On 16 April 2026 Ajinomoto announced it had replaced transferrin in serum-free cell culture media with hinokitiol, a low-molecular-weight plant compound. The company’s own release describes transferrin as “difficult to manufacture and a significant driver of culture media costs.” It puts no percentage on it.
Coverage supplied one. Green Queen’s write-up states that albumin and transferrin “have been found to account for 95% of the costs associated with manufacturing cultivated meat,” and hyperlinks the claim to a Good Food Institute analysis.
We fetched that analysis. It says something different, about a different denominator:
“Reducing the cost of growth factors is indeed critical in early R&D because they currently comprise over 95% of the cost of representative animal component-free media formulations.”
Growth factors, not albumin and transferrin. Media cost, not manufacturing cost. And in the cost table inside that same document, transferrin is $85,600 of a $7,535,958 batch — about 1.1%.
This is not a gotcha about one outlet. It is a load-bearing number in how the sector’s input supply chain gets discussed, and getting it wrong sends procurement attention to the wrong line item. vegconomist ran the same Ajinomoto story the same day and used no percentage at all, which turns out to have been the right call.
The table the 95% comes from
GFI’s analysis, authored by Liz Specht and last updated 9 February 2020, models Essential 8 medium at a 20,000-litre batch. Reproduced in full:
| Component | Concentration (mg/L) | Cost per g | Supplier | Cost per 20,000 L | Share of medium cost |
|---|---|---|---|---|---|
| DMEM/F12 basal | 1X | $156 for 50 L | Thermo Fisher | $62,400.00 | 0.8% |
| AA2P | 64 | $7.84 | Cayman Chemicals | $10,035.20 | 0.1% |
| Sodium bicarbonate | 543 | <$0.01 | Alibaba | $2.39 | <0.1% |
| Sodium selenite | 0.014 | $0.10 | Alibaba | $0.03 | <0.1% |
| Insulin | 19.4 | $340.00 | Sigma | $131,920.00 | 1.8% |
| Transferrin | 10.7 | $400.00 | Sigma | $85,600.00 | 1.1% |
| FGF-2 | 0.1 | $2,005,000.00 | R&D Systems | $4,010,000.00 | 53.2% |
| TGF-β | 0.002 | $80,900,000.00 | R&D Systems | $3,236,000.00 | 42.9% |
| Total | $7,535,958 | 100% | |||
| Cost per litre | $376.80 |
Component costs and the total are as published. The share column is our calculation from those figures. The document’s own summary agrees with the arithmetic: “just two — those present at the lowest concentrations, FGF-2 and TGF-b — account for almost all of the cost of the growth factors, comprising over 96% of the total cost of the medium in this base case.”
Two caveats the document states and citers drop. The pricing “data were gathered in April 2017” — nine years old. And the transferrin line is “recombinant, expressed in rice (not purified from serum),” so it is one production route at one supplier’s bench list price, not a market price.
The peer-reviewed literature reproduces the same structure. Quek et al., writing in npj Science of Food in 2024, state that “in the Essential 8 medium, nearly 98% of the media cost can be attributed to FGF-2 and TGF-β,” and that for the beefy-9 formulation “albumin, FGF-2, and insulin, collectively representing around 60% of the total media cost.”
Practical consequence: if you are modelling media cost, the sensitivity lives in FGF-2 and TGF-β at bench pricing, and in albumin at scale. A media formulation that eliminates transferrin and changes nothing else moves roughly one point of the medium bill.
Where the 95%-of-manufacturing-cost idea does have a source
The confusion is not invented from nothing. The same GFI document contains a separate, explicitly hedged claim about total product cost:
“Estimates we have heard from these sources range from 55% to over 95% of the marginal cost of the product attributable to the cost of the medium.”
So “95%” and “manufacturing cost” do appear together — describing medium as a share of product cost, sourced to unnamed expert consultation and given as the top of a range that starts at 55%. Quek et al. put the same relationship more conservatively: serum-free media “makes up at least 50% of variable operating costs in cultivated meat manufacturing.”
