Rethink Priorities and Metaculus have launched the Cultured Meat Tournament, which aims to provide more accurate timelines for philanthropic funders, biotech researchers, and farmed animal advocates. In this essay I present a short summary of cultured meat production, two growth projections, and then a set of new questions alongside my reasoning and probabilities.
Notes: For brevity, I use M for a million (10^6) and B for a billion (10^9). I view clean/cultured/cultivated/cell-cultured/in-vitro/lab-grown as interchangeable terms.
From the petri dish to the plate
According to Kinniment (2020), the research field of cultured meat arguably started in the 1950s with Willem van Eelen, the first person to conceive of a way of actually culturing cells for use as food. While NASA had the idea of investigating in-vitro meat in the 1970s, it was not until advances were made in tissue engineering that the plans could be executed on. In the 1990s, NASA funded research for cultured meat on long-haul space flights. The first successful result was turning goldfish cells into edible fish filets in 2002. In 2004, Jason Matheny, who founded New Harvest, convinced the Dutch government to spend €2M on cultured meat research. One of the research groups funded by the Dutch government included Mark Post, who went on to demonstrate growing muscle from cells in 2009.
Google co-founder Sergey Brin funded Post to create a cultured burger, which was presented in a 2013 live tasting, costing an estimated $2.3M/kg to produce (Colbin 2021, Kinniment 2020, CBC 2013). The first sale of cultured meat occurred in a restaurant in Singapore in December 2020 for ~$17 USD for two chicken “bites” made of ~70% cultured meat, at a loss to the producer, Good Meat (Scipioni 2020). The company again sold cultured chicken curry at a loss for ~$3 USD for a promotional event in February 2022 (Firlany & Lei 2022, Soh 2022, Vegconomist 2022). According to the Good Food Institute (GFI), at least 94 companies are working on producing cultured meat or products for cultured meat producers, and it could be as many as 160 (McGill 2022).
Good Meat may have only produced a couple of hundred kg of cultured meat for sale in Singapore in 2021 (Baker 2021, Fassler 2021) and only 700 people have eaten the chicken, according to the company (Severson 2022). Uma Valeti of Upside Foods said there are only “maybe a thousand-plus people that have tasted [this type of chicken] so far in the world” (the Economist 2021). In November 2021, GFI (2021) wrote, “the industry is currently producing cultivated meat at scales better measured in kilograms than tons.” David Humbird, who wrote a techno-economic analysis (TEA) of scaling up cultured meat, believed cultured meat production was in the 1–10 kg/year range (2020 page 9).
In comparison, US plant-based meat production in 2020 was between 90,000 and 180,000 metric tons (the former according to Shapiro 2020 and this paywalled page from Meatingplace cited in this newsletter, the latter according to data I obtained from FoodTrending.com). 13M metric tons of alternative protein (meat, seafood, milk, eggs, and dairy, excluding pulses, tofu, and tempeh) were consumed globally in 2020 (BCG 2021). ~545M metric tons of conventional meat, including seafood, is produced each year (according to OurWorldinData), mostly via the industrial farming of animals.
When will meat made mostly from cultured animal cells start being produced in the millions to tens of millions of metric tons per year?
Projections of cultured meat growth
Projection 1: Cultured meat is on a high growth trajectory
Many cultured meat companies are claiming they will be able to produce tens or thousands of metric tons of cultured meat per year within this decade. The largest production volume planned appears to be BlueNalu's plan (2019) to be able to produce 8,000 metric tons per year by 2029. (See other claims and plans from MicroMeat 2022, Aleph Farms 2022, CellMeat 2021, Good Meat 2021, BioTech Foods 2021, Future Meat 2021, MeaTech 2021, Mosa Meat 2021, Upside Foods 2021, WildType 2021, BlueNalu 2020, SuperMeat 2020, compiled in this table).
A techno-economic analysis using confidential data from cultured meat companies (Vergeer, et. al 2021) argued it is feasible to bring cultured meat to sufficiently low cost ($6.43/kg) at industrial scale (10,000 metric tons/year per facility) by 2030.
