AI and Cultivated Meat. Near-Term Impacts of AI on the Commercial Viability of Cultivated Meat

Main Takeaways

We assessed how current and near-term AI tools affect the commercial viability of cultivated meat. We reviewed existing and emerging AI applications across the sector’s main bottlenecks—technical, economic, regulatory, political, and consumer-facing—and asked which bottlenecks AI can meaningfully address and which it cannot. Our goal is to help funders and strategists decide where to direct resources in light of rapid AI progress.

Key Findings:

• AI could provide a large relative uplift for cultivated meat, but actual adoption is far below the ceiling, constrained by data scarcity, company secrecy, funding shortfalls, and a small workforce.
• Open-access data is the binding constraint. Even as AI systems become more capable, they can only help cultivated meat R&D if sector-specific data exists to train and apply them.
• AI currently appears to help the least with addressing negative sentiment from policymakers or consumers. Insofar as investors have fled due to uncertainty about regulation or demand, AI advances do not clearly translate into progress for the sector.
• Broader AI trends (agentic systems, improving LLMs, and automated labs) could close the gap between AI’s theoretical and realized impact, primarily by lowering the expertise threshold for deploying specialized tools. But conventional meat producers have more data, capital, and infrastructure to use these same tools, so the net effect on displacement is ambiguous.

Recommendation:

Funders who consider cultivated meat promising should prevent further political bans, invest in open-access data and AI tools, and measure actual purchasing behavior and meat displacement as products reach market. As AI compresses technical timelines, non-AI institutional work—regulatory capacity, political strategy, consumer research—may become comparatively more important and neglected, and more in need of philanthropic support.

Executive Summary

Context

• There is considerable disagreement about when, if ever, cultivated meat will displace conventional meat at scale.
• Increasingly sophisticated large language models (LLMs) and AI agents may help address the cultivated meat industry’s most pressing bottlenecks.
• On the other hand, it may be that AI is largely irrelevant to some of these challenges. Or, perhaps, transformative AI (TAI) would meaningfully accelerate progress, but today’s AI tools are insufficient.

What We Did

• We reviewed what AI tools are currently available or in development to benefit the cultivated meat industry as of April 2026.
• We assessed the extent to which deployment of these tools could and will address the industry’s most pressing bottlenecks over the next ~5 years.
• We highlight which bottlenecks deserve greater attention because they benefit less from AI progress.

What We Found (see Table 1 for summary)

• AI could provide a large relative uplift for cultivated meat compared to many other industries, because the sector faces numerous well-defined scientific and engineering problems where AI tools already exist or are emerging.
• Yet actual AI adoption in cultivated meat appears far below this ceiling. The constraints are data scarcity, company secrecy, funding shortfalls, and the small size of the workforce.
• AI currently has no obvious application for addressing negative sentiment from policymakers or consumers. Insofar as investors have fled due to uncertainty about regulation or demand, advances in AI do not obviously translate into progress for the sector.
• Broader trends in AI, including improving LLM capabilities, agentic AI, or fully automated labs, could reduce the gap between theoretical and realized AI impacts in cultivated meat by minimizing uptake barriers for specialized AI tools.

Table 1: AI’s current and potential impact on cultivated meat bottlenecks

* Assuming no AGI/TAI in the next five years

Strategic Implications

• For funders who consider cultivated meat a promising technology, a couple of priorities persist, whether or not you think AI makes a difference to cultivated meat strategy:
  • Preventing further bans. This is a precondition for the sector reaching consumers and could also help attract more funding to the space.
  • Measuring consumer acceptance and conventional meat displacement as products become available will help assess whether the sector delivers on its stated benefits and update strategy accordingly
• If you think AI is necessary for making cultivated meat competitive with conventional meat or that the gap between AI’s potential and actual contribution to cultivated meat should be closed, you should consider prioritizing generating open-access data and AI tools for the cultivated meat sector.
• If the potential of AI tools for cultivated meat is realized, we expect R&D bottlenecks to be relatively less neglected and institutional or consumer bottlenecks to be the binding constraint.

How This Analysis Could Be Wrong

• Cultivated meat companies are keeping quiet about major productivity improvements facilitated by AI. Cultivated meat companies arguably have a strong incentive to overstate their progress in order to improve investor sentiment. That said, they can boast about breakthroughs without necessarily giving due credit to AI tools. Companies may be remaining coy about internal AI use to preserve competitive advantages.
• Short timelines for transformative AI render current limitations irrelevant. We restricted our projections to only 5 years, because longer forecasts suggest a non-trivial probability of TAI by then (e.g., see Wynroe et al., 2023). We hope to do some follow-up work on how TAI changes the picture we have depicted here, but do not yet have firm views.

About This Report

Below, we go through the bottlenecks in turn, describing progress that has been made on the bottleneck, regardless of AI progress, and then examining theoretical and demonstrated AI applications that could improve the bottleneck.

Methods

We drew on three main types of evidence to produce this report. First, we conducted a broad reading of academic literature, industry reports, and news articles on cultivated meat bottlenecks and AI applications. Second, we attended the Bezos Centre for Alternative Proteins conference and conducted semi-structured interviews with five industry experts and researchers, including those working at cultivated meat companies, academic labs, and non-profit organizations. We also had experts (including AI in cultivated meat specialists and a wider alternative protein sector expert) review a draft of this report to check for mistakes or areas where views and assumptions differ. Third, we reviewed existing techno-economic analyses and funding landscape data.

We did not conduct a comprehensive systematic review. Where we state that we “did not find” evidence of a particular application, this reflects the scope of our search rather than definitive absence. We flag throughout the report where our evidence is thin and where our assessments reflect judgment rather than empirically demonstrated claims.

Scope

We provide a non-exhaustive list of the bottlenecks cultivated meat faces before it could meaningfully displace animal agriculture.

We categorize these key bottlenecks into five broad categories:

• Technical or scientific: how to make cultivated meat
• Economic and cost: making it affordable to produce
• Regulatory and political hurdles: getting permission to sell
• Consumer perception and acceptance: getting people to buy cultivated meat
• Capital and funding: having money to develop facilities and products

It is necessary to overcome these five bottlenecks for reaching the ultimate goal of displacement, whether cultivated meat actually replaces conventional meat in consumer diets.

These categories are not mutually exclusive, and there are feedback loops between them. We treat them separately for clarity, but note that progress on one can create traction with another. For example, a challenge related to reducing the cost of a component may also be a scientific challenge, such as finding a new formulation that is cheaper than the standard one. We categorize bottlenecks based on our impression of what challenge the bottleneck most contributes to, but we concede that this can be somewhat messy in practice.

We describe each bottleneck, highlight any progress that has been made so far, then assess AI’s impact on each bottleneck category in turn, covering what AI already does, what it could plausibly do based on today’s capabilities, and where it appears to offer little help.

