Almost nothing in the precision fermentation cost stack is public. We have documented that at length: four techno-economic models in the whole literature, spanning under $20/kg to $15,000/kg, none of them sitting where the industry actually operates.
There is one exception. The carbon feedstock — the sugar that goes into the tank — has a quoted market price published by two governments on a regular schedule. It is the only input in this industry a stranger can look up.
So we looked it up, and then we checked it against the assumption used in the leading published model. Two things came out of that.
The first is that the widely repeated claim that feedstock accounts for around half of production cost does not survive contact with the sources. On the Good Food Institute’s own case-study assumptions, feedstock works out at roughly 19% of unit cost for a food-grade protein, and under 5% for a high-purity one.
The second is that the price of sugar is not really the variable that matters. The yield is.
Methodology and limits
Three limits govern everything below, and they should be read before the numbers.
We are comparing a sucrose price to a glucose assumption. The published government price series cover refined beet and cane sugar — sucrose. The techno-economic models assume glucose or dextrose, which is produced from corn starch and tracks the corn market rather than the sugar market. USDA does publish a corn-sweetener price series, but it is distributed as a spreadsheet we were not able to parse, and we are not going to substitute a vendor estimate for it. The sugar prices below are therefore an indication of the order of magnitude of industrial carbohydrate, not a like-for-like input price. Where that distinction changes a conclusion, we say so.
We have not converted euros to dollars. We did not source an exchange rate for the relevant period, so the EU and US figures are presented in their own currencies and not compared arithmetically.
The share-of-cost calculations are ours, derived from figures GFI publishes. GFI does not state a feedstock percentage in the form we give it. The derivation is shown in full so it can be checked or disputed.
What sugar actually costs
| Series | Price | Period | Publisher |
|---|---|---|---|
| Midwest refined beet sugar, spot (lower range) | 45 ¢/lb ≈ $992/tonne | 10 June 2026 | USDA ERS, citing Sosland’s Sweetener Report |
| Northeast refined cane sugar, forward range | 49–51 ¢/lb ≈ $1,080–$1,124/tonne | late May 2026 | USDA ERS, same source |
| US #16 raw sugar futures | 34.81 ¢/lb ≈ $767/tonne | April 2026 | Reported from ERS data by Southern Ag Today |
| World #11 raw sugar futures | 14.05 ¢/lb ≈ $310/tonne | April 2026 | Reported from ERS data by Southern Ag Today |
| EU white sugar, EU weighted average | €501/tonne | June 2026 | European Commission, DG Agriculture |
Per-tonne conversions from cents per pound are ours, at 2,204.62 lb per tonne.
The gap in the middle of that table is the thing an operator should notice. In April 2026 the US domestic raw sugar contract sat at 34.81 cents per pound while the world contract sat at 14.05 — the domestic price roughly two and a half times the world price. Nor is that a one-month artefact: over January to April 2026 the US #16 contract averaged 32.97 cents per pound.
That gap is not a market anomaly. It is the operating result of the US sugar programme, under which imports enter through tariff-rate quotas with a much higher duty applying above the quota. We have not cited a figure for that over-quota duty because we could not source one from the documents used here — and it is worth noting that USDA’s own import table shows substantial volumes still entering under the high-tier tariff, so it is not an absolute barrier.
Practical consequence: where you site a fermentation plant changes your carbohydrate bill before you have hired anyone. A plant inside the US tariff wall buys at the supported domestic price; a plant outside it buys at the world price. For a process consuming several kilograms of sugar per kilogram of product, that is not a rounding difference.
The claim that feedstock is half the cost
This claim is everywhere, and we could trace it to exactly one source: an August 2025 Synthesis Capital post, which states that “Feedstock costs can easily make up over half of the total cost per kilogram of end product.” No methodology, dataset or model is cited for the figure.
The two nearest GFI-published statements say something narrower and different.
From GFI and Hawkwood Biotech’s June 2025 review of 55 published techno-economic models: “Feedstock, raw feedstock processing, and raw material costs were the primary COP drivers in just over half of the published models”. That is a statement about how often feedstock ranks first across models — not about what proportion of cost it represents in any of them.
