Assuming we end factory farming before 2100, which factor will have contributed the most?

Alternative proteins

Social moral changes

Bullet point summary:

• Cultivated meat could have a price between $15/kg and $30/kg according to reputable technoeconomic analysis, which we review and explain.
• We present an interactive demand-side economic model where we translate that price to market share: https://pabloamc.github.io/Cultivated_meat/interactive.html
• Some species, like pork and especially chicken, are tough to replace with cultivated meat. Others, like cows and seafood, are more significantly more tractable. This could be good for shrimp and fish.
• Beachhead products exist, in particular foie-gras and high-end fish.
• Disruptive innovation economic theory says cultivated meat will have a hard time because there is no pressing problem that mainstream people perceive, but it is feasible in the same way electric cars are disrupting petrol cars.
• Some of the main advantages of cultivated meat include significantly addressing vegan attrition rates, by lowering the perceived social or health taxes. Long term, this should make conventional meat less and less socially acceptable: no longer a “necessary evil”.
• If electric cars are a good guide, government (or China) support will be important.
• In the end, the forecast is that both advocacy (to create demand) and cultivated meat (to provide supply) will need to work together to disrupt conventional factory farming. Either of those factors working alone would likely fail.

Screenshot of the model in https://pabloamc.github.io/Cultivated_meat/interactive.html, with knobs to tweak the parameters. I set plant-based products to be at the price and taste parity.

Summary: This blog post aims to recalibrate the expectations of cultivated meat, helping end factory farming. The blog contains two parts. The first part revisits high-quality technoeconomic analysis from the literature, updating the 2021 Humbird pessimism with a brighter outlook supported by the empirical analysis by Pasitka in 2023. The second part takes those price estimates and feeds them to a demand-side economic model aiming to calibrate the potential market share of cultivated and plant-based products. The model has many parameters, some of which are pinned down by empirical observations, and others are left for the user to tweak and explore different scenarios. Overall, the key conclusion is that cultivated meat is not a silver bullet, but should be able to capture a large enough market share to progressively erode the legitimacy of factory farming as “a necessary evil”. As such, I believe it could be combined with effective advocacy to progressively eliminate factory farming. However, this will be extraordinarily hard on many levels, from the science and technology to the social aspects, and will unfortunately take a very long time.

This post was originally intended as a reply to "Cultivating doubt: why I no longer believe cultivated meat is the answer".

AI model usage: Claude Opus 4.8 was used to vibecode the economic model presented on the webpage https://pabloamc.github.io/Cultivated_meat/interactive.html, including its technical descriptions on the webpage. The blog post was written by me, with only minor polishing from AI (less than 5%), especially for technical terms and descriptions in the economics literature.

Introduction

I have recently been thinking hard about how we could end factory farming, as have other effective altruists. Historically, there have been two broad patterns in how humanity solved large moral issues. The first and most commonly cited among the animal advocates are societal moral changes, such as the end of slavery or the civil liberties movement. Meanwhile, less famous changes happened due to technological advancements; for example, the end of whale hunting with the invention of kerosene.

The intuitions of the effective altruism community seem largely split on the feasibility of either path for ending factory farming. A technologically-driven approach to animal liberation seems easier for some, as it does not require convincing large swaths of the population to go vegan, a strategy that has otherwise produced few results for decades. On the other hand, the few concrete successes achieved over the last two decades have validated corporate campaigns as a particularly promising way of improving some animal welfare conditions (for example, egg-laying hens). Still, it is unclear how widely or far they will generalise; there is only so much you can ask of companies without a strong public buy-in. Meanwhile, initial plant-based hopes have yielded rather disappointing results, at least compared to the expectations in 2021 (McKinsey predicted $25 billion by 2030). Further, the high-quality techno-economic analysis produced by David Humbird in 2021 highlighted acute cost challenges to make cultivated meat cheaper than its conventional counterpart, under reasonable assumptions [5]. These intuitions were later reflected in mostly pessimistic forecasts for cultivated meat.

This post aims to revisit the prospects of cultivated meat. The first part will review technoeconomic analysis (TEA) for cultivated meat. While David Humbird’s TEA was high quality and pessimistic, a later empirical TEA led by Pasitka found reasons for optimistic updates [8]. As we shall cover, the main cost factors described in those analyses continue to be reduced, which suggests we might be living in the cheaper worlds allowed by the original David Humbird analysis. We also provide an appendix with detailed analysis explaining the key cost variables in such TEAs.

The second part of the blog post will present an economic model predicting the demand for cultivated and plant-based products for reasonable price ranges motivated by the TEAs, as well as other key parameters derived from the literature. This model has wide uncertainty, but it is designed to capture key intuitions of people’s behaviour. It provides estimates for market shares, diffusion dynamics, ceilings, geographies, animal species and beachhead products. We also discuss the importance of innovation economics to design the best strategies to substitute conventional meat products.