We distinguish between how AI could help in theory and how it actually helps, based on uptake and actual deployment of AI tools. For the latter, we generally focus on commercial applications of AI, as these provide the most relevant examples of actual AI use. While we give many examples, we do not provide an exhaustive list of all current and potential AI applications for each area.

What AI means in this report

We define “AI” broadly in this report to encompass the range of machine learning (ML) techniques and capabilities that could affect cultivated meat development. This includes classical ML methods (such as Bayesian optimization for experimental design), purpose-built scientific models (such as AlphaFold for protein structure prediction), and contemporary LLMs. It also includes AI agents—systems that can autonomously plan multi-step workflows, call specialized tools, and iterate on results with limited human oversight—and automated laboratories where AI controls physical instruments directly. We discuss these broader AI trends in a dedicated subsection because the distinction between individual AI tools, tool-integrated agents, and autonomous laboratory systems affects how quickly technical bottlenecks could plausibly be resolved.

We note that AI capabilities are developing rapidly, and our assessment reflects the state of the field and near-term projections as of April 2026.

Why we focus on cultivated meat

Cultivated meat is further behind than other alternative proteins and faces more unsolved technical problems, making it a useful case study for understanding AI’s impact on an emerging food-tech sector.

1. Technical or Scientific: How to Make It

1.1. Designing the Biological Inputs

The process of making cultivated meat depends on several biological building blocks working together, including cell lines and cell culture media.

The industry has made progress in developing good cell lines. Key challenges include finding or engineering cell lines that can:

• divide many times without losing function (immortalization)
• grow in serum-free and animal-component-free media
• be adapted to suspension culture for scalable bioreactor production
• have the ability to differentiate into many cell types

Progress on cell lines has been made in a few directions. First, spontaneous immortalization of various species cells has been demonstrated, including for chicken and bovine cells, and several publicly accessible cell lines are now available for research (e.g., Stout et al., 2023; Saad et al., 2023). In 2025, GFI acquired SCiFi Foods’ CRISPR-immortalized, suspension-adapted^[1] bovine lines for open distribution via Tufts University Open Cell Bank—the first time suspension bovine cells have been publicly accessible. However, cell lines from a wide range of species that have demonstrated production capabilities are still needed.^[2]

Cell culture media—the nutrient liquid in which cells grow—historically relied on fetal bovine serum (FBS), a product extracted from the blood of unborn calves during slaughter. FBS contains a complex mixture of growth factors, hormones, and nutrients that support cell growth. However, the industry is moving away from FBS for several reasons: FBS is very expensive (e.g., Lee et al., 2023), it has high variability batch-to-batch (making consistent bioprocesses difficult), some evidence indicates FBS can cause undesirable adaptations in cells (for example, if the cells being grown are not bovine), and it makes cultivated meat products non-vegan (e.g., Venkatesan et al., 2022).

Reformulating media without FBS requires identifying which specific components of FBS the cells actually need and then sourcing or engineering non-animal alternatives for each. Researchers and companies have pursued several strategies, and at least six companies have now achieved approval for products with serum-free media formulations.

The remaining challenges for serum-free media are primarily optimization problems rather than scientific barriers. Each cell line and production approach typically requires its own media formulation, optimized for specific growth characteristics, differentiation behavior, and cost targets. Optimization can include several dimensions—for example, the cells (engineering cells to grow more efficiently in specific media), formulation of the media (finding the right combination of nutrients, growth factors, and supplements), and sourcing of raw materials. While companies have demonstrated that specific serum-free media is achievable, the prevailing challenge is whether it can be made cheaply and reliably enough across the range of species and cell types the industry needs.

AI accelerates biological input design, but published applications in cultivated meat remain scarce

Below we give a handful of examples of AI tools that are or can be used in cultivated meat R&D. We likely could have given more examples, but we assume many readers are already on board with the idea that AI tools can and do significantly uplift science R&D. If you would like to see more examples, we recommend this recent review of AI applications in cultivated meat.

AI for media optimization applications likely exists. Companies usually need to test thousands of media formulations to find combinations that support reliable cell growth—a process that traditionally requires running each formulation manually, culturing cells, and evaluating outcomes. As researchers note, since cell culture media have many components and need to meet multiple requirements (such as price, stability, and effect on cell growth), media design can be treated as a hyperparameter optimization problem.

AI can help companies maximize the value of limited experimental budgets—if a company can only afford 100 experiments out of 1,000 possible, AI-guided experimental design can help select the most informative 100 of the bunch. Multus’ MediOP platform addresses this by using machine learning models to optimize formulation by extracting data from thousands of automated parallel experiments, with the model’s dataset growing each time new media is made. Clever Carnivore has also cited AI as contributing to their media cost reductions, though details are limited.

Another bottleneck is monitoring cell cultures as they grow. Counting cells accurately and non-invasively is essential for knowing when to harvest, when to adjust media, and how well a culture is performing. Traditional methods require staining cells with dyes (which can be costly and interfere with growth) or removing samples from the bioreactor. Hoxton Farms, which produces cultivated fat, made a model called BrightQuant that can estimate cell density from unstained images and be used across different cell lines.

There are also several tools that plausibly could help with cultivated meat R&D, but we could not find published applications by cultivated meat companies.^[3] The cultivated meat literature identifies several protein-engineering needs that these tools could plausibly address. Some growth factors degrade quickly in high temperatures, and enhancing their thermostability is a recognized priority—a 2025 review discussed machine learning as an emerging approach for improving growth factor thermostability, specifically for cultivated meat applications (Mainali et al., 2025). Scaffolding for structured cultivated meat products requires proteins with specific cell-adhesion and mechanical properties. Tools such as AlphaFold 3, which predicts structures of proteins, and AlphaProteo, which designs novel proteins that can bind to target molecules, could, in principle, help optimize growth factor thermostability and generate novel scaffold proteins optimized for food-grade use. However, we did not find any published papers or public announcements from cultivated meat companies reporting applications of AlphaFold, AlphaProteo, or similar frontier protein-engineering tools, though some groups are generating the growth factor data that could unlock such applications.^[4]

Cell line engineering—optimizing cell lines for the traits described in Box 1—is another area where AI could contribute, though we found no cultivated meat-specific examples yet (Todhunter et al., 2024, §3.4, discusses this further).

1.2. Making the Science Hold at Scale

Scaling up means growing cells in progressively larger bioreactors, from bench-scale (a few litres) to pilot-scale (hundreds of litres) to commercial-scale (thousands of litres or more). At each step, the physical environment inside the bioreactor changes in ways that can harm or kill cells.

In a small vessel, nutrients and oxygen can reach cells easily. In a larger vessel, mixing can become uneven, oxygen gradients form (some cells get too much oxygen, others too little), shear stress from stirring can damage animal cells, and temperature gradients can build up. Currently, each increase in scale still requires running real bioreactors and validating that cells behave as expected.