The same report, describing a private single-cell-protein case study using corn glucose at $0.40/kg, states that “Feedstock costs and CapEx together make up 51% of COP”. Feedstock and capital expenditure, combined.
Neither supports the claim that feedstock is half the cost. What they support is that feedstock is frequently the largest single line — a much weaker statement, and entirely compatible with feedstock being a fifth of the total.
Working the numbers instead
GFI Europe and Arthur D. Little’s July 2025 report models a food-grade alpha-lactalbumin case with two assumptions that can be combined: a substrate (glucose) cost of $0.60/kg and a target yield on substrate of 0.13 g/g. It gives a unit cost of $24/kg at food-grade purity (~95%) and $100/kg at high purity (>99%).
At 0.13 grams of product per gram of substrate, one kilogram of protein requires 7.7 kilograms of glucose. At $0.60/kg that is $4.62 of feedstock per kilogram of product — which is 19% of the $24/kg food-grade cost and 4.6% of the $100/kg high-purity cost. The substrate and yield assumptions are identical across both purity tiers in GFI’s model; only the downstream processing differs.
So on the most detailed public model available, feedstock is roughly one fifth of a food-grade unit cost, and the difference between food-grade and high-purity — which is downstream processing, not sugar — is worth four times as much.
Practical consequence: if you are trying to cut cost per kilogram in precision fermentation, negotiating a better sugar contract is one of the smaller levers available to you. Purification strategy and yield are much larger ones.
Why yield, not price, is the variable
The yield figures in GFI and Hawkwood’s model review are the most consequential numbers in this article, and they are strikingly wide.
| Process | Yield on feedstock | Sugar per kg of product |
|---|---|---|
| Precision fermentation, public models | 0.01–0.20 g/g | 5 kg – 100 kg |
| Precision fermentation, private models | 0.03–0.17 g/g | 5.9 kg – 33 kg |
| Fungi, continuous (biomass fermentation) | 0.3–0.5 g/g | 2 kg – 3.3 kg |
| Yeast, batch (biomass fermentation) | 0.5 and 0.65 g/g (two models) | 1.5 kg – 2 kg |
| Microalgae, batch | 0.28 g/g | 3.6 kg |
Sugar-per-kilogram figures are ours, being the reciprocal of the yield.
Read the top row against the bottom rows. At the pessimistic end of the published precision fermentation range — 0.01 g/g — a kilogram of product consumes 100 kilograms of glucose, which at $0.60/kg is $60 of feedstock alone, more than twice the entire $24/kg food-grade unit cost in GFI’s own case study. At the optimistic end, 0.20 g/g, it consumes 5kg and costs $3.
That is the whole argument. The price of sugar moves within a band of roughly two to three times between the world market and the US domestic market. The yield moves within a band of twenty times. Feedstock cost is dominated by the biology, not by procurement.
It also explains, without any further data, why biomass fermentation and precision fermentation have such different economics. A mycoprotein producer converting sugar at 0.3–0.5 g/g is using between a third and a twentieth as much substrate per kilogram of output as a precision fermentation producer expressing a single protein. That is not a difference of degree. Any comparison of “fermentation” costs that does not separate the two is comparing different industries — the same reconciliation failure we found when assembling capacity figures across four incompatible units.
Alternative feedstocks, and why substitution is slow
The obvious response to a feedstock bill is to buy a cheaper carbon source. GFI’s October 2023 sidestreams analysis identifies the leading candidates — corn stover, soy straw, sugarcane trash and bagasse, barley straw and husks — and is unusually candid about why they have not displaced refined sugar.
The technical barrier is that lignocellulosic material has to be broken down before it becomes fermentable, and the breakdown itself creates problems: “each sidestream lignocellulosic saccharification could produce compounds, such as phenolics and furfural, that could be inhibitory to cell culture.” The report also notes that pentose sugars such as xylose are not consumed until glucose is exhausted in many organisms, so that fraction of the carbon is effectively paid for and wasted.
The regulatory barrier is separate and slower: “For upcycled ingredients that do not have a robust history of human consumption, it will be necessary to evaluate their allergenicity, toxicity, and other relevant safety metrics.”
