
|---|---|---|
| dsm-firmenich (Glycom) | None found | Esbjerg, Denmark |
| Novonesis (Chr. Hansen / Jennewein) | None found | Bonn / Rheinbreitbach, Germany; a Rayong, Thailand site acquired from Meihua |
| IFF | None found | Not disclosed; historic contract manufacture with Lonza |
| BASF | None found | “German-based production,” site unnamed |
| FrieslandCampina Ingredients | None found | A dedicated 2’-FL plant “being built in Italy” (2018) |
| WACKER / BENEO | None found | “Production will be based in Europe” (19 May 2025) |
The only tonnage figure attached to a non-Chinese HMO producer that we could verify is 300 t/yr at Thai Kyowa (Kirin) in Rayong, live since November 2022.
These are among the largest ingredient companies in the world, several of them listed, all of them competing in the same category on the same molecules. Between them they disclose one capacity number, and it belongs to none of them.
The price
In November 2025, Baolingbao Biology answered an investor question on the Shenzhen exchange’s interaction platform. Its stated price for 2’-FL: RMB 400,000 to 600,000 per tonne — RMB 400 to 600 a kilogram — with the company noting prices “change dynamically.”
Chinese financial press reporting the same week described the pre-entry level, when supply was dominated by dsm and Chr. Hansen, as an imported raw material price “as high as RMB 1,000 per kilogram,” nicknamed 奶黄金 — milk gold — and reported that domestic cost reduction had “driven [it] below RMB 500,000 per tonne.”
At roughly 7.1 yuan to the dollar that is about US$56 to US$85 a kilogram, down from around US$140. A fall of 40% to 60%, disclosed by a listed company on an exchange platform, in a category where no Western producer has ever published a price at all.
Practical consequence: if you are formulating with HMOs, the contract you signed against 2023 pricing is now materially off-market, and there are at least six newly cleared US suppliers. If you are selling them, the transparency asymmetry is no longer working in your favour.
Where that puts HMOs against commodity protein
The obvious question is how a halved HMO price compares to bulk dairy protein. The answer is less dramatic than expected, and the reason is that both ends are moving.
From USDA Agricultural Marketing Service Dairy Market News for the week of 10–14 August 2026:
| Product | Range | Approx. per tonne |
|---|---|---|
| WPC 34% | $1.88–2.75/lb, mostly $2.00–2.24 | ~$4,400–4,900 |
| WPC 80% | $12–13/lb | ~$26,500–28,700 |
| Whey protein isolate | $14 to upper $14s/lb | ~$30,900–32,900 |
| Acid casein (US imports) | $4.25–5.25/lb | ~$9,400–11,600 |
| Rennet casein | $4.40–5.35/lb | ~$9,700–11,800 |
And from the EU Milk Market Observatory, week ending 7 June 2026: whey powder at €143 per 100 kg — €1,430 a tonne, up 45% year on year.
So 2’-FL at roughly US$70/kg sits about 2.5 times WPC 80% at about $27/kg, and about 15 times WPC 34% at about $4.6/kg.
We are deliberately not turning that into a headline multiple. The two figures are not like-for-like: one is a Chinese domestic price during an aggressive local capacity build-out, the other is US spot in an unusually tight whey market. The defensible statement is narrower and more interesting: the gap is closing from both ends simultaneously — HMO prices falling on new entry, whey protein prices rising sharply. That is a more useful fact for a formulator than any single ratio.
The counter-argument: the end market is contracting
The strongest case against reading HMOs as a growth template is that the category is a supply-side success story attached to a shrinking demand base.
The buyer is deprioritising it. Reckitt reports Mead Johnson Nutrition under a heading titled “non-core,” with FY2025 revenue of £2,119m and like-for-like growth of 3.8% decomposed as volume −2.3%, price/mix +6.1% — and Q4 volume at −4.6%. MJN’s 61.1% gross margin is below Core Reckitt’s 62.2%. Infant formula is not even the group’s best-margin business.
The demographics are unforgiving. China’s National Bureau of Statistics recorded 7.92 million births in 2025, a crude birth rate of 5.63 per thousand, and a population decline of 3.39 million. a2 Milk reports the China-label infant formula market declined 5.6% in FY25.