Companies can increase total production capacity in two main ways. Some pursue scale-up (building individual bioreactors that are as large as possible), which maximizes the output per vessel but runs directly into the biological challenges described above. Others pursue scale-out (running many smaller bioreactors in parallel). Scale-out avoids the biological problems of very large vessels but increases operational complexity, facility footprint, and capital requirements.

AI could help scale-up science through digital twins and bioprocess control, but adoption is early-stage

As described in section 1, scaling up cultivated meat production is difficult because the physics of a bioreactor change fundamentally at larger volumes—mixing, oxygen transfer, and heat dissipation all behave differently. A digital twin is a computational model of a biological system (in this case, a bioreactor and the cells growing in it) that integrates multiple types of data to simulate how the system behaves under different conditions. Companies can use digital twins to run “virtual experiments”—testing the effect of changing a parameter like temperature, stirring speed, or media composition in simulation before committing to expensive physical runs.

While digital twins are not necessarily AI-driven, it is likely that machine learning models could better learn complex, non-linear relationships from data and more readily adapt based on new data. Gourmey partnered with DeepLife to create a digital twin of avian (bird) cells, which can simulate cell behavior under a variety of conditions, allowing optimization of different attributes and inputs without running wet-lab experiments. They report that “these optimizations are actively informing commercial-scale production”.

AI tools can also be used to directly control the conditions inside bioreactors (e.g., Rajasekhar et al., 2024), which could minimize errors and the level of oversight needed. Cultivated meat company Magic Valley partnered with Pythag Tech to integrate AI in bioprocesses throughout scale-up and optimize media formulations. They report that this will lead to “reduced waste from manual experimentation, lower per-kilogram production costs, and shortened research and development timelines.”

1.3. Achieving Sensory Parity

Turning cultured cells into products that taste, look, cook, and feel like animal meat remains a major challenge—particularly for structured products like steaks and chicken breasts, which are also currently more expensive to produce than unstructured products. Structured products require a scaffold: a three-dimensional structure that cells can grow into, align within, and differentiate across, replicating the layered architecture of real muscle tissue. Identifying scaffolding materials that meet food safety requirements and can accurately replicate the structure of natural meat at scale remains an active area of scientific research.^[5]

Some pragmatic near-term paths avoid the full sensory parity challenge by targeting products where human sensory expectations are lower, such as pet food (e.g., Meatly).

AI could speed up sensory optimization, but we found no current applications

Traditional food product development involves preparing physical samples, recruiting and running sensory panels (groups of trained human tasters), collecting and analyzing their responses, and then reformulating.

Machine learning models can predict sensory attributes like taste, texture, aroma, and consumer acceptance from non-sensory data (chemical composition, physical measurements, spectroscopic profiles), potentially reducing the number of human sensory panels needed (Nunes et al., 2023). Electronic tongues—sensor arrays that mimic aspects of human gustatory perception—can generate standardized data that ML models interpret to classify and predict food qualities (Hao et al., 2025). See Weinstock (2026) for a more pessimistic view, suggesting that AI will not be technically capable of predicting human taste perception. Weinstock suggests the industry should continue relying on human panels instead.

We did not find published examples of cultivated meat companies using these tools for sensory optimization specifically, though there is early academic scoping work (Todhunter et al., 2024; Thomas et al., 2025). NECTAR is actively working to build the paired sensory preferences datasets and AI tools needed to predict how changes in formulation affect how a plant-based product tastes and feels. The underlying capability could, in principle, be applied to cultivated meat products as the necessary training data accumulates. Integrating AI tools for sensory analysis largely depends on generating the paired sensory-composition datasets that ML models need for training, which largely do not yet exist for cultivated meat.

2. Economic and Cost: Making it Affordable to Produce

To understand whether cultivated meat could meaningfully displace conventional meat, we need to understand whether and how much price affects consumer purchasing habits. Willingness-to-pay (WTP) for cultivated meat is heterogeneous, and the key uncertainty is whether there exists a sufficiently large segment of consumers with high enough WTP to make products commercially viable. Multiple studies have found that substantial consumer segments value cultivated meat below conventional meat—that is, they would require a discount to choose it (Van Loo et al., 2020; Kantor & Kantor, 2021; Asioli et al., 2022; Yu et al., 2025). The field consensus, as conveyed to us by industry experts, is that cultivated meat cannot sustain a price premium over conventional meat and must reach at least price parity to succeed.

For cultivated meat to be sold at low prices, it must be produced as cheaply as possible. The CEO of Fork & Good argued that scaling up production while keeping costs down requires both technological innovation and infrastructure development (Mridul, 2026). We address each below.

2.1. Scaling Production With Profitable Unit Economics

As well as being a scientific problem, scale-up is also an economic issue. Cell culture media have historically been, and in large part still remain, the single largest operating cost. Expert estimates compiled by GFI suggest media accounts for roughly 55%–95% of the per-unit operating cost of cultivated meat production, depending on the media formulation and production approach used (Specht, 2020, p. 6; Gu et al., 2025, §4.2.4).^[6]

Patterns in media cost have been promising. A widely cited techno-economic analysis of large-scale cultivated meat production (Humbird, 2021) estimated a production cost of $37 per kilogram at a global production volume of 100 million kilograms per year, with growth factors contributing $3 per kilogram (Table 2). Secondary sources have characterised Humbird’s most optimistic scenario as implying media costs of roughly $2.50 per litre and a production cost floor of $16–21 per kilogram (Lever VC, 2025), though Humbird has disputed some of these characterizations and cautioned against comparing media costs on a per-litre basis without accounting for how much media is needed per kilogram of product.

Several companies have claimed costs substantially below Humbird’s predictions. The CEO of Meatly recently reported at the Bezos Centre for Sustainable Protein Conference in March 2026 that they work with cell culture media costing 22 pence (GBP) per litre, which could reach 1.5p when producing at scale. A Believer Meats-funded,^[7] peer-reviewed analysis found media costs of $0.63 per litre (Pasitka et al., 2024). A report from Lever VC compiled claims from several other companies of costs 10–30x cheaper than Humbird’s prediction. The lowest public claim we found was from Clever Carnivore at $0.07 per litre. We note that most (though not all) of these figures are self-reported by companies and investors with clear incentives to present optimistic numbers, so we cannot easily verify the veracity of these claims.

We also note that the cost of media per litre alone can be misleading. Conversion ratio—the amount of media per kilogram of cultivated meat product—also matters. A cheaper medium that requires more volume per kilogram of output may not represent an improvement over a slightly more expensive medium with better conversion efficiency.

Growth factors, one component of cell culture media, used to be prohibitively expensive because the only existing manufacturing infrastructure produced them for pharmaceutical applications, where production at very small scales resulted in high prices per unit. Pharmaceutical production requires extremely high purity (“pharma-grade”), which demands high-quality materials, sterile equipment, and rigorous batch testing—specifications far beyond what food products require. Producing these proteins at food-grade purity and food-industry scale has brought costs down substantially (Swartz, 2023, p.37).