The key conclusions I take from this exercise are that (i) cultivated meat should be able to capture a significant share of luxurious products – with foie gras being perhaps the most promising case – (ii) beef and seafood could capture double-digit market shares, and (iii) pork and especially chicken are unfortunately particularly impermeable to cultivated meat. This suggests a challenging but ultimately feasible path: a concerted effort to both lower the price and raise the quality of cultivated meat, while simultaneously pushing for animal welfare and political reforms that render factory farming socially unacceptable. There are a few reasons for my optimism here. First, I think cultivated meat may be identified by consumers as their familiar meat, something that plant-based products have not achieved and may be important socially and psychologically [16]. Second, I think it is much easier to end factory farming if we provide consumers with a more expensive but otherwise identical product: the load-bearing assumption here is The Good Food Institute's (GFI) belief that consumers eat meat despite how it is currently produced, not because of it. In other words, I am hypothesising here that for many people, there is an important psychological separation between meat being “real meat” and coming from an animal. Third, as values often follow behaviour, providing consumers with an option to avoid killing animals without changing anything else in their lives should make animal killing much less socially acceptable over time, eroding the legitimacy of factory farming and ultimately creating a more humane society.

Technoeconomic analysis

A techno-economic analysis (TEA) is an in-depth study that aims to predict the cost of cultivating meat under different assumptions. The most famous one is perhaps David Humbird’s 2021, commissioned by Open Philanthropy [5]. That TEA was broadly pessimistic relative to prior expectations and arguably played a key role in sparking the funding winter in the cultivated meat sector. However, while the difficulty of achieving cost parity was the most common takeaway, the key contribution of David Humbird was to analyse the viability from first principles, including an in-depth analysis of the thermodynamic limits of cultivating meat.

The clearest example of this limit is perhaps one of Bruce Friedrich’s most emphatic quotes: chicken is the most efficient animal, yet it ingests 9 calories of feed for each calorie output in the form of meat or eggs. In terms of feed, however, a modern frankenchicken consumes only 2 times as much feed (cereals) than meat it produces (Lewis Bollard on the Dwarkesh podcast). Cultivating meat also requires media with amino acids, and such amino acids will – in one of the best-case scenarios – be extracted from soy or other plants. Since feed represents a large portion of the cost of growing chicken, there is a reasonably hard limit on the cost of cultivating meat. In this example, the cost floor of cultivated chicken is likely no less than half the cost of frankenchicken meat.

On the other hand, one criticism of David Humbird's TEA is the difficulty for the reader to separate what are strong scientific constraints from reasonable assumptions based on current and foreseeable technology. For instance, in his model, the cost of cultivated media is largely dominated first by the cost of the cultivation media and the protein source in particular; and by the choice and limits of bioreactors in the second place. Other common talk points like growth factors or the need for pharma-like hygienic conditions represent only marginal costs to the overall expectations, $3/kg to $4/kg to in each case, out of approximately $23/kg in the best case scenario. This conclusion contradicts previous TEAs that suggested growth factors as the dominant cost of cultivating meat. This separation of hard scientific constraints vs reasonable assumptions based on foreseeable technology led people to state their opinion more strongly than they should probably have: the cost of his model was dominated by the cost of the amino acid source in the media, which later work has found to be far too conservative.

Perhaps one of the highest quality subsequent TEAs (in my non-expert opinion) was the peer-reviewed TEA by Pasitka et al in 2024, carried out by the collaboration of a university group and a cultivated meat company [8]. While Humbird assumed the best-case cost of the cultivated media to be $1/L, Pasitka et al. empirically showed they could produce a hydrolysate at just $0.63/L. In 2024, Supermeat (another cultivated meat company) claimed a cost of under $0.5/L [13], and the Good Food Institute 2026 report indicates that a few companies claim a media as cheap as $0.2/L [15]. While it is not possible to verify those last claims independently, it is not unreasonable to consider a cost floor around this point or lower. Thus, if anything, it seems we live in one of David Humbird’s most optimistic scenarios, out of the general pessimism of his report. Despite this, it will be key to figure out how much cheaper the media can become, as it is still a dominant driver of the cost. Let me emphasise this: it is great news that the dominant cost factor in Humbird’s report was both (i) not a strong constraint, and (ii) probably the point where his estimates are most unfoundedly pessimistic. This should thus positively update the reader about the feasibility of cultivated meat working.