Two true statements — growth factors are ~95% of media cost; media may be 55–95% of product cost — collapsed into one false one about transferrin.
Why transferrin still matters: volume, not cost share
The correction does not make Ajinomoto’s work unimportant. It relocates the reason.
GFI’s January 2023 analysis, led by Elliot Swartz with co-authors from Laurus Bio, Core Biogenesis, LenioBio and NTU FRESH, models the volumes of recombinant protein a cost-competitive industry would need:
“The analysis found that 96.6% of production volume is expected to be attributable to albumin, 2.42% to transferrin, 0.97% to insulin, and only 0.02% to all other growth factors.”
And the price targets:
“albumin would need to be produced at $10/kg, insulin and transferrin at $1,000/kg, and growth factors at $100,000/kg. Furthermore, cell culture media would need to be used efficiently, defined as 8 to 13 liters of media per kilogram of cultivated meat.”
Against that $1,000/kg target the report’s extrapolated price for transferrin is $642.98/kg on average, in a range of $1,639.22 to $94.30/kg — and it warns that “the costs for transferrin alone are higher than the 10% of the total production costs for cultivated meat in the high media use scenario.”
That is the honest case for attacking transferrin: at industrial volume, with growth factors engineered down or out, transferrin becomes a binding constraint on a $1/kg growth-factor-and-recombinant-protein budget. It is a scale problem arriving later, not the dominant cost today. Andrew Stout of Tufts named the same set from the bench side — companies building minimal media that “don’t use [expensive components such as] insulin or transferrin, maybe don’t use albumin, the real thorns in the side for media development right now.”
What Ajinomoto actually claims
The primary release is disciplined and worth separating from its coverage:
- Hinokitiol “binds to iron and delivers it into cells,” and Ajinomoto “confirmed that the function conventionally performed by transferrin can be substituted, maintaining high cell proliferation rates in serum-free media.”
- Results were “presented at the International Scientific Conference on Cultured Meat (ISCCM), held in November 2025.”
- The technology is “patent pending” and has “already undergone verification with prototypes,” with market introduction targeted “within the next several years.”
- Hinokitiol “is listed on Japan’s List of Existing Food Additives, confirming its safety for human consumption.”
- On performance, the figure is in the figure caption: conditions (b) and (c) “yielded approximately three times the cell volume of (a).” vegconomist rendered this as proliferation “approximately three times higher than in a control group without either compound.”
Ajinomoto also states, unusually plainly, that hinokitiol is chemically stable where “high-molecular-weight transferrin… is prone to quality fluctuation between production batches.” For a food manufacturer, batch-to-batch consistency in a media input is a qualification-cost problem as much as a unit-cost one — and it is the claim least likely to be overturned by better cost data.
Practical consequence: the regulatory position is the sharpest near-term signal here. An input already on Japan’s existing-additives list carries a materially shorter approval path in that market than a novel recombinant protein. That, not the cost share, is what a scale-up lead should be reading.
The input layer is being restructured in two opposite directions
The reason to care about any of this is structural. Cultivated meat’s input supply chain — cells, media, media components — is being reshaped by two moves that pull opposite ways on the same cost line.
Ajinomoto is monetising it. The release is explicit about the business, and it is not meat:
“As an initial step, Ajinomoto Co. will draw on the technology and expertise accumulated through its biopharmaceutical and regenerative medicine culture media business to pursue entry into the media component sale, culture technology development, and production support segments.”
An incumbent amino-acid and biopharma-media supplier is positioning itself as the arms dealer of the sector, on the same logic that made the pharma capacity argument we examined in the China capex claim so contested: existing regulated fermentation and media infrastructure is worth more to food than food can usually afford.