While costs will likely come down with economies of scale, other approaches to growth factor cost reduction involve engineering. The first is engineering the growth factors themselves—optimizing the microbial organisms that produce growth factors to increase yields and reduce costs, or modifying the growth factor proteins to improve their properties. Second, some researchers are investigating whether cells can be engineered to not need growth factors.^[8]

The shift from pharma-grade to food-grade production is a recurring theme across cultivated meat economics. It applies not just to growth factors but also to bioreactor design, facility construction, and quality control systems (see “Reducing Capital Cost of Infrastructure” below).

AI could accelerate specific aspects of the production cost challenge

Optimizing inputs like growth factors, minimizing wet-lab experiments, and controlling bioreactor conditions in the ways described above could reduce costs that scale with production (the more cells you grow, the more growth factor you need). There are a few other ways AI could help improve unit economics, though we did not find evidence of cultivated meat applications yet.

Bioprocess optimization to reduce batch failure rates (Todhunter et al., 2024 §6.1). AI-driven image processing and predictive models could reduce batch failure rates by identifying early warning signs of culture decline and adjusting parameters in real time. AI sensors and computer vision models could also detect contamination early (e.g., Chelvam et al., 2025).

Yields could be improved through better feeding strategies. AI models that predict optimal nutrient feeding schedules—when to add media, how much, and what composition—could increase the amount of biomass produced per litre of bioreactor capacity.

AI-driven process control could plausibly reduce energy consumption for heating, cooling, and mixing, and optimize water and nutrient recycling—though we did not find published studies quantifying these gains for cell culture specifically. We expect these gains to be modest in the near term compared to the gains from engineering redesign (such as the food-grade vs. pharma-grade transition described above), but they could have cost-saving effects as production scales up.

2.2. Reducing Capital Cost of Infrastructure

Insofar as unit economics are improving, infrastructure is the more binding economic constraint. Costs for a single commercial facility can be in the dozens of millions, and companies often don’t have existing cash flows from other products to finance these investments—they have to raise the capital on the promise that one day there will be a commercially viable product.

Bioreactors designed for pharmaceutical-grade production are expensive (due to the stringent requirements for sterility), and food-grade bioreactors are not widely available, so they lack economies of scale. A large number of new facilities are needed for cultivated meat production to meaningfully increase. One analysis estimated that producing around 0.3% of projected 2030 global meat production would require about 11 to 44 times the current bioreactor capacity of the pharmaceutical industry, at current levels of cell-culture productivity (though productivity levels continue to improve).

Some companies are making headway on reducing costs by applying the same pharma-to-food-grade logic to infrastructure. Meatly trialed a new 320-liter bioreactor costing £12,500 (typical costs can be £250,000); Vow developed a 20,000-liter bioreactor for under $1M using former SpaceX engineers—the largest operational cultivated meat bioreactor to date.

While cultivated meat companies are increasingly designing purpose-built food-grade equipment rather than repurposing pharmaceutical bioreactors, the scale of the infrastructure gap could still be compounded by the fact that cultivated meat is not the only sector seeking to build large-scale bioprocessing capacity—pharmaceutical industries and precision fermentation are too. The sector could draw on overlapping pools of skilled bioprocess engineers and stainless steel fabrication capacity. Whether this competition for shared inputs does or would persistently constrain cultivated meat infrastructure costs—particularly given that pharmaceutical companies can typically afford to pay more for the same talent and materials—is an open question we have not seen addressed in the literature.

Other approaches are being trialed too. Shared infrastructure models, where pilot-scale facilities (which are the first scale-up step from lab-scale facilities) are open-access and are already in operation in the UK, the Netherlands, and Singapore. Practitioners in the field consistently rank these as a high priority.

Nevertheless, capital expenditure can still strain companies. One example is Believer Meats, which achieved FDA and USDA approval and built a facility with 12,000,000 kg annual capacity, but filed for bankruptcy in December 2025 after allegedly failing to pay $34 million in construction bills.

We found limited evidence of AI applications for infrastructure cost reduction

We searched for examples of AI tools being applied to reduce cultivated meat infrastructure costs (such as AI-assisted facility design, construction planning, or equipment optimization) and did not find publicly accessible evidence of applications.

AI could, in principle, help with specific aspects of facility design and construction, such as modeling factory layouts and material supply chains. Our impression is that the core bottleneck in infrastructure cost appears to be engineering insight and capital access. The dramatic cost reductions achieved by some companies seem to have come from domain-specific decisions about materials and sterilization methods, not from computational optimization of existing designs.

One area where AI-adjacent technology could plausibly matter is advanced robotics and automation for bioreactor manufacturing itself. Companies like Meatly and Vow achieved cost reductions partly through custom design and fabrication. If the companies building bioreactors and processing equipment used AI-enabled manufacturing (e.g., robotic welding, automated quality control), they could potentially produce equipment more cheaply and pass savings on to cultivated meat producers. This is speculative—we did not find evidence of this happening yet, but we did not do a comprehensive search—but it could be an indirect pathway through which AI could reduce infrastructure costs without cultivated meat companies themselves adopting AI tools.

3. Regulatory and Political Hurdles: Getting Permission to Sell

Even technically and economically competitive products can be blocked from the market by political barriers.

3.1. Regulatory Capacity and Timelines

Cultivated meat companies face several distinct regulatory challenges. The first is the time to approval: food safety agencies are often poorly resourced and slow, and the evidence companies must provide to secure approval is not always obvious before starting the process.

At a panel at the Bezos Centre for Sustainable Protein Conference 2026, company leaders were asked to choose between a hypothetical fast-track regulatory approval, a 100,000-litre bioreactor, or a massive drop in media cost. Most chose regulatory approval, citing that it unlocks revenue, investor confidence, and consumer feedback. The CEO of Ivy Farm Technologies also recently named regulatory timelines as a key challenge.

Regulatory pathways for novel foods are frequently unclear. The United Nations Food and Agriculture Organization recently called for regulators to simplify processes for cultivated meat (FAO, 2026, p. 29–30), suggesting some attention is being paid to the issue. The UK Sandbox programme was created in part to try to improve these issues. It ensured dedicated Food Safety Agency capacity to evaluate submissions and provides pre-application advice to companies that may seek approval. Recently, the UK FSA suggested that it could approve cultivated meat for sale within the next five years.

The second challenge is whether products will pass the approval process. Different production methods entail varying requirements. Genetically modified components, for example, would have a higher regulatory burden. Some commentators have raised concerns about contamination risks in bioreactors—the warm, nutrient-rich environment that supports animal cell growth could also support the growth of pathogens if sterility is compromised. We are uncertain about how serious this risk is in practice (companies have strong incentives to maintain sterility, and contamination would typically be detected through standard quality control), but regulators will need to be satisfied that production processes are sufficiently safe each time they produce.