Consultancies that have published reports on cultured meat production timelines (AT Kearney 2021, McKinsey 2020, BCG 2020) estimate production could fall in the range of 1,000-90,000 metric tons by 2025, 116,000-3.5M by 2030, 6M-8.8M by 2035 (see my calculations here).
GFI produced a graphic of a potential industry life cycle for cultured meat in their state of the industry report (2020).
In a presentation about the state of the industry report with this graphic shown on screen, Blake Byrne of GFI said (timestamped video GFI 2021)
“if we're applying concepts that really come from the industrial biotechnology sector thinking about the scale-up of industrial biotechnologies, of which of course cultivated meat is one of them, we see that the industry really sits, at least the leading edge of the industry sits, at the pilot scale of production scale of production so within kind of this year next year we would expect the industry to be producing hundreds of metric tons of meat at least in these initial facilities and we would also expect to see the initial first wave of salable or commercialized products. In the coming years what we will see is companies are developing, are building, manufacturing facilities that really sit kind of in the demonstration scale and eventually industrial scale. So the demonstration scale, again which will begin within the coming years, is basically the opportunity for companies to gather key process engineering understanding for their production of cultivated meat to really understand what is the cost profile of producing cultivated meat at scale. Now, in the next five, seven, potentially 10 years we will see likely industrial scale production of cultivated meat. This is the point at which the industry is producing millions of metric tons of cultivated meat”.
(A GFI employee wrote to clarify that GFI has not made any official statements/forecasts for cultivated meat production timelines).
"Supplying 10% of the global meat market (estimated to be 40M metric tons in 2030) is ‘certainly an ambitious target’, according to Blake Byrne, former Business Innovation Specialist at GFI, but [. . .] it is ‘not out of step’ with some industry analysts, consulting firms, and investment banks, which have made projections in this industry” (Southey 2021).
The premium and specialty meat and seafood market may be ~6M tons/year (Specht et al 2021). Therefore, the 100,000 metric tons modeled in a techno-economic analysis (TEA) by Humbird (2020) “represents only 0.017% of total production, or 1.7% of the high-end market. It therefore seems entirely possible that cultivated meat could replace [100,000 metric tons per year] globally of strictly high-end products (sushi-grade tuna, steak tartare, etc.)” (Swartz 2021b).
Projection 2: Cultured meat is limited to small production volumes
Techno-economic analyses by Humbird (2020) and Risner, et al. (2020) were very skeptical that >100,000 metric tons/year will be produced in the near future, if ever, and especially not at prices close to those of conventional meat.
A GFI survey (2021), fielded in December 2020, of cultured meat companies found (of 17) 35% expected to produce 0kg to 1kg over the next 12 months (by December 2021), 41% anticipated 1kg to 10kg, 18% 10kg to 100kg, and 6% 100kg to 1,000 kg. Taking the upper bounds, this implies just 1.4 metric tons total for 2021 among these 17 companies.
As part of a project at Rethink Priorities, two Rethink Priorities researchers (Linch Zhang and Neil Dullaghan), five top Metaculus forecasters and a cultured meat scientist issued probabilities on cultured meat products (where >51% of the meat is cultured meat) reaching annual production volumes in metric tons (>100,000;>1M; >10M; >50M) by a certain year (2031, 2036, 2051). The details of the project are available in Dullaghan & Zhang (2022). The aggregated probabilities suggested annual production volumes below 100,000 metric tons were more likely than higher volumes over the next 30 years. Here is a screenshot from the report.
These probabilities appear to be shaped by some key cruxes (Dullaghan & Zhang 2022). Specifically, their beliefs/skepticism that:
- Cultured meat is a case where technological innovation can outperform the efficiency of the biological system it is trying to replicate.
- Cultured meat still faces difficult fundamental challenges.
- Cultured meat is analogous to high-growth reference classes, especially ones that were supported by government funding (such as photovoltaics and genome sequencing).
- Humbird (2020) is wrong about some fundamentals, especially media costs.
