A Tonne of Duck in One Run: What Parima and Vow Actually Proved

If you saw the headline last week (“cultivated meat produced at 99% lower cost”, “the world’s largest food-grade bioreactor”, “the final barrier removed”) and thought “right, but what does any of that mean at the bottom of a 22,000-litre tank,” this post is for you.

Thank you for reading this post, don't forget to subscribe!

I spend a fair amount of my time building techno-economic models for exactly this kind of process. So when a cultivated meat milestone lands wrapped in three superlatives and a percentage, my first instinct isn’t to celebrate or to sneer. It’s to go looking for the mass balance. Let me read the Parima and Vow announcement the way I’d read it if someone handed it to me across a bench: which numbers are load-bearing, which are decoration, and what the whole thing tells us about where this industry has quietly moved.

What actually happened

France’s Parima (the parent company of Gourmey, cultivated duck, and Vital Meat, cultivated chicken) ran a tonne-scale batch of duck in a single run on Australia’s Vow 22,000-litre production line, which Vow describes as the largest food-grade cell-culture bioreactor in the world. According to Parima, the run came in at 99% lower cost than their earlier production, hit its numbers on the first attempt, and showed no performance loss relative to smaller scales. The two companies started collaborating in 2025; this was the payoff.

That’s the news. Now the engineering.

The number everyone’s quoting is the wrong number

Let’s deal with the 99% first, because it’s the one doing the heavy lifting in every headline and it’s the one you should trust the least.

A 99% cost reduction sounds like a law of physics has been repealed. It isn’t. It’s a reduction relative to Parima’s own earlier runs, and everyone in this field knows that early cultivated-meat runs were priced somewhere between “absurd” and “please don’t ask.” When your starting point is a few thousand dollars per kilo, knocking off 99% is genuinely impressive engineering and still leaves you needing to explain where you actually landed. Relative reductions off a huge base make wonderful press releases and terrible planning inputs.

The honest numbers are the absolute ones, and to Parima’s credit they (and their analysts at Arthur D. Little) put them on the table. Finished-product cost in the low €40s per kg in 2026, with a modelled path to roughly €10/kg by the end of the decade, about a 75% reduction from here. A separate ADL figure floats a floor of $3.43/lb (roughly €7/kg) for this kind of setup. Those are the numbers I’d build a model around. Not the 99%.

Here’s the trajectory as the analysts frame it, with my translation in the third column:

Milestone	Figure cited	What it means on the floor
Today’s finished cost	low €40s / kg	Fine for premium, nowhere near mince
Modelled floor for this setup	~$3.43/lb (~€7/kg)	A model, not an invoice
End-of-decade target	~€10/kg	Requires volume, offtake, cheaper cell feed
Food-grade growth media	approaching €0.20 / L	The genuinely startling one, see below
Cell densities achieved	55 to 100 g/L	Microbial-fermentation territory, not fragile-mammalian

The real lever: they took the expensive molecules out of the media

If you want to understand cultivated meat economics, ignore the tank for a minute and look at what’s dissolved inside it. Media has always been the cost killer. In a conventional animal-cell process, the growth medium, and specifically the recombinant proteins in it, routinely accounts for the overwhelming majority of the cost of goods. Growth factors like FGF2, recombinant albumin, insulin: these are the molecules that make a litre of medium cost more than a litre of decent whisky.

Buried in the Parima announcement, past the tank-size bragging, is the sentence that actually matters: the cost gains came from removing the most expensive compounds, namely growth factors, albumin, and insulin. That’s not a tweak. That’s pulling out the exact line items that have kept every honest cultivated-meat TEA stubbornly in the red for a decade.

And then there’s the ADL figure I keep coming back to: food-grade growth media approaching 20 cents per litre. If that number is real and reproducible at volume, it reframes the entire problem. For years the sceptic’s trump card was “yes, but the media will never be cheap enough.” Twenty cents a litre is the sound of that card being quietly put back in the deck. I want to see the formulation and the assumptions behind it before I fully believe it (more on my scepticism later), but if it holds, this is the breakthrough, not the bioreactor.