Finally, companies also face a sequencing problem. Securing regulatory approval requires detailing the production methods that will be used when the product is produced at commercial scale. Therefore, companies can struggle to secure approval without scale-up data, but they struggle to fund scale-up without the revenue that approval would enable. A further complication is that novel food approvals are typically granted for a specific product produced under specific conditions (e.g., in the EU: Monaco, 2025; Johnson & Monaco, 2025). If a company subsequently wants to change its cell line, media formulation, or production process, it may need to submit a new or amended application and go through a lengthy review process again. If companies have limited ability to iterate on their products once approved, it could slow the pace of improvement that would otherwise drive down costs and improve consumer acceptance.

AI may help with some aspects of regulation

Some AI tools are already being adopted by regulatory agencies. In December 2025, the FDA announced deployment of “agentic AI” for all agency employees to assist with complex tasks, including safety reviews, inspections, and compliance. The FDA also deployed an internal generative AI tool called “Elsa” to help staff with reading, writing, and summarizing internal documents. Whether these tools meaningfully accelerate the throughput of novel food applications—as opposed to medical device or drug reviews, which are the FDA’s primary focus—is unclear, but it nevertheless demonstrates that regulatory bodies are open to using AI to accelerate processes.

AI could also help on the applicant side. Cultivated meat regulatory filings require substantial safety data—toxicology assessments, allergenicity testing, nutritional analysis, and evidence that production processes are consistent and controlled. AI tools could help companies design more efficient safety and toxicity experiments, analyze complex datasets from those experiments more quickly, conduct literature reviews, and prepare regulatory dossiers faster. They could also help companies understand complex regulatory pathways and processes faster and run dossier completeness checks against published guidance requirements, flagging gaps before submission reaches a human reviewer. We did not find published examples of cultivated meat companies using AI in this way, but we expect that several may be doing so with widely available LLMs (here is some evidence in favor of this).

We remain uncertain about the net effect of AI on regulatory throughput. The binding constraints on regulatory timelines appear to include agency staffing levels, the inherently sequential nature of safety review, legal accountability requirements (where human decision-makers must ultimately sign off), and political dynamics. AI may compress some parts of this process, but we think it is somewhat unlikely to eliminate the need for human judgment at critical decision points, at least for now.

We speculate that AI could reduce approval timelines by months rather than years, but we have low–medium confidence in this estimate. Our reasoning is that the longest phases in a typical novel food approval are likely missing data (which AI may be able to identify will be an issue before the application is submitted), data generation by the applicant (which AI could accelerate), expert scientific review by the agency (which AI could assist but currently human reviewers are required to be in the loop), and political or bureaucratic queuing (which AI does not address).^[9] If AI halved the time companies spend preparing dossiers and halved the time agency reviewers spend evaluating scientific evidence (both likely optimistic assumptions), this would compress perhaps 6–12 months out of a 2.5-year process on our best guess.

3.2. Cultivated Meat Bans

Political opposition to cultivated meat has intensified in recent years. In the United States, eight states have banned the sale of cultivated meat (see Figure 1)—seven of which also banned production (see Appendix for more details). Approximately 21.5% (~75 million people) of the American population lives in states where production or sale is banned, with a further 5% in states where a bill is currently proposed.^[10] Governor Ron DeSantis framed Florida’s ban as resistance to the “global elite’s plan to force the world to eat meat grown in a petri dish or bugs to achieve their authoritarian goals.”^[11]

Figure 1: Map of US states showing which states have enacted permanent bans (red), temporary bans (orange), have proposed a bill (yellow), or have proposed a bill that did not progress (light blue).

Italy and Hungary (around 15% of the EU population)^[12] have both banned cultivated meat production and sale, and Romania has proposed a ban.^[13] France proposed a ban in 2023, but the bill did not progress. In 2026, the EU banned the use of multiple meat-related terms for the sale of plant-based proteins and cultivated meat (while “burger,” “mince,” “sausage,” and “nuggets” are allowed, terms like “chicken,” “beef,” “pork,” and “bacon” are not), though no cultivated meat products are approved for sale in the EU.

In the EU, novel foods like cultivated meat first have to pass the European Food Safety Authority risk assessment. If a product passes, the European Commission then drafts an act to approve the product, which the Standing Committee on Plants, Animals, Food, and Feed then votes on (GFI, 2024b; Monaco, 2025). The committee is composed of representatives from all EU member states, and qualified majority voting is used—the draft act can be blocked if four member states that represent more than 35% of the EU population vote against it (abstentions also count as votes against).

In January 2024, delegations from Italy, Austria, France, and 11 other member states issued a note to the Council of the EU expressing concerns about cultivated meat safety. The three author countries account for around 30% of the EU population. Experts we spoke to suggested that there is a real risk that the EU could ban cultivated meat.

How AI affects political bans is unclear at this stage

The bans described above likely reflect economic interests (protecting conventional meat producers) and cultural politics (framing cultivated meat as unnatural or elitist). AI could, in principle, help advocacy organizations analyze political dynamics, identify proposed bans before they gain momentum, target communications more effectively, or identify persuadable legislators. However, the conventional meat industry would equally benefit from these same tools, so how this influences the current asymmetry is unclear. The ceiling for uplift is higher for less-resourced groups like animal advocates and cultivated meat companies, but only if those groups prioritize adopting AI tools.

4. Consumer Perception and Acceptance: Getting People to Buy It

Even if manufacturing and production problems are solved, the response from consumers remains unclear.

Currently, awareness of cultivated meat is moderate—studies find around 43–60% of surveyed consumers report being aware of cultivated meat (e.g., see GFI, 2024a, p. 5 for a review of various sources).^[14] A 2024 poll found that only 27% of US adults consider themselves “very” or “somewhat” familiar with cultivated meat, even after the concept is explained to them (GFI, 2024a, p. 5). GFI Europe found familiarity rates ranging from 23% in Greece to 61% in the Netherlands, while studies in China and across 12 African countries found rates of 66% and 64%, respectively.

Normally, a reliable way to understand consumer demand for a product is to observe purchasing behavior in a functioning market. Cultivated meat is not yet widely available, which means we cannot measure demand the way we ordinarily would, via revealed preferences in competitive retail environments. Instead, we must rely on stated preferences: survey responses, hypothetical choice experiments, and willingness-to-try or willingness-to-pay measures.

However, consumers’ reported hypothetical willingness-to-try can differ substantially from actual consumer habits. One study using a discrete choice design (in which participants chose between specific meals) found that 59% of meals selected in a hypothetical choice were meat-free, compared to 36% in actual sales data (Brachem et al., 2019, note that cultivated meat was not included in this study).