Beyond the media, another key cost factor was the bioreactor. David Humbird studied two types of bioreactors, stirred and ATF perfusion, see the appendix for more details. The bioreactor cost efficiency was limited mainly by the cell density achievable. He concluded that batch-operated stirred bioreactors represented the likely most cost-efficient option. However, when the cost of the bioreactor starts to dominate over the cost of the cultivated media, continuously operating bioreactors could become more cost-efficient [9, Fig. 6]. Part of the reason is that these new reactors can keep higher cell densities than assumed possible by David Humbird: close to 90 million cells per mL, instead of 60 million cells per mL, and to source them continuously instead of by batches. It seems this has led at least some companies to shift to perfusion bioreactors, and in particular to TTF instead of ATF. Pasitka et al. report that the reusable filters of TTF biorreactors could make TTF biorreactors up to 33% cheaper than ATF biorreactors. Despite the empirical implementation of Pasitka and following David Humbird's rigorous methodology, there is still work to be done to verify the cost when scaling bioreactors to sizes close to 20 m3, which David Humbird agreed was a reasonable target size.

A list of techno-economic analyses might be found in the bibliography. For comparison purposes, the reader might be particularly interested in [3]. In my opinion, Ref. [4] represents an excellent introduction to the technical aspects of cultivating meat. In the appendices, we also provide tabulated estimates for different input costs for David Humbird’s TEA, Pasitka et al. TEA and 2023 Negulescu’s. All three of them seem to follow the rigorous methodology of David Humbird, and provide reasonable estimates.

My conclusions

My read of the current literature is that cultivated meat seems possible at a not-unreasonable biomass cost of $25/kg. Thus, cultivated meat seems to be just a factor of 2-4 away from the cost of conventional meat. Cultivated medium seems to be the largest driver of the cost, and the cost in this line might be improved via either cheaper media or more metabolically efficient cells.

All in all, my understanding is that price parity with certain classes of meat seems doable, though beating the cost of factory-farmed meat by an order of magnitude is likely impossible. In the next section, I discuss what a price close to parity would entail for sales of cultivated meat. However, let me reemphasise that despite these seemingly optimistic predictions, science is and always will be extraordinarily hard, and both the research and diffusion to make this a reality will be extraordinarily lengthy, resource-intensive and will look desperately long – decades – for those of us who care about animal welfare (in the absence of the unpredictable effects of AI).

From price to consumer behaviour

While the above analysis focuses on the price of cultivated meat – or its wet biomass – we are ultimately interested in its market share. In this section, we present an economic model that aims to predict it. The model we present can be seen at https://pabloamc.github.io/Cultivated_meat/interactive.html. This model assumes certain features influence consumers' choices among conventional, cultivated, and plant-based meat. The three standard factors mentioned in the literature are price, taste and convenience. The model provides a way to tweak all those parameters, where the price of cultivated meat from Pasitka’s parameters (including, if desired, the need for a clean room as modelled by David Humbird). We also add other parameters like the additional cost for scaffolding or the additional retail cost.

However, there is strong evidence that price, taste and convenience are not sufficient to predict consumer choice [16]. For this reason, we introduce four more parameters. Two of them are the psychological or social value that consumers provide to cultivated (or plant-based) meat, being actual “meat”; and the value they provide to it being traditionally authentic, as in coming from animals. We also model a safety-related neophobia (with a tunable decay rate) that fades over time, as well as a constant neophobia. The intuition from these parameters is that for low and mid-quality food, consumers might have some psychological attachment to “real meat”, but probably don’t care so much where it comes from. Luxury meat might, in contrast, be assigned some value due to the provenance, as shown by the existence and premiums that Designations of Origin command. This separation between “real meat” and “meat that comes from living animals” tries to reflect GFI’s thesis that consumers eat meat despite how it is produced. Additionally, consumers might be initially reluctant to try out “lab-produced meat” for safety reasons, or they might have some other reason for neophobia that does not dissipate with time.

The model we put forward is a random-utility discrete-choice demand model with a diffusion overlay. Price is treated as exogenous — pinned by the TEAs of the previous section — rather than cleared by supply and demand; we do not model supply, economies of scale, or market equilibrium, except for letting the model user set the medium or reactor cost per kg of wet biomass, as well as scaffold cost and similar parameters. This is a deliberate scope choice: we ask "at a given price, what share would consumers choose?", not "what price would the market settle on?".

To estimate market share, we use a standard discrete-choice rule — a multinomial logit 

from the random-utility theory, where $V_j$ is the value assigned to each feature.  We compare the features of cultivated meat as a whole against the cost and features of different types of conventional meat, both by species (chicken, pork…) as well as some luxury products that may clear the bar earlier on (foie gras, salmon…). We also model two target types of consumer – ethically minded and mainstream – and different geographies with their respective conventional meat prices (US, China, Europe, Brazil, India and Nigeria). The model accounts for both the stable equilibrium and the diffusion speed.