A more radical answer is to leave sugar out entirely. Solar Foods has done exactly that — its organism takes its carbon from carbon dioxide rather than from sugars, and the company has said publicly that in large installations electricity becomes the dominant share of its production cost. That is a genuine escape from the sugar market. It is also a substitution of one commodity exposure for another, and electricity prices are no more stable than sugar prices.
Practical consequence: treat sidestream feedstock as a research programme with a regulatory dependency, not as a near-term procurement option. The inhibitor chemistry has to be solved per sidestream, and the food-safety case has to be made per ingredient.
One useful non-finding
GFI’s model assumes the same substrate at the same price for both the food-grade and the high-purity tier. In other words, in that model, food-grade output does not require a special grade of input sugar — the purity requirement is met downstream. We had expected food-grade certification to constrain feedstock choice, and on this evidence it does not.
The counter-argument
Two objections deserve stating properly.
The first is that GFI’s $0.60/kg glucose assumption may simply be too low, in which case our 19% figure is too low as well. It is below the US refined sugar spot price in the table above — though, as flagged in the methodology note, that is a sucrose price rather than a glucose price, and dextrose from corn is not priced off the sugar programme. If the true delivered glucose price is closer to $1.00/kg, feedstock at 0.13 g/g becomes $7.69/kg, or 32% of a $24/kg unit cost. Still not half, but materially more than a fifth. Anyone modelling this should substitute their own delivered price rather than trusting either figure.
The second is that being the primary cost driver in just over half of published models is a perfectly reasonable thing for an industry to compress into “feedstock is half the cost” in conversation. The compression is wrong, but the underlying observation — that feedstock is where cost-reduction efforts most often get pointed — is not.
The third is the most serious, and it comes from GFI’s own sensitivity analysis of a single-cell-protein model. There, moving feedstock cost from a $0.40/kg base to $0.56/kg — a 40% increase — raises cost of production by 14.3%, while a 20% worsening of yield raises it by around 8%. Within one model at one operating point, in other words, feedstock price is not the minor term this article’s framing might suggest. That is a fair correction to make.
What it does not overturn is the cross-model finding. A twenty-fold range in grams of product per gram of substrate, observed across the published literature, is a far wider spread than any plausible dispute about the price of a gram of substrate. Local sensitivity within one model and dispersion across many models are different questions, and the second is the one facing anyone deciding whether a process can work at all.
What we could not establish
- A current US wholesale price for dextrose or high-fructose corn syrup per tonne from the USDA series. The relevant workbook is a binary spreadsheet we could not parse, and we declined to substitute a commodity-intelligence vendor estimate whose methodology is not published. This is the single biggest gap in this piece: it means the sucrose prices above cannot be converted into the glucose price the models actually use.
- Any fermentation protein company’s disclosed feedstock cost, or the identity of its sugar supplier. We looked specifically and found none. That absence is consistent with the broader disclosure pattern in this sector.
- The current over-quota (high-tier) duty rate on US sugar imports. We could not source it from the USDA documents used here and have therefore not published a number for it.
- What lies behind the $0.60/kg substrate assumption. The GFI Europe and Arthur D. Little report attributes it to “Industry interviews / triangulation”, which tells us it is drawn from what companies say they pay rather than from a published series — but not what contract terms, volumes or geography sit behind it.
- The exact percentage breakdown of raw materials in that report’s own charts; text extraction scrambled the chart labels, so we derived the share from the report’s stated substrate cost and yield instead, which is shown above and can be checked.
What to watch
- Whether anyone publishes a delivered feedstock price. One company disclosing what it actually pays per tonne, on what contract terms, would do more for this industry’s credibility than another round of cost-parity projections.
- Yield disclosures. Given that yield dominates the feedstock line by a factor of twenty, a company that publishes grams of product per gram of substrate is telling you more about its economics than one that publishes a cost per kilogram without a boundary.
- US plant siting decisions. With the domestic raw sugar contract at roughly two and a half times the world price, a fermentation plant choosing a US location is accepting a known and quantified input penalty. If US-sited capacity keeps being announced anyway, that tells you feedstock is not the binding constraint — which is exactly what the numbers above suggest.