In the largest single market, the buyer is a monopsony and prices are falling. GAO found that more than half of all US infant formula is purchased by state agencies through WIC, that federal law requires single-supplier competitive contracts, that two manufacturers held almost all WIC contracts as of August 2024, and that inflation-adjusted US formula prices fell 11% between 2020 and 2023. The “high price point” that made HMOs work is under pressure from the demand side too.
The efficacy base is contested. A review of twelve randomised trials in Advances in Nutrition (April 2026) concluded that evidence for clinical benefits of human-identical milk oligosaccharides is “limited,” with studies “heterogeneous and [having] key limitations.” A 2023 BMJ survey of 608 infant formula products across 15 countries found that 74% provided no scientific reference for their health claims, and that 90% of claims citing registered trials carried a high risk of bias.
Put together: falling ingredient price, contracting birth cohort, consolidated and price-sensitive buyers, and a benefit narrative under academic pressure. A supplier entering now is entering late.
What we could not establish
- Any Western or global transaction price for 2’-FL. The only attributable prices in existence are Chinese. Every “EUR 350–500/kg” style figure we found traces to vendor market-research listings with undisclosed methodology, and we have not used them.
- Any capacity figure for dsm-firmenich, Novonesis, IFF, BASF, FrieslandCampina or WACKER. Not “we could not find one this week” — we could find no instance of any of them ever publishing one.
- The acquisition and capex figures for this category from a source we could retrieve. Reported values for DSM’s purchase of Glycom and Chr. Hansen’s purchase of Jennewein, and Novonesis’s investment in the Rayong site, are widely repeated, and we removed them from an earlier draft of this piece because we could not resolve a working URL for the original announcements. They are omitted rather than cited loosely. Nothing above depends on them.
- The Chinese regulatory chronology from a primary document. NHC Announcement No. 8 of 2023 (2’-FL and LNnT as nutrient fortifiers, 7 October 2023), a February 2026 approval for 3’-SL, and NHC Announcement No. 1 of 2026 widening permitted production strains, all rest on IFF’s own press release plus Chinese trade press. We fetched no NHC or CFSA document.
- The duration of the EU Article 26 data protection period. It is standardly five years; we did not read that on a fetched page and are not asserting it.
- Whether HMO businesses are profitable. No company discloses HMO-level P&L. The claim that this category is a commercial success therefore rests entirely on data nobody outside the companies has seen.
- A reliable market-size figure. The two most-circulated numbers measure different things — roughly US$180m for the ingredient in 2022 versus roughly US$4.2bn for finished HMO-containing formula in 2025. Both are vendor projections. Placing them side by side without stating the basis is precisely the measurement error that makes this sector’s figures incomparable.
- A claim that Novonesis is divesting its HMO business. We searched for it specifically and found nothing; the evidence runs the other way, since it bought HMO-capable capacity in Thailand in April 2026. It appears to be a conflation with dsm-firmenich’s Animal Nutrition & Health divestment.
- A published lactoferrin price index. None exists. The widely quoted US$300/kg figure refers to 2011.
What to watch
- Whether any Western HMO producer responds to Chinese price disclosure with disclosure of its own. Opacity is a defensible strategy only while you set the price.
- Whether FDA rejects a second Chinese HMO notice. GRN 1281 is currently a single data point. Two would suggest a dossier-quality pattern; one may be an outlier.
- Whether a precision-fermented animal protein completes EU novel food authorisation before the end of 2027. Remilk’s application, NF 2023/19996, is the leading candidate and the cleanest test of whether the HMO route generalises.
- Whether the 2.4 g/L US inclusion ceiling moves. It has not in twelve years. If it does, the supply-demand arithmetic in this category changes overnight.
Our position, stated for later checking: HMOs are a worked example of getting a precision-fermented ingredient to market, but the conditions that made it work — milligram inclusion, one application, a fixed regulatory ceiling, and buyers who specified the product — are largely absent for bulk protein. The most transferable lesson is not the growth. It is what happens eighteen months after a category becomes contestable: the price disclosure comes from the challengers, and it comes down.