To make these claims concrete, the figure below shows the model's penetration by type of meat for the United States, evaluated at the cost floor (the optimistic end of the cost range, chosen here to show the ceiling on what demand allows when price is most favourable). The pattern is intuitive: cheap, efficiently-produced meats are the hardest to displace, not the easiest. Premium foods, on the other extreme, pose resistance to cultivated meat due to a higher value placed on “authenticity”, despite a much easier price parity clearance. The ordering across species is chicken < pork < seafood ≈ beef, driven almost entirely by how close cultivated cost can get to each species' conventional price.

Our model suggests that the initial market (before diffusion) share might be tiny, but (in the absence of bans) could be able to capture a non-negligible market share of certain species. Pigs, and especially chickens, could be particularly challenging markets due to their efficiency in converting feed into meat. However, cows and seafood might be significantly more promising. This view is pessimistic for chicken, but could be really good news for shrimp. Note that shrimp in particular have an additional advantage: taking the exoskeleton and head from the shrimp is annoying, dirty and takes time; cultivated shrimp meat enjoys a mundane but practical advantage (not accounted for in the model).

We also model the role of luxury foods as a beachhead for cultivated meat. We find foie gras as a particularly promising product: it is luxurious, so it commands a high price, it is unstructured (so it does not need much scaffolding), and is already considered publicly as a particularly barbaric product, similar to fur farming, which has led to its ban in several geographies. Other seafood products, like tuna or salmon, could also be quite promising.

The model can also ask what happens if people become much richer (e.g. because AGI brings abundance): would they buy cultivated despite a price gap? Our analysis suggests it helps modestly — our price term is a Berry–Levinsohn–Pakes form in which richer consumers are less price-sensitive [17]. Rising incomes pull the laggard regions up the curve, and also help increase the market share of cultivated meat broadly, but it does not completely overhaul the conclusions even up to very large incomes, in part because increased incomes are damped by a logarithmic functional form.

This model has two limitations. First, the diffusion component is a standard Bass S-curve, a form designed for the first purchase of durable goods rather than repeat grocery purchases; we use it only to govern how fast adoption approaches its ceiling, never to set the ceiling itself, which is fixed entirely by the static choice model. The equilibrium shares are therefore robust to the diffusion form, even though the precise adoption timing is not. Second, and perhaps more importantly, we calibrate price sensitivity to today's marginal elasticities and then evaluate it at prices and at a degree of market maturity far from where those elasticities were measured. If consumer preferences harden or soften as cultivated meat shifts from novelty to staple, the true response could differ from ours in ways no reduced-form model estimated on a not-yet-existing product can rule out. These limitations are somewhat mitigated by the use of Monte Carlo simulations to predict confidence intervals.

Demand and disruption

When I was young and read about development economics, I was amazed by how economists would build intricate and complex theories that seemed to explain a lot, yet equally enticing theories could pin down the causes of economic development elsewhere. Here, I intend to do partly that: argue that the model above is inaccurate.

I must confess I am not just an avid reader of development economics; I feel equally attracted to the economics of innovation. And the standard economics of innovation does not support innovations disrupting incumbents by just improving on the main features. Quite to the contrary, that is more often than not a strategy for failure, since sustaining innovations are tilted in incumbents' favour. Instead, disruptive innovations typically address underserved – and not mainstream – markets. That is, there must exist a group of consumers who are fairly unsatisfied with the current available options. For example, lactose intolerants with cow milk.

This small market allows the disruptive technology to improve its key features until it can take on the mainstream market. It does not take all at once, however. Instead, it typically addresses the least profitable, less attractive segments of the mainstream market. Incumbents typically respond by leaving those less profitable segments and focusing instead of highly profitable segments.

This description feels somewhat confusing: do most disruptive innovations really attack from the cheap end? That does not feel right, after all, the electric car (and Tesla) attacked from the top, not from the bottom. What is going on here? Electric cars are indeed one of the most complicated stories of disruptive innovation: cars represent a huge upfront cost, they used to have little range, and the most benefit is for the climate, not for the user. This is in contrast to kerosene displacing whale oil, which provided much stronger and immediate benefits to the user.

However, the book “The Innovator’s Dilemma”, one of the basic books on the topic, still devotes a whole chapter to explaining how disruptive innovation really applies to electric cars. The answer – I believe – is in part the government being convinced that we needed to electrify transport, and in part Tesla addressing an underserved market: wealthy drivers who enjoy fun-to-drive cars and who want to be at the forefront of technology.