GFI is de-monetising part of it. On 16 October 2025 GFI announced it had bought SCiFi Foods’ biological assets out of the startup’s insolvency and given them away. Tufts Now, which carried the specifics, reports GFI “had the winning bid, buying eight cell lines and the recipes for two serum-free… media formulations competitively at auction.”
The chronology is worth stating precisely, because the announcement lagged the transaction by more than a year. Per GFI: “In late August 2024, that bid was accepted, and by mid-September, the cells and media were successfully transferred to TUCCA for storage and eventual distribution.” SCiFi had ceased operations in June 2024 after raising $40 million.
What was bought, per the Tufts Open Cell Bank listing: three lines are offered initially — TU-GFI-SCL1, TU-GFI-SCL3 and TU-GFI-SCL4, all Bos taurus connective-tissue fibroblasts, immortalised with TERT and P53 knockout, adapted to suspension. SCL3 was “reportedly used by SCiFi Foods in a submission to the FDA”; SCL4 was “reportedly cultured by SCiFi Foods in a 500L bioreactor format.” Three media are listed — FSF4, FSF6 and FSF6-EZ — of which FSF4 and FSF6 are flagged “not yet validated at Tufts.”
The replacement-cost case is made by the seller. SCiFi co-founder Joshua March: “It took us four years and tens of millions of dollars to develop these cells into commercial cell lines that grow quickly in suspension and in serum-free media.” GFI’s estimate of the avoided cost: cell line development “is estimated to cost between $2-10 million per startup,” so “for every 10 cultivated meat startups, access to these materials could save the industry anywhere from $20 to 100 million.”
GFI’s Elliot Swartz makes the standardisation argument that is the real prize: “CHO cells produce something like three quarters of the world’s biologic drugs because everyone had the same cell line in their hands and could be speaking this sort of common language… But there’s no standard cell line yet for cultivated meat.” On media, he adds: “Most media formulations are trade secrets and the formulations that are [publicly] available are designed for low-density adherent cell culture that’s really not commercial.”
Access is staged: academic researchers first, “with Tufts and GFI working to facilitate access to the cells by commercial entities at a later phase.” No date is given for commercial access.
This is the same asset class as the physical plant we tracked through Believer Meats’ two auctions, with the opposite outcome. Steel found no buyer at a price the estate liked. Cell lines found a buyer whose purpose was to remove them from the market entirely.
What media actually costs, as disclosed
The most useful thing an operator can take from the literature is the spread of disclosed per-litre figures, which is far wider than any single citation suggests.
| Figure | Basis | Source |
|---|---|---|
| ~$400/L | Essential 8, retail at bench-scale quantities | GFI/Specht |
| $376.80/L | Essential 8 modelled at 20,000 L, April 2017 list pricing | GFI/Specht |
| below $1/L | GFI projection with bulk, food-grade basal components | GFI/Specht |
| $0.63/L | “Believer Meats demonstrated that a serum-free medium can be produced at a cost as low as USD $0.63 per liter” | Quek et al., citing Pasitka et al. |
| $0.24/L | GFI’s most optimistic modelled scenario | GFI/Specht |
| $0.07/L | Company-disclosed, current, pilot scale. “That’s what we’re paying today, including our in-house growth factor production, water purification and mixing” — Paul Burridge, Clever Carnivore | AgFunderNews, June 2025 |
Three of those are models, one is a retail list price, and two are company disclosures. The $0.07/L figure is the most striking because it comes from a company that had raised $9 million and was running two 500-litre bioreactors — a fraction of the capital of firms quoting worse numbers. Note also that Believer Meats demonstrated $0.63/L media and shut down anyway: cheap media is necessary and not sufficient.
Set against GFI’s efficiency assumption of “8 to 13 liters of media per kilogram,” $0.07/L implies $0.56–$0.91 of media per kilogram of product, and $376.80/L implies $3,014–$4,898. Those are our calculations from the two sourced inputs, and the four-order-of-magnitude spread between them is the actual state of public knowledge. It is the same disclosure vacuum we found in Meatly’s medium cost disclosure and in the wider cost-data gap.