Deception studies—where participants are told that conventional meat is cultivated or vice versa and their responses are measured—offer another way to study consumer reactions even before products are widely available. One deception study (Rolland et al., 2020) found that consumers ranked a burger they thought was cultivated (but was actually conventional meat) higher for taste than the conventional burger.^[15]

Because cultivated meat is not yet widely available, we cannot yet accurately validate hypothetical willingness-to-try with actual purchasing habits, limiting what current metrics can tell us about future adoption.^[16]

Still, some patterns can be observed that might be important for encouraging consumers to switch away from cultivated meat.

A common assumption in the alternative protein space is that if cultivated meat can match conventional meat on price, taste, and convenience—sometimes called “parity”—consumers will switch. A lot of alternative protein strategies are based on this premise, and many techno-economic analyses and projections implicitly adopt this framing. However, achieving price, taste, and convenience parity may not be sufficient for cultivated meat to displace conventional meat consumption at scale. A range of evidence on cultivated meat attitudes suggests that social and psychological factors, including culture, social norms, neophobia (aversion to new things), and naturalness heuristics, play important roles in food decisions (for reviews see Bry-Chevalier, 2026; Onwezen et al., 2021; Peacock, 2026; Siegrist & Hartmann, 2020).

Some estimates suggest that willingness to try cultivated meat sits below that for plant-based alternatives (e.g., Food Standards Agency, 2022). Among US consumers unlikely to try cultivated meat, GFI’s 2024 data (p. 22) found the top reasons were: preferring conventional meat (37%), not thinking cultivated meat is “natural” (36%), favoring helping farmers (26%), and expecting to not like the taste or texture (25%). Overall, no one objection seems to consistently rank as most common, suggesting that the barriers to adoption could be diffuse rather than concentrated around only price, taste, and convenience.

Second, while one might think that cultivated meat could hold a distinct appeal for consumers who reject plant-based alternatives (e.g., because plant-based options are not “real meat”), some evidence (e.g., Slade, 2018) found that consumer groups interested in plant-based and cultivated meat overlapped considerably. A separate study found that there was a portion of consumers who would not eat plant-based meats but would eat cultivated meat, but it was small (9% for “would buy” and 9.2% for “might buy” cultivated meat). The evidence is self-reported and from small samples, but may complicate the claim that cultivated meat will reach an otherwise unreachable audience.

Third, a recent meta-analysis found that food technology neophobia—resistance stemming specifically from the technological nature of cultivated meat rather than simple unfamiliarity with novel foods—was a strong predictor of rejection. There is also growing concern that cultivated meat is vulnerable to being categorized as ultra-processed products in the minds of consumers, which is frequently ranked as a top food concern (e.g., in the UK: FSA, 2026a), though we lack strong empirical evidence on the magnitude of this effect for cultivated meat specifically.

As cultivated meat becomes commercially available, researchers can begin measuring actual consumer behavior rather than relying on stated preferences—and these data will sharpen both advocacy and marketing strategies. Marketing research, message testing, and engagement with specific consumer segments are already being done to some extent by organizations. Measuring whether cultivated meat sales actually reduce purchases of conventional animal products will be essential for understanding whether the industry is achieving animal welfare, sustainability, and other goals. This research can begin even before full commercial availability: some cultivated meat and non-meat (e.g., milk) products are already on the market in some jurisdictions and could provide early evidence on substitution patterns.

It remains possible that widespread availability and familiarity will shift attitudes, but companies are currently unable to demonstrate demand, which can dampen investment.

AI could change some barriers to consumer acceptance, but this is uncertain

Whether AI could influence “naturalness” heuristics and food technology neophobia depends in part on how amenable consumers will be to change in light of new information. Evidence is mixed, but points toward slow change rather than rapid shifts. Repeated exposure could reduce neophobia, and the resemblance of cultivated meat to conventional meat could increase willingness-to-try for the first time, though several positive experiences may be needed (Rolland et al., 2020), and social factors around eating the food play a role (Szakály et al., 2021). AI could plausibly accelerate the rate at which consumers encounter cultivated meat products (by improving products and reducing prices), but whether availability influences neophobia is unclear.

Positive information about taste, health benefits, and sustainability can also increase willingness to try novel foods, including cultivated meat (Szakály et al., 2021; Rolland et al., 2020). AI-powered marketing personalization could help companies identify the most effective framings for different consumer segments. AI could also accelerate product improvement (better taste, texture, nutrition), which could indirectly improve consumer acceptance by making products more appealing. Though again, conventional meat companies could also use these AI capabilities to prevent uptake of cultivated meat.

AI could also influence consumer acceptance indirectly through price. If AI-enabled bioprocess optimization drives down cultivated meat production costs (as discussed above), cheaper products could attract some consumers. Separately, if AI-driven economic growth raises incomes substantially—as some analysts project—wealthier consumers might be more willing to try premium cultivated meat products, though they might equally increase consumption of premium conventional meat. The direction of this effect is ambiguous, and we do not yet have strong evidence either way.

5. Capital and Funding: Having Money to Develop Facilities and Products

Annual funding for cultivated meat companies in 2025 was 93% lower than in 2021. Compared to annual investment in cultivated companies in 2021 (the largest year), investment in 2024 was an order of magnitude lower (Battle et al., 2025, p. 14). Cultivated meat investment peaked at roughly $1 billion in 2021, then declined sharply to approximately $922 million in 2022, $226 million in 2023, $139 million in 2024, and $74 million in 2025 (Battle et al., 2025, p. 14). Investment in the last three years combined amounted to just less than the sector raised in 2021 alone.

The 2021 peak coincided with a record year for venture capital across all sectors, driven by low interest rates and abundant liquidity. Agrifoodtech investment as a whole was up ~85% in 2021 compared to the previous year (AgFunder, 2022). From 2022 to 2023, investment fell around 48%—declines steeper than the 35% drop in venture capital overall (Marston, 2024; Battle et al., 2026, p. 15). Cultivated meat’s funding collapse can therefore be seen as partly a correction from an anomalous peak, and likely partly a sector-specific loss of confidence.

Funding constrains progress on most other bottlenecks—the CEO of Hoxton Farms called scaling and funding the biggest challenges, describing them as “inseparable.”

Without capital, companies cannot scale production or survive long regulatory waits. But the sector cannot easily attract funding without demonstrating progress on those bottlenecks. As GFI argues (p. 16):

A category-wide shift in private capital tides likely requires a handful of cultivated meat companies to successfully de-risk their operations by increasing production, lowering costs, and demonstrating a path to profitability.

GFI goes on to state that venture capital alone will not be sufficient to get cultivated meat to market, instead needing a mix of funding sources, including governmental and philanthropic funding.

How AI impacts funding is unclear

AI’s relationship with cultivated meat funding is complex. The most direct effect is that AI tools could reduce the amount of capital cultivated meat companies need. If AI-guided experimental design reduces the number of wet-lab experiments required, or if LLMs allow small teams to handle regulatory preparation and literature review at lower cost, then the effective purchasing power of each dollar invested increases.