Below I will describe my areas of uncertainty and reasoning and encourage the community to add their probabilities and reasoning to a new set of questions. These are my personal views and do not represent the views of Rethink Priorities as a whole.
While I have spent a few months researching this topic, I am not a domain expert in either tissue engineering or forecasting. I tend to be under confident on average in my probabilities. Therefore, I expect and hope to update based on the engagement from the Metaculus community.
It is unclear whether companies are saying they could produce meat mostly from cultured meat as modeled in the TEAs. For example, FutureMeat is aiming to have a 45% chicken product & a 60% lamb product, and the rest 40-55% plant protein and water respectively (Watson 2021). Producing 100,000 metric tons of a plant-based meat with some cultured muscle or fat to add to the “meatiness” is quite different from producing products where the majority of the meat is from animal cells. Forecasting the hybrid scenario requires us to also forecast trajectories of plant-based meats. I’m mostly interested in cultured meat products where >51% of the “meat”, by weight, is from cultured cells for reasons discussed in Dullaghan & Zhang (2022).
Forecasting production volumes
I looked at a range of pieces of evidence to get a sense of what some reasonable upper and lower bounds for production could be.
How much weight to put in other forecasts and projections?
The Metaculus community medians for "the lowest retail price, in USD per kg, of any product containing 50% clean meat" in US or European retail supermarkets are $20 for 2026 and $11 for 2029. The estimates for how much revenue the US market for products containing at least 1% cultured meat, by weight, will generate are $3.4 billion in the fiscal year 2027 and $7.4 billion in 2030. Assuming all the US revenue came from products containing 50% cultured meat sold at the lowest price implies 170,000 metric tons for 2027 and 672,727 metric tons for 2030. Since not all this US revenue will come from 50% cultured meat products and not all 50% cultured meat products will be sold at the lowest price in the preceding year, the metric tons for such products in the US market would be lower. On the other hand, as this is just the US market and the Metaculus questions above do not include seafood, global production could be in the millions of metric tons.
The aggregated probability from the forecasting panel assembled in Dullaghan & Zhang (2022) expected annual production of >51% cultured meat to not exceed 100,000 metric tons over the next thirty years (54%). Even without my own probabilities the aggregate is still similarly low for high production volumes. However, my probabilities for the 2051 questions were 7-12 percentage points below the aggregate of the panel for those questions but similar for the other years. So perhaps I am not accounting for how much progress can be made given more time. The claimed progress from cultured meat companies on animal-free serum, non-pharmaceutical-grade amino acids, alternatives to stainless steel stirred-tank bioreactors and that the entire bioreactor system need not be Class 6 & 8 ISO clean rooms are positive signs that innovation is moving closer to biological efficiency in ways pessimists in the panel might have dismissed due to a reliance on Humbird (2020). However, I put a lot of weight in forecasters over claims from companies so this panel still anchors my probabilities quite low.
The projections from AT Kearney (2021), McKinsey (2020), and BCG (2020) lack the transparency needed to understand the reasoning behind their calculations and I do not know if their track record of projections is good in general. Their reports also make some unsubstantiated claims about cheap media and capital requirements that seem wrong to me based on what I have read elsewhere. So I do not put much weight on their estimates and treat them as maximum upper estimates given the lower estimates from the other lines of evidence. The claims about future production from cultured meat companies seem less credible given their poor track record on claims about products coming to market (Dullaghan 2021).
How much growth is needed?
The table below shows the compound annual growth rate (CAGR) needed to reach a given production volume, assuming annual cultured meat production in 2021 was 1 or 10 metric tons (which seems likely to be an overestimate and true production is in the hundreds of kilograms).