Practical consequence: treat any single quoted media cost as unusable without three qualifiers — bench or bulk pricing, food or pharma grade, and whether in-house growth-factor production is included. Clever Carnivore’s figure is low partly because it makes its own growth factors, which is a capability claim, not a purchasing claim.
The counter-argument
Several defences of the “95%” framing deserve stating properly.
First, it is directionally right about where cost sits: media dominates cultivated meat’s variable cost, and expensive recombinant proteins dominate media. A reader who came away believing “the expensive stuff is the proteins in the media” got the correct lesson from a wrong number.
Second, the underlying GFI page that Green Queen links for the biological role of transferrin does group albumin and transferrin together as proteins “used at much higher concentrations in cell culture than other recombinantly produced growth factors” — and by volume that grouping is exactly right, as GFI’s 2023 report confirms with albumin at 96.6% of anticipated volume. The error is a category slip between volume and cost, not fabrication.
Third, the 2017 pricing underlying our 1.1% figure is stale in both directions. Recombinant transferrin prices have not been published at scale since; a food-grade, plant-expressed transferrin at food volumes could be far cheaper than Sigma’s bench list, which would make transferrin’s share smaller still — or the reverse, if growth-factor costs have fallen faster. Nobody has published a current table, which is itself the finding.
What does not survive the counter-argument is the specific claim as printed: that albumin and transferrin account for 95% of manufacturing costs, cited to a document in which growth factors account for 95% of media cost and transferrin accounts for about 1%.
What we could not establish
- A current cost table. Every component breakdown we found traces to April 2017 bench pricing. No supplier or company has published a comparable table since.
- Any food-grade transferrin price. No price sheet located. GFI’s $642.98/kg average (range $1,639.22–$94.30) is an extrapolation, and its $1,000/kg is a target, not a quote.
- What GFI paid for the SCiFi assets. Not disclosed in the GFI release, the AgFunderNews report or Tufts’ coverage.
- How many groups have requested the cell lines, and the split between companies and academic labs. Widely repeated figures of “more than 20 groups” do not appear in any of the four primary or trade sources we fetched; we have therefore not used them. Only a waitlist is published.
- When commercial entities get access. Staged after academic access, with no date.
- Ajinomoto’s cost claim, quantified. The release claims “significant reduction” without a figure, a per-gram hinokitiol price, or a replacement ratio against transferrin. No pricing exists because the product is not on the market.
- Whether hinokitiol substitution holds at scale. Verification is stated at prototype level; no bioreactor volume is named.
- The ISCCM presentation itself. We relied on the company’s description of the November 2025 presentation; we did not obtain the underlying data.
- Two secondary sources we could not read. ScienceDirect’s 2026 media review returned an empty response to repeated fetches (live, bot-protected, not dead), and the Nature Reviews Bioengineering comment by Ho et al. is paywalled at $39.95 — we cite its abstract only and have taken no figures from it.
What to watch
- A hinokitiol price. Ajinomoto is entering media component sales. The moment it quotes a per-gram or per-litre figure, the transferrin line becomes checkable rather than argued.
- Whether anyone publishes a 2026 cost table. The nine-year-old Essential 8 table is doing work far beyond its design life. A current one would settle several disputes at once, including this one.
- Commercial access to the Tufts open cell bank. Academic access de-risks research; commercial access is what would actually compress startup cell-line spend by the $2–10m per company GFI cites. Watch for a date and for licence terms.
- Whether the 95% claim propagates. It is now in circulation attached to a reputable citation. The test of whether this sector’s numbers self-correct is whether the next round of coverage repeats it.
- Copycat asset rescues. Swartz says “other philanthropic nonprofit entities” are interested in “acquiring the assets of companies and sharing them openly.” Given the pattern of supply-layer shutdowns, there will be no shortage of estates to buy from.