So, what does the literature of innovation economics suggest if we want to replace animal meat? It suggests it will be an uphill battle, unless we convince the public and governments that supporting alternative proteins is a good idea, thus generating an underserved market. There are strong arguments for this beyond animal welfare: climate, pandemics and reducing antimicrobial usage.

Another clearly underserved market is all those vegetarians and vegans who struggle to remain so because of social pressure or health concerns. Having cultivated meat will likely allow them to avoid animal suffering. Also, restaurants that want to avoid the vegan veto: entire groups skipping restaurants because one or two friends are vegan and would not be able to eat there. Other groups could be perhaps new parents who are particularly sensitive about the food of their newborns, and want to avoid all chances of bacterial contamination in typical animal products, as well as in hospitals.

But perhaps the biggest strength of cultivated meat is not the strength of the pull, but the ability to retain consumers. It is said that electric car drivers typically struggle to go back to driving petrol cars because the power, smoothness and reaction times of driving an electric car make driving more comfortable than driving combustion cars. Similarly, cultivated meat could retain a lot of consumers by combining less pathogenic risks with no animals being killed. Once someone is used to avoiding animal suffering, he or she will not change that behaviour unless a strong pull appears. Today, that pull is mainly social pressure, health concerns…, and the reason why vegetarian and vegan attrition rates are fairly high. But if cultivated meat is available, then that social pressure stops working: you no longer renounce anything (except potentially a higher price).

Meanwhile, I think the alternative protein community and companies should create the conditions to make plant-based and cultivated products not as an ethical sacrifice but as a new culinary experience, and grow, for instance, kitchen clubs and other social experiences on alternative proteins. Not to duplicate the meat experience, but to offer something new, exotic or fun. This way, we could make plant-based eating a social asset, not a social drag.

However, it is probably important that cultivated meat does not get associated directly with the identity of being vegetarian or vegan. Part of the reason why plant-based foods stalled is (among price, taste, etc) that many people consider themselves non-vegetarian and also consider plant-based products to be designed for vegetarians.

Ultimately, though, I think we will need to rely on the animal welfare movement to push conventional factory farming meat outside of supermarkets and restaurants. Once a strong non-industrial alternative exists, the animal welfare movement will need to convince the public that animal suffering is not necessary and should end.

Conclusion

The review of the TEAs provides reasonable estimates for the potential cost of cultivated meat. Such estimates are often above price parity, but depending on the product, they could still capture a significant market share. Unfortunately, some of the species that suffer the most (chicken and pork) will not be affected much. For this reason, we find it unlikely that cultivated meat left to the market forces could fully replace the conventional meat market.

However, this does not mean it is useless: quite to the contrary, by offering a close-enough alternative, it makes advocacy efforts much more likely to succeed. It could also reduce a lot of the attrition of vegetarians and vegans: currently, a significant fraction of vegans end up dropping back to omnivores because of health concerns or social pressure. Allowing cultivated meat to get hold of an important fraction of some categories of animal products, it should make it much easier for vegans to remain so for longer (possibly indefinite) amounts of time, since the only constraint is a not too exorbitant price. This could make becoming vegetarian an increasingly popular and easy option, leading to increased pressure on conventional products and perhaps even the ban of all but the highest welfare rungs. When it becomes sufficiently popular, animal advocates could even use some of its other perks (like public health or no use of antimicrobials) to further push for the ban of conventional meat.

Future work should (i) validate the model presented here as accurate, (ii) refine the economic parameters presented, and (iii) extend TEAs to precision fermentation so we can also model the demand for precision fermented eggs, milk and other animal-derived products. Since high-quality TEAs in this area do not exist, we find it hard to pin down the costs. I suspect carrying out a high-quality TEA of precision fermentation would thus be quite valuable to the EA community.

Some other key limitations of the model are that some concrete traditions will likely become highly entrenched politically and culturally, and some factory farmers might try to weaponise this to protect their inhumane practices. For example, 22 out of the 27 countries in the EU have banned the production of foie gras (and force feeding in general), yet France considers it a cultural and gastronomical tradition of high value. Typically, such traditions do not hold the sympathy of the general population, but are surprisingly hard to eliminate. For example, a large majority of Spanish people oppose bullfighting (according to a study by Ipsos); and consider it barbaric. Yet there is a committed minority with support from the most traditional political parties that have managed to keep this outside the public debate, and somehow anchored their identity (and the conception of their national identities) to those practices. Unfortunately, most countries seem to have somehow attached their national identities to meat, and probably to a large extent to the production methods too. At some point, we will need to figure out how to overcome this landlock situations. I suspect stressing compassion and humanity should play a key role.