CAGR needed for cultured meat projections
An intuitive comparison is plant-based meat. It is unclear what proportion of the 90,000 to 180,000 metric tons of US plant-based meat in 2020 was from the new generation of plant-based meats launched ~2015 from Beyond Meat and Impossible Foods, but it seems likely there was a CAGR >100%. However, all plant-proteins, by weight, had a global CAGR 2013-2018 of 5.8% and a forecasted 2018-2023 CAGR of 4.3% (Kerry 2019). For meat substitutes excluding tofu, forecasts for annual global market growth vary between 6.8% and 9.4% CAGR 2019-2025 (FAIRR 2019). GFI has highlighted how a global supply squeeze of key ingredients poses a threat to plant-based meat reaching 25M metric tons by 2030 (if it had a global retail sales CAGR of 18%) (GFI 2022) and plant-based meat companies Beyond and Impossible have already run into supply issues (Shanker 2021, Capritto 2019). This particular reference class makes me think high CAGRs are possible, and maybe even likely, in the initial stage as products come to market, but also that production growth might slow dramatically before reaching 1% of meat demand.
I made a list of seven reference classes that are often touted as analogous to cultured meat (production of solar energy, electric cars, GM crops, biofuels, margarine, nuclear fission energy, lithium-ion batteries). Looking at their CAGRs, examples of the CAGR needed to hit thousands or millions of metric tons exist, but not many (just Teslas sold 2011-2021: 158% 10 year CAGR). Even solar terra-watt hours production does not appear to have had such rapid growth, and growth seems to slow after reaching 1% of energy production (Jewell and Cherp 2021). However, I recently read that Geltor, a company producing animal-free collagen via precision fermentation, has scaled up from 10,000 liters of fermentation in 2019, to 100,000 in 2020, and to "millions" in 2021 (Watson 2022), suggesting 10x growth per year. So I would not be surprised if a more systematically collected database of reference classes showed high production volume CAGRs were more common than my quick list suggested.
How much bioreactor capacity would be needed?
Vergeer, et. al (2021) suggested that in order to produce 10,000 metric tons/year it would take 2.3M liters of cell culture produced across 560 bioreactors of 2,000 to 10,000 liters of working volume each, plus another 50 or so smaller ones. I have some reasons to think the analysis in Vergeer, et. al (2021) is overly optimistic though (see Zhang & Dullaghan 2021 and Hughes 2021). Risner, et al. (2020) estimated producing 121,000 metric tons/year would require 30B to 140B liters of media spread across 360 to 5,205 20,000 liter stirred tank bioreactors in three of their four scenarios (the 4th requiring only 50 bioreactors was an intentional pie-in-the-sky model). Humbird estimated that to produce 100,000 metric tons/year with 20,000 liter bioreactors would require a global bioreactor volume of ~10M liters, but with 200,000 liter bioreactors the global installed volume would be 20M liters because larger bioreactors will produce less meat per batch due to contamination issues (2020, page 36). Good Meat claimed an unstated number of 100,000 liter bioreactors would be needed to produce ~22M metric tons/year (Baker 2021). In stark contrast to the estimates above, Mosa Meat (2021) wrote, “roughly 30 billion liters of bioreactor volume would be needed to satisfy the global yearly meat consumption of ~300 million tons.”
Is this within grasp or exceptionally large?
Mosa Meat (2021) wrote, “yearly wine consumption is ~15 billion liters, so with an average residing time of 2 years in the vessels, that compares quite well to a 30 billion liter capacity for meat.” Humbird wrote that 10M liters in 20,000 liter bioreactors “would be quite a bit larger than the entire biopharmaceuticals industry” and 100 200,000 liter “fermentors” (a bioreactor used in microbial culturing) “would be roughly enough to supply all the baker’s yeast in Europe (30% of the world)” (2020, page 36). Humbird offers a figure showing the required global volume to produce a varying amount of cell mass (2020, page 36). The diagonal lines show the bioreactor size (20 m3 = 20,000 liters, 200 m3=200,000 liters). Horizontal lines indicate the estimated installed volumes of key biotechnology sectors.