AI could also affect investor behavior, though the direction is ambiguous. On one hand, cultivated meat companies that credibly integrate AI into their R&D may be better positioned to attract funding from investors currently focused on AI-adjacent opportunities. On the other hand, AI is absorbing a growing share of total venture capital—roughly half of global venture funding in 2025 went to AI-related companies (Teare, 2026)—and one industry analyst described the current environment as “VC flight out of everything not AI” (Marston, 2025).

One further possibility is that if AI-driven productivity growth raises economic output broadly, this could expand the pool of capital available for sectors like cultivated meat. This is highly uncertain because whether macroeconomic growth translates into investment in a politically contested food technology depends on many factors beyond aggregate wealth.

AI Adoption in Cultivated Meat May Lag Due to Data Scarcity and Company Secrecy

There are numerous scientific and computational problems in the design and manufacture of cultivated meat that AI models could help solve. Yet we see very few applications currently.

One plausible explanation is that much of the relevant data does not exist yet and is expensive to generate. High-quality biological output data—how different media compositions affect taste, how cells behave at different bioreactor scales, multi-omics data across species—is costly to produce.

Even where data exists, it can also fail to replicate in new settings or processes, which could lower the value of open-source datasets, unless AI tools are capable enough to handle uncertain external validity.

Moreover, companies keep cell line data, media formulations, and process parameters as trade secrets. We found that cultivated meat employees were reluctant to discuss AI use, which could indicate that companies view AI applications as competitively sensitive, or it could reflect that specialized use is limited. While we cannot distinguish between these explanations with our current evidence, company secrecy likely means each new entrant repeats mistakes others may have already solved.

Some groups are working to make progress on shared data—for example, Food System Innovations is funding projects designed to produce open-access datasets that can help the industry improve various aspects of alternative protein development. The Bezos Earth Fund has also funded a cultivated meat AI hub to provide open-source data and tools.

We remain uncertain about the level of counterfactual benefit that companies would get from using shared datasets. GFI acquired cell lines from SciFi Foods when it went out of business, which GFI estimated saved the industry “millions of dollars and years of cell line development time.” How accurate these claims are is unclear, as they have not been independently estimated.

Broader AI Trends Could Change the Strategic Picture

The previous sections assessed AI’s impact on cultivated meat bottleneck by bottleneck, and found that realized applications are sparse relative to theoretical potential. This section considers whether several broader trends in AI development could close that gap. We describe four such trends, but note upfront that, similar to the above, each is largely prospective for cultivated meat: we found limited evidence of adoption in the sector for any of them.

Improving LLM capabilities

Frontier AI models are improving substantially in biology and chemistry with each generation. This means that each new model could provide more detailed and accurate advice, with various steps of cultivated meat development. For example, an LLM asked to help design a media optimization experiment will give more useful answers as its scientific reasoning improves. More capable models also make fewer errors and synthesize information from more sources, which compounds across the many stages of cultivated meat R&D where researchers currently rely on manual literature review, hypothesis generation, and data analysis.

Perhaps more importantly, LLMs are now proficient coders. These capabilities allow cultivated meat company employees to very rapidly build custom analysis tools, automate data pipelines, and create internal dashboards, without hiring additional dedicated software engineering staff.

AI-enabled troubleshooting

AI could also help with cultivated meat R&D through troubleshooting practical laboratory tasks. Bioreactor operation, cell culture monitoring, contamination detection, and harvest timing depend on knowledge built up over years of laboratory training. AI systems that can interpret many different types of data and that can hold the context for a given experiment, laboratory, or bioprocess can reduce the expertise threshold for operating cell culture systems.

AI systems can also be used for troubleshooting other AI tools or software, potentially lowering barriers to deploying the narrow biological tools described above.

We expect that AI troubleshooting will supplement rather than replace experienced operators for some time. But the directional effect—lowering the expertise threshold for competent bioreactor operation—could matter for a sector where experienced cell culture technicians are scarce.

Agentic AI

While we think the AI tools presented in previous sections could make cultivated meat R&D faster, each tool currently requires specialized technical expertise to operate. Each has its own interface, data format, and workflow. Integrating these tools with an AI agent allows researchers to access specialized capabilities without needing to master each tool independently.

AI agents can use a suite of specialized tools and subagents at machine speed, greatly speeding up R&D and potentially performing complicated computational workflows better than expert humans. Platforms that connect LLM-powered agents to hundreds of biological tools and databases have emerged in academic settings, and AI companies have released connectors that let researchers perform complex bioinformatics analyses through conversational interfaces.

Based on our preliminary search, the cultivated meat sector has not widely adopted these platforms yet. We think this trend matters most for smaller companies. A well-funded cultivated meat company with dedicated computational biologists can already deploy specialized tools effectively, while a startup with a handful of employees may be knowledge- or time-constrained. That said, we expect all companies—regardless of size—will be able to greatly speed up their computational R&D activities by adopting AI agents at scale.

Fully automated laboratories

The most ambitious extension of these trends is the fully automated laboratory, where AI agents control laboratory instruments directly and run closed-loop experiments. Fully automated laboratories are further from practical deployment than the other trends described above. Setting up autonomous laboratory systems requires substantial technical infrastructure and expertise, and is costly. Since the cultivated meat sector is currently funding-constrained, we expect fully automated labs are most likely to become relevant over a somewhat longer time horizon (though, given rapid AI improvements, some experts may still expect this to be realized in less than five years).

Hybrid systems combining AI-driven decision-making with human oversight at critical junctures represent the most practical near-term trajectory (Helleckes et al., 2026).

What Broader AI Trends Mean For Cultivated Meat Strategy

These four trends have different adoption barriers, different timescales, and affect different bottlenecks. But taken together, they point toward a few strategic conclusions.

First, agents could accelerate the bottlenecks where the time costs of trial and error and iteration are high. Media optimization, bioprocess control, and sensory prediction all involve iterating across many variables, integrating data from multiple sources, and translating results into actionable process changes. AI agents could speed up these kinds of workflows by designing an experiment, analyzing the output, updating a predictive model, and proposing the next experiment without a human researcher managing each hand-off between steps. Moreover, AI agents can carry on working while human employees are off the clock.

While there are likely some steps that involve irreducible physical time (for example, cells could take days to grow), agents are likely to speed up planning, analysis, interpretation, and coordination steps. We do not have good estimates of what fraction of total R&D time this kind of work represents, but we are fairly confident in the directional claim.