I also think it will be very important to figure out what strong reasons and situations are, where people feel a strong need to switch to alternative proteins. So far, most work has been dedicated to making the products better and to replicating the meat experience as faithfully as possible. However, products that have achieved strong and swift market penetration did so in part because there was a situation where a consumer had an urgent need that could not otherwise be addressed previously.

Ultimately, cultivated meat won’t be a silver bullet. This entails that the path to a factory-farming-free society will be hard, long, and in many cases disappointing. However, it will not be useless either. I think that, managed well, it could create an opportunity to make people more compassionate with animals, and to stop considering their torture as a necessary evil; something mainstream society likely believes. I think that, combined with political advocacy and corporate campaigns, it could open the door to a society where factory farming is no longer acceptable, a more humane society.

Bibliography

[1] Ashizawa, R., Rubio, N., Letcher, S., Parkinson, A., Dmitruczyk, V., & Kaplan, D. L. (2022). Entomoculture: A preliminary techno-economic assessment. Foods, 11(19), 3037. https://doi.org/10.3390/foods11193037

[2] Garrison, Greg L., Jon T. Biermacher, and B. Wade Brorsen. "How much will large-scale production of cell-cultured meat cost?" Journal of Agriculture and Food Research 10 (2022): 100358. https://doi.org/10.1016/j.jafr.2022.100358

[3] Goodwin, Corbin M., William R. Aimutis, and Rohan A. Shirwaiker. "A scoping review of cultivated meat techno-economic analyses to inform future research directions for scaled-up manufacturing." Nature Food 5.11 (2024): 901-910. https://doi.org/10.1038/s43016-024-01061-3

[4] Gu, Huiwen, et al. "Scaling cultured meat: challenges and solutions for affordable mass production." Comprehensive reviews in food science and food safety 24.4 (2025): e70221. https://doi.org/10.1111/1541-4337.70221

[5] Humbird, David. "Scale‐up economics for cultured meat." Biotechnology and Bioengineering 118.8 (2021): 3239-3250. https://doi.org/10.1002/bit.27848 
[6] Lever VC, “A Second Generation of Cultivated Meat Companies Breaks Through Projected Cost Barriers”, 2025. https://www.levervc.com/wp-content/uploads/2025/04/Lever-VC-A-Second-Generation-of-Cultivated-Meat-Companies-Breaks-Through-Projected-Cost-Barriers-4.pdf

[7] Negulescu, Patrick G., et al. "Techno‐economic modelling and assessment of cultivated meat: Impact of production bioreactor scale." Biotechnology and Bioengineering 120.4 (2023): 1055-1067. https://doi.org/10.1002/bit.28324

[8] Pasitka, Laura, et al. "Empirical economic analysis shows cost-effective continuous manufacturing of cultivated chicken using animal-free medium." Nature Food 5.8 (2024): 693-702. https://doi.org/10.1038/s43016-024-01022-w

[9] Quint, Yossi, et al. (Ark Biotech), “Cultivated meat’s path to price parity”  (2023) https://uploads-ssl.webflow.com/628c84a07fc3884718667dbf/64acb656a98483e9a5921b83_Ark%20TEA%20v20230710.pdf

[10] Risner, Derrick, et al. "Preliminary techno-economic assessment of animal cell-based meat." Foods 10.1 (2020): 3.https://doi.org/10.3390/foods10010003

[11] Risner, Derrick, et al. "Techno-economic assessments of cellular agriculture." Cellular Agriculture. Academic Press, 2024. 469-486. https://doi.org/10.1016/B978-0-443-18767-4.00002-0

[12] Vesgeer, Robert, et al. (CE Delft) “TEA of cultivated meat” (2021) https://cedelft.eu/publications/tea-of-cultivated-meat/

[13] Supermeat, A Deep Dive Into Cost-Effective Cultivated Meat Production, 2024, https://supermeat.com/wp-content/uploads/2024/11/A-Deep-Dive-Into-Cost-Effective-Cultivated-Meat-Production-V.pdf

[14] Zhang, Linch and Dullaghan, Neil “Cultured meat: A comparison of techno-economic analyses”, (2021) https://forum.effectivealtruism.org/posts/y8jHKDkhPXApHp2gb/cultured-meat-a-comparison-of-techno-economic-analyses

[15] The Good Food Institute, “State of the Industry: Cultivated meat, seafood, and ingredients” (2026).

[16] Jacob R. Peacock Price-, taste-, and convenience-competitive plant-based meat would not currently replace meat, https://rethinkpriorities.org/research-area/price-taste-and-convenience-competitive-plant-based-meat-would-not-currently-replace-meat

[17] Berry, Steven, James Levinsohn, and Ariel Pakes. "Differentiated products demand systems from a combination of micro and macro data: The new car market." Journal of Political Economy 112.1 (2004): 68-105.