GFI’s Blake Byrne said the largest facility that has ever been produced for animal cell culture is 250,000 to 350,000 liters so a 2.3M liter facility as envisaged by Vergeer, et. al (2021) would require "10 times the volumetric capacity for cell culture than the largest facility that’s ever been produced for those purposes" (Southey 2021). McKinsey (2020) suggests current global pharmaceutical cell culture capacity is 10-20M liters. I calculated then that at 2.3M liters per 10,000 ton/year facility, replicating current global pharmaceutical cell culture capacity would be enough to produce 87,000 metric tons/year. GFI’s Elliot Swartz (2021b) estimated (based on McKinsey analysis) to produce 1.5 M metric tonnes, we need bioreactor capacity 11x to 22x the current global pharma industry. George Peppou (2021) of VowFoods noted that "For a sense of scale, 22x the global cell culture space would fit inside one of Tesla's gigafactory". Producing 40M metric tons by 2030 (10% of projected meat demand) would require 4,000 of these facilities (Southey 2021), which would be 9.2 billion liters of cell culture capacity (42x the current global pharma industry). On the one hand this seems like it requires a huge scale up in capacity, but on the other hand the global pharma industry is smaller than I expected. So 100,000 metric tons/year doesn’t seem like such a monumental feat to think a concerted effort of public funding, talented scientists and engineers, and luck could not achieve in one person’s lifetime
How long would it take to scale up production?
Bioreactor construction appears to take 12 months to go from initiation to completion (Watson 2022) (but one cultured meat company claims to have "bought a large scale line with <3 week lead time" (Peppou 2022)). Claims from companies imply 2-5 years are needed to build entire operational facilities (Aleph Farms 2022, Askew 2021, Shanker 2021, BlueNalu 2019). Another alternative protein process, precision fermentation, appears to be bottlenecked by a lack of manufacturing capacity (Warner 2022). New commercial fermentation facilities take ~36 months to build and retrofitting existing facilities can take~ 18 months, but there is a very limited supply of existing facilities (Warner 2021). Plant-based milk company, Oatly, had significant problems scaling up (TheJournal 2022). Plant-based meat companies had some issues meeting demand while they were relying on co-manufactures, which left them competing with other bigger companies for a shift in the production line (Bollard 2021). It was not until they got capital to build their own production facilities that they could clear that bottleneck. The world’s largest plant-based meat focused contract manufacturing facility has a capacity of 55,000 metric tons (Askew 2020).
It seems plausible the major bottleneck for cultured meat will be a mechanical engineering one, not a tissue engineering one (Wolf 2022). I would want to see ground broken on much larger production facilities by 2028-2029 and plans to do so before then to think we are on track for high growth trajectories. Even using the seemingly more optimistic but cheaper factory costs estimates in Vergeer, et. al (2021), 10,000 metric ton/year facilities could cost $450M so ramping up to 100,000 or 1M seems like the costs can not be swallowed by venture capitalists. Public funding is needed but seems to be only in the range of $12M-$200M by 2021 so it could take a few years just to lobby for, allocate, and distribute much larger amounts.
How far are companies now from where they need to be?
So far, none of the concrete public plans from cultured meat companies are for facilities with production capacity greater than 8,000 metric tons/year, and neither are they plans for many smaller facilities instead. Many of the claims from cultured meat companies are about production capacity, which I read as being the amount they could produce if it were economically viable to do so. This is not the same as saying once the facilities are operational they will be consistently producing that amount. The claims about future production capacity seem less credible given the poor track record on claims about products coming to market (Dullaghan 2021).
Humbird wrote “today, animal cell culture is not practiced in bioreactors larger than [25,000 liters”] (2020 page 5) and designs larger than 200,000 liters “are not practicably sterilized to a level suitable for animal cell culture”(2020 page 7). GFI’s Blake Byrne said going above 10,000 liters in the animal cell culture space is “extraordinarily rare" (Southey 2021). Actual production is occurring in bioreactor systems much smaller than this. SuperMeat is using 200-1,000 liter reactors for their tasting restaurant (SuperMeat 2022), FutureMeat claims to have installed 1,200 liter bioreactors (Leichman 2019) while Good Meat claim to use 1,200 to 5,000 liter bioreactors (Baker 2021).