The quiet thesis: cultivated meat just became a fermentation problem

Here’s the part that made me sit up, and it’s the part nobody’s putting in a headline because it doesn’t photograph well.

Parima grows its cells in suspension, non-GMO, as undifferentiated biomass. No genetic modification, no scaffolds, no coaxing cells into structured tissue. Their own description is that it works “in a manner similar to biomass fermentation.” Read that again if you’re a bioprocess engineer, because it’s a bigger claim than the tonne of duck.

For most of the last decade, cultivated meat was really a tissue-engineering project wearing a bioprocess costume. The dream was structured tissue: scaffolds, differentiation, perfusion of delicate anchorage-dependent cells into something with the grain of a steak. It’s beautiful biology. It also doesn’t scale, because every one of those requirements fights you the moment the vessel gets big.

Growing undifferentiated biomass in suspension is a completely different animal, and one I actually know how to run. It means you get to use the entire playbook of industrial microbial fermentation: high agitation, aggressive aeration, high cell density, the same Monod kinetics and mass balances I lean on for a lysine run. The tell is right there in the density figures: 55 to 100 g/L. That’s not fragile-mammalian-cell territory (a good CHO perfusion is measured in cells per mL and treated like fine china). That’s the same order of magnitude I’d expect from a robust microbial fermentation. When your “meat cells” behave like well-mannered production microbes, scale-up stops being a moonshot and starts being an engineering exercise with known failure modes.

Two paradigms, side by side, since this is the whole story:

	Tissue-engineering approach	Fermentation-style approach
Product	Structured tissue	Undifferentiated biomass
Cells	Anchorage-dependent, delicate	Suspension-adapted, robust
Scaffolds	Required	None
Media	Growth-factor heavy	Growth factors removed
Scale-up	Fights you at every litre	Uses the microbial playbook
Who can run it	Specialist tissue engineers	Any competent bioprocess team

Cultivated meat spent ten years trying to be the first column. Parima’s result is a bet on the second, and the economics are following the bet.

Why “first attempt, no performance loss” is the impressive part

Scale-up is where cultivated meat usually goes to die, so let me give the 22,000-litre claim its due, and its caveat.

At that volume, the things that quietly kill a process aren’t exotic. They’re the same four horsemen every scale-up engineer loses sleep over. Oxygen transfer (your k_La doesn’t scale linearly, and a strict aerobe at high density is thirsty). Mixing time (blend time gets longer as the tank grows, so cells experience the vessel as a set of neighbourhoods, not one uniform environment). Gradients in dissolved oxygen, pH, and substrate: pockets that a cell at the top never sees the same way as a cell near the sparger. And dissolved CO₂ accumulation, the silent titre-killer in big vessels, because you can’t strip it out as easily as you’d like.

Getting to 22,000 L on the first attempt with no performance loss is exactly the claim that should be hard to believe, which is precisely why the suspension, high-density, robust-cell strategy matters. You can hammer a tough suspension culture with the agitation and aeration needed to keep k_La up and CO₂ down. Try that with anchorage-dependent cells on scaffolds and you’d shear them into soup. The believability of the scale-up claim is downstream of the boring biological choice they made years earlier. That’s usually how it works: the impressive result at the end was bought with an unglamorous decision at the start.

Caveat, and I’ll say it plainly: one successful run is a demonstration, not a manufacturing process. I’ve watched too many “it worked!” batches fail to reproduce three runs later. A tonne in a single run proves the process can close. It doesn’t prove it closes reliably, at GMP, batch after batch, which is the only thing a supply chain actually cares about.

The business model is the other half of the story

There’s a second engineering decision hiding in here that has nothing to do with cells, and it’s just as important.