Appendix: The science of cultivated meat

Cultivated meat production involves five conceptually distinct stages:

1. establishing a cell line from a biopsy,
2. proliferating that cell line in bioreactors,
3. differentiating cells into muscle or fat,
4. structuring the tissue on a scaffold (for non-minced products), and
5. downstream processing.

TEAs focus on the second step only, though scaffolding might be challenging too.

Cultivating meat is a technically challenging endeavour. Animal cells evolved to live inside a body — bathed in blood, anchored to tissue, receiving precisely calibrated hormonal signals, and protected from the external environment by an entire immune system. Removing them from that context means replacing all of those components with engineering. The cells themselves are the first problem: unlike bacteria or yeast, which can be grown in simple mineral broth and double every 20 minutes, animal cells require a chemically complex medium containing 15–20 amino acids, growth-promoting proteins (also called growth factors), and sources of energy like glucose, and even then divide only every 24–48 hours. Animal cells need significantly stricter temperature, pH and other environmental concentration conditions; and they also lack a cell wall that would protect them from sheer stress from the cultivated medium. Finally, the cultivated media bathing them is also the perfect place for other microorganisms to proliferate, which means strong sanitary conditions to avoid any bacterial contamination.

Cell lines

Apart from the species they belong to, there are different types of animal cells we may care about. Depending on their differentiation, we may have embryonic stem cells and adult somatic cells. Adult cells might be reprogrammed as induced pluripotent stem cells. Another major technological distinction is how many times a cell can divide before it naturally stops growing (senescence):

• Primary Cells: Cells harvested directly from a living animal's tissue. They exhibit natural biological limits and will only undergo ~50 cellular divisions (known as the Hayflick limit).
• Immortalised Cell Lines: Cells that have bypassed natural senescence and can divide indefinitely. These can be naturally/spontaneously immortalised (like certain chicken fibroblasts) or genetically engineered to break the Hayflick limit.

In addition to immortalisation, animal cell lines may also be reprogrammed to live in suspension, as they otherwise prefer to replicate anchored to other cells or to a scaffold. Further changes include modifying “wild-type” cells to achieve improved metabolism. Such improved metabolism results in lower CO2, NH3, or lactate production, which would otherwise limit the achievable cell density.

Cultivated media

The cultivated medium is an aqueous medium in which cells grow and reproduce. The media should contain abundant O2 because cells are voracious for oxygen, while keeping CO2, NH3 and lactate at a minimum. The other key components of the media are:

1. Energy sources, primarily glucose or pyruvate.
2. Signalling proteins and growth factors: proteins that signal to the cell that it should multiply.
3. Amino acids: the building blocks of proteins that cells use to grow (desired) or as sources of energy (undesired).
4. Other micronutrients or additives are used to improve the conditions of the media, such as stabilisers and others.

The growth factors are the single most expensive component per kg, but they are used in minuscule quantities. The most commonly mentioned include:

• Albumin: Often the most abundant and costly recombinant protein used, accounting for a large portion of traditional media expenses. It is highly valued because it provides shear protection against the mechanical stress of bioreactors, acts as a lipid carrier, and mitigates oxidative stress. Because of its high cost, companies are actively replacing it with cheaper alternatives like rapeseed protein isolate, or synthetic combinations like methylcellulose (for shear protection) and hydroxypropyl β-cyclodextrin (as a lipid carrier).
• Insulin and Transferrin: These are crucial recombinant proteins –proteins generated by microbial generation– frequently used to replace animal-derived serum in modern formulations. They consistently rank among the top four most costly ingredients in scaled-up production models.
• Growth Factors (e.g., FGF-2, TGF-β, NRG1): These are hormones that stimulate glucose uptake and instruct the cells to rapidly divide. While cells only require them in microscopic quantities, their raw cost is astronomical—sometimes reaching millions of dollars per gram. Cost-reduction strategies involve engineering the proteins to be more potent and heat-stable, or producing them in-house using cheap microbial fermentation.

Amino acids are the physical building blocks of protein. While cheaper per gram than growth factors, they are consumed in massive quantities. They are divided into essential and non-essential, which the cells cannot generate themselves. To bypass the cost of synthesising highly pure individual amino acids, researchers are exploring complex, bulk plant hydrolysates (like soy or yeast extracts) to provide a cheap amino acid base.

Biorreactor

The role of the bioreactor is to facilitate the conditions that promote cellular growth and replication, keeping conditions stable, the media oxygenated and removing the CO2, NH3 and other catabolites (byproducts of cellular metabolism). There are broadly three types of bioreactors considered in the literature, depending on how they propagate nutrients: stirred, airlift and perfusion.