However, Good Meat look to be installing 10,000 to 12,000 liter bioreactors in their upcoming Doha facility with a claimed production capacity of 4,500 metric tons per year (Quint 2022). GFI’s Elliot Swartz argued that reactors >20,000 liters are usually not the stirred tanks usually modeled in TEAs: "They are something else like bubble columns and air-lift reactors, which mix cells and liquids differently” and points to a study that “suggests a single 300,000L air-lift reactor could produce enough [cultured meat] for 75k people" (Swartz 2021 citing Li, et al 2020). However, that study assumed an annual cultured meat consumption of 10kg/person, so it amounts to only about 7,500 metric tons/year. This doesn’t seem that large a production volume for such a massive reactor.
The Metaculus community median forecast for the largest cultivated meat product production capacity, in metric tons per year, of a single production facility, has trended down. From September 2021 to January 2022 the median estimates moved from 124 metric tons down to 58 metric tons for a facility in January 2023 and 30,000 metric tons to 10,000 metric tons for a facility in January 2030. GFI’s Elliot Swartz, when talking about the facilities modeled in the Vergeer, et. al (2021) which produces 10,000 metric tonnes annually of cultured meat, said "the facility in the [study] is very large in terms of its volumetric capacity - it is highly likely that most facilities in the next 10 years will be smaller than this" (Neo 2021). So it appears the Metaculus community median for single production facility capacity in 2030 has now moved to be in line with that but it is unclear why the community assumed it could be higher before.
I would want to see at least ten facilities with at least a capacity of 10,000 metric ton/year before 2028 to believe 100,000 metric tons/year can be actually produced in 2030, but so far there does not seem to be even one such plan.
Is it just a bioreactor manufacturing problem?
It is worth noting that some of the forecasters in Dullaghan and Zhang (2022) believed that cultured meat production at scale may run up against hard biological limits, not just engineering ones, that make it a case where technology can not beat the biological system it is trying to replicate. Concerns about the viability of cheap alternatives to pharmaceutical-grade media (especially amino acids) also nudges my estimates down. A GFI survey (2020) of cultured meat manufacturers and suppliers of culture media components or formulations provided predictions on the lowest production cost of media that could be achieved by December 2026. The majority of those surveyed think the best price they can get is “<$5/kg”, frustratingly in a range that covers both higher than Humbird's (2020) lowest estimate ($3.39/kg) and the lowest estimate of what is needed (<$1). In that survey 72% of respondents indicated that cell growth media represented over 50% of their operating costs, and 38% said growth media represented 80% or more of operating costs.
One concern raised, which I share, is that the business incentives are there to “find alternatives to the largest cost-drivers so that they can begin to produce regularly at smaller scales without going bankrupt” but the required innovation and cost reduction across the entire value chain is unlikely to be driven by private profit seeking—such as moving “to a system where the main basal media components not currently a major cost driver are grown via agriculture as opposed to biotech” (Avacyn 2020).
Overall, it seemed as though the most rigorous and quantified estimates from the Humbird (2020) TEA and the forecasting panel point towards production volumes of >51% cultured meat of lower than 100,000 metric tons. The estimates suggesting millions of tons were the least rigorous and quantified, only indirect estimates in the case of the Metaculus questions, or vague enough that they could apply to hybrid products. Based on these estimates and the general sense that there are not many precedents for such rapid CAGR (but with large uncertainty), I gave low probabilities for production volumes above 100,000 metric tons/year, for ≥51% cultured meat.
The forecasting panel in Dullaghan and Zhang (2022) seemed to give half the odds to <$10/kg as “any price” cultured meat, so I halved my production volume estimates for these cheaper products after 2040 and made even bigger reductions for products in the next 15 years. ≥20% cultured meat volumes could be higher if they are just adding some meatiness to a more established plant-based meat industry. I didn’t invest time into also building a model of other alternative protein meats and invested more weight into the high growth projections from consultancies and companies than in the ≥51% forecasts. I still have some doubts that all consumers who would choose cultured meat would choose hybrids. Survey respondents interested in cultured meat are heavy meat-eaters, while those more interested in plant-based meat are those reducing their meat intake (Bryant and Barnett 2020). My intuition is that the heavy-meat eaters would still view hybrid products as plant-based meats and opt for them less often, so this nudged my estimates down.