Vow has pivoted from making its own finished food to co-manufacturing for others, renting out that big iron as a service. Parima, meanwhile, keeps what it’s actually good at: the cell lines, the media, the process. It manufactures in-house in France up to around 2,000 L (400 L bioreactors in Paris, 2,000 L running daily near Nantes, a 5,000 L fermenter on the side) and rents the commercial-scale capacity from partners like Vow for the big runs.

Pharma worked this out decades ago. It’s the CDMO model: the company that owns the molecule doesn’t necessarily own the factory. Parima’s CEO framed it as the shift from early-stage vertical integration (where you build everything yourself because the supply chain doesn’t exist) to specialisation as an industry matures. He’s right, and it matters more than it sounds, because it decouples the product from the capex. When cultivated-meat funding halved in 2025, “you don’t have to raise €200M to build a factory before you can sell a gram” stopped being a nicety and became survival. Capital efficiency is a feature now, not a footnote.

The go-to-market is quietly clever, too

One more thing the TEA lens makes obvious: leading with duck, and with Gourmey’s foie gras, is not an accident. At a finished cost in the low €40s per kg, you cannot sell mince (you’d be laughed out of the chiller aisle). But you can absolutely sell a premium delicacy that already commands those prices and comes with genuine ethical baggage in its conventional form.

Premium-first is the correct sequencing. You enter where the cost structure already works, bank the revenue and the learning, and ride the cost curve down toward the mass market as media and volume improve. It’s the same reason new technologies tend to show up in luxury goods before they show up in commodities. The honest TEA supports this ordering; the marketing just happens to agree with the maths for once.

Where I’d hold my applause

I like this result. I also build models for a living, which means my job is partly to be the person at the table who asks the annoying questions. So:

• One run isn’t a process. Give me three consecutive runs at 22,000 L within spec and I’ll get properly excited.
• Where’s the downstream? Cultivated-meat announcements love the upstream titre and go strangely quiet on harvest, dewatering, and formulation. Those are real unit operations with real cost, and “tonne-scale biomass produced” is not the same as “tonne-scale product shipped.”
• The 20-cents-per-litre media: under what assumptions? At what volume? With what purity spec? A media cost is only meaningful with its bill of materials attached, and I don’t have it.
• Regulatory reality. Parima has approvals in Singapore, but is still waiting on FSANZ (the Australian timeline has already slipped), the EU, the UK, Switzerland, and the US. The biology may be solved; the paperwork is very much not.
• “No performance loss” needs a mass balance. I’d want the yield on substrate, the CER trace, the DO profile, the actual media formulation. Those are the numbers that turn a press release into an engineering claim. Until I see them, I’ll file this under very promising rather than proven.

None of this is cynicism. It’s the checklist. A milestone earns its adjectives when it survives the checklist.

The bit I keep relearning

Parima also partnered with an AI company, DeepLife, to build an avian digital twin for optimising production, which is catnip for someone who spends his evenings putting ML and model predictive control on top of bioreactors. But notice the order of operations, because it’s the whole lesson.

The digital twin isn’t what made this work. What made this work was a stack of deeply unglamorous fundamentals: a robust cell line that grows in suspension, a media formulation with the expensive proteins engineered out, and enough respect for oxygen transfer and mixing to hold performance from a few litres to twenty-two thousand. The AI layer earns its keep after that foundation is solid. It helps you run a good process better; it cannot rescue a bad one.

I said something similar at the end of a lysine post a while back and I’ll stand by it here: the biology does the work, and our job is to create the conditions for it to do its best work. Cultivated meat spent a decade being run as a tissue-engineering moonshot. It’s finally being run as a bioprocess (media cost, oxygen transfer, mass balance, a tough cell line in a big tank), and that, far more than any single tonne of duck, is why the numbers have started to move.

Get the conventional part right first. The models come after. They always did.

---

Sources: Green Queen, “Parima Validates Commercial-Scale Cultivated Meat Production with Vow” and Vow (eatvow.com). Cost trajectory figures attributed to Arthur D. Little via Green Queen’s reporting.