Types of bioreactors discussed in the technoeconomic analysis.

• Stirred-Tank Reactors (STRs): STRs are the most commonly cited bioreactors for cultivated meat production, utilised primarily for bulk cell expansion and differentiation. They utilise mechanical agitation, such as blade impellers, to mix the culture fluid.
• Airlift Reactors (ALRs): Airlift reactors are proposed as an alternative to STRs for exceptionally large-scale production (e.g., exceeding 20,000 litres). Instead of using mechanical impellers, ALRs circulate fluid using air sparging (air bubbles) to create a circular flow. ALRs may use bubbles of pure oxygen or air, which deliver the O2 and remove the CO2. Air bubbles may exert higher stress on cells than gentle stirring.
• Perfusion Bioreactors are designed to achieve exceptionally high cell densities by continuously replenishing fresh nutrients and removing toxic metabolic waste while retaining the cells inside the reactor. There are two types:
  • Alternating Tangential Flow (ATF): Uses a hollow-fibre filter and a diaphragm pump, but faces limitations in flow rate and scalability.
  • Tangential Flow Filtration (TFF): Uses a centrifugal pump that allows for much higher filtrate flux rates, which have successfully supported higher cell densities and scaled up to 5,000-litre systems.

Beyond their mechanical shape, reactors may also differ in their operation mode and biosecurity level:

• Fed-Batch vs. Continuous: Reactors like STRs are often run in a fed-batch mode, where nutrients are added incrementally without removing the culture fluid until the final harvest. This is contrasted with continuous processes (like perfusion), where media flows in and out constantly. Typical perfusion bioreactors may accumulate cells and feed them in batches, or harvest cells continuously.
• Pharma-Grade vs. Food-Grade: Reactors are also distinguished by their hygiene and material standards. Pharma-grade bioreactors are highly expensive and typically made of high-grade 316 stainless steel to prevent any contamination. Food-grade bioreactors are proposed as a cost-saving measure, potentially utilising lower-grade 304 stainless steel, alternative alloys, or even durable non-metallic polymers, assuming standard clean-in-place (CIP) and steam-in-place (SIP) sterilisations are sufficient.

The sterilisation is a key feature of these facilities, as even non-pathogenic bacteria would quickly outcompete the animal cells and deplete the media of valuable resources.

Humbird, 2022

Reference: https://doi.org/10.1002/bit.27848

Humbird employs a rigorous, theoretical techno-economic analysis (TEA) built on first-principles bioprocess engineering and standard practices from the pharmaceutical industry. Because cultivated meat does not yet exist at a commodity scale, Humbird bases his biological models on the known physical and metabolic limits of well-studied mammalian cells (like Chinese hamster ovary, or CHO, cells). His method heavily emphasises strict engineering constraints—such as oxygen mass transfer, catabolite inhibition, and carbon dioxide accumulation—to theoretically cap bioreactor sizes at around 20,000 Litres. Furthermore, he uses regression modelling to estimate how much raw materials (like amino acids) might cost if their supply chains were drastically scaled up to meet commodity meat demands.

Negulescu, 2023

Reference: https://doi.org/10.1002/bit.28324

Negulescu and colleagues also use a theoretical TEA approach, but their methodology is designed to explore the absolute limits of economies of scale by relaxing some of the strict pharmaceutical constraints assumed by Humbird. For example, their model assumes that carbon dioxide accumulation will not inhibit cell growth at large scales, allowing them to theoretically model massive bioreactors ranging from 42,000 Liters up to 262,000-litre airlift reactors. Additionally, their method involves altering facility assumptions, modelling a food-grade production environment rather than an expensive, pharmaceutical-grade clean room. They then apply Humbird's bulk-pricing media formulas to these much larger theoretical bioreactor scenarios to see how far the cost of goods can be driven down.

Pasitka, 2024

Reference: https://doi.org/10.1038/s43016-024-01022-w

Unlike Humbird and Negulescu, who rely primarily on theoretical assumptions about cell behaviour, Pasitka et al. use an empirical economic analysis. Their method began in the laboratory, where they generated real-world, empirical data by running a continuous manufacturing process using immortalised chicken fibroblasts in an animal-component-free (ACF) medium for over 20 consecutive days. By utilising a Tangential Flow Filtration (TFF) perfusion system, they successfully recorded actual cell densities (up to 130 ✕ 106 cells/mL), actual doubling times, and precise media consumption rates. They then took this empirical data and plugged it into a theoretical 50,000-Litre factory model (based on Humbird's financial framework) to project the costs of a scaled-up facility grounded in proven, replicable biological performance.