Second, LLMs and AI agents lower the barrier to using specialized AI tools. Even if narrow AI tools exist and would be useful, a startup with limited resources and few employees could face challenges in adopting AI tools, whether due to knowledge, skills, or time constraints. For example, a protein-design model, a digital twin, and a Bayesian optimization framework could all require their own technical expertise to set up and operate. Agents that can select and call tools could reduce the adoption barrier. While companies may still need some technical capacity to configure agents, provide them access to data, and validate outputs, the costs of integrating AI into workflows could be substantially lower for agentic AI. Today’s LLM models with strong coding capabilities can already achieve this in part by allowing users to create custom tools.

Third, the return on open-access data investment increases as AI capabilities improve. More capable models—including those integrated with specialized tools—extract more value from each dataset than simpler tools can. Analyzing a media formulation dataset with basic Bayesian optimization today could be enhanced with a much richer multi-objective optimization or an end-to-end agentic workflow as capabilities advance. Investments in open-access data made now will compound in value as the models trained on that data improve. Data also remains the binding constraint as AI model capabilities grow. As LLMs and other AI tools improve, they can be adopted only for cultivated meat R&D if specific data exists to make them relevant for the problems in this sector. Data generation infrastructure, therefore, becomes increasingly important.

Finally, agentic AI could mean fewer employees are needed to build a successful cultivated meat company, which could reduce the sector’s funding needs and allow more to be achieved with fewer resources.

One caveat is that symmetric access to agentic AI for conventional and cultivated meat complicates the picture. Conventional meat and animal agriculture companies have vastly more data, more capital, and more established infrastructure than cultivated meat startups. They can use the same AI tools—and agentic systems—to optimize their own production, reduce costs, and improve products.^[17]

If AI makes both sides more efficient, the net effect on displacement is ambiguous. Whether AI disproportionately helps cultivated meat depends on whether it solves problems that are specific to cultivated meat (e.g., media optimization, novel cell line engineering) more than it helps conventional producers with their already-mature processes.

We suspect—but cannot demonstrably prove—that AI’s marginal value is higher for cultivated meat because it faces more unsolved technical problems. Cultivated meat development also faces the type of problems that AI systems are currently best-placed to solve—well-defined scientific and engineering challenges with measurable outputs, such as optimizing a media formulation or predicting cell behavior under different bioreactor conditions^[18]—while the remaining efficiency gains in conventional meat production are more incremental and more constrained by the physical realities of raising live animals. That said, we expect there are remaining unsolved problems in aquaculture and in farming new species previously deemed unviable to farm that AI tools could improve.

Conclusions

AI has the potential to provide large relative uplift for cultivated meat, but we expect actual adoption is far below this ceiling. The gap is driven by data scarcity, company secrecy, funding constraints, and a small workforce. While there are seemingly some breakthroughs in the cultivated meat space related to the price of inputs, we have little evidence that AI is driving those breakthroughs.

A couple of priorities persist, whether or not you think AI makes a difference to cultivated meat strategy:

• Prevent further bans: For funders who consider cultivated meat a promising technology, preventing further bans is a precondition for the sector reaching consumers. Preventing further bans could also help attract more funding to the space. We found limited evidence on which political interventions work in this context, so funders may want to invest in systematic analysis of tractability by jurisdiction before committing larger resources.
• Measure consumer acceptance and conventional meat displacement as products become available: Measuring whether cultivated meat displaces conventional meat matters for assessing whether the sector delivers on its stated benefits—for animal welfare, climate, and food security. Assessing real-world purchasing habits and displacement will provide oversight on whether the theory of change for cultivated meat is impactful and cost-effective, and what can be done to improve these metrics.

If you think AI is necessary for making cultivated meat competitive with conventional meat or that the gap between AI’s potential and actual contribution to cultivated meat should be closed, you should consider prioritizing generating open-access data and AI tools for the cultivated meat sector. Several experts we consulted emphasized that if open-access datasets existed, the modeling required would often be relatively straightforward. This constraint persists even as AI systems become more agentic and capable of coordinating multiple tools because an agent that can orchestrate an entire optimization workflow is still limited by whether the underlying data exists. This means that AI investments should focus primarily on data generation and standardization, with model development as a secondary priority.

AI tools for media optimization, growth factor engineering, sensory prediction, and bioprocess control all depend on high-quality training data, but generating that data is expensive, and individual companies keep their datasets as trade secrets. Therefore, without philanthropic intervention, this data likely will not exist or will remain proprietary. Data generated by academic work can help, but needs to be generated in commercially relevant settings (i.e., at scale-up quantities and bioreactors). The return on open-access data investment increases as AI capabilities improve, because more capable systems—including those that integrate multiple tools into coordinated workflows—extract more value from each dataset than narrower tools can.

Some work already exists for generating open-access datasets, but we saw fewer examples of open-access tools. While data is the main constraint, we think open -access tools that are easy to use could still help cultivated meat companies, particularly early startups, make progress faster.

Finally, AI may change the strategic priorities in cultivated meat development. AI currently seems to offer substantial help with the bottlenecks that have historically made the most progress (biology, media, bioprocess), while offering limited help (currently) with the bottlenecks that, for the most part, have not yet improved substantially: regulatory throughput, political opposition, and consumer acceptance. The continued improvement of AI capabilities—from more powerful base models to tool-integrated agents—could widen this gap further by compressing technical timelines. We expect that technical R&D is becoming less neglected as AI can handle more of the science, agentic systems and AI-enabled troubleshooting lower the barrier to deploying specialized tools, and the companies best-placed to use AI tools are already the ones attracting private capital. Work to improve regulatory capacity, political strategy, and consumer acceptance is likely becoming comparatively more neglected, and AI currently offers fewer solutions in these domains. These patterns suggest that non-AI institutional interventions (regulatory capacity, political strategy, consumer research) may be comparatively more neglected and more in need of philanthropic support. Alongside this, investment in the open-access data infrastructure that determines whether agentic AI tools can function effectively for cultivated meat should also be prioritized.

Information gaps that would sharpen these conclusions

Cost-effectiveness of regulatory interventions. We recommend advocacy for regulatory capacity building with high qualitative confidence, but we have no cost-per-outcome estimates for this type of work.

Political tractability by jurisdiction. We can describe the political landscape but cannot currently say which interventions are tractable where, what it would take to reverse existing bans, or whether there are high-leverage moments in EU or US political processes.

Counterfactual impact of open data versus direct AI R&D grants. Our recommendation on open data infrastructure is based on the public goods logic, assuming private capital will not fund it. But we have limited evidence on how large the bottleneck actually is in practice, what it would cost to address it meaningfully, and what the downstream impact on AI tool effectiveness would be.

Economy-wide AI adoption baselines. We have limited data on how cultivated meat’s AI adoption compares to other R&D-intensive sectors of similar size and funding levels. This comparison would help calibrate whether the sector’s low adoption is unusual or expected given its constraints.

Net effect of AI on displacement. While we evaluate how AI could help cultivated meat companies here, we have not assessed whether AI disproportionately helps cultivated meat producers relative to conventional animal agriculture companies. Conventional producers have more data, capital, and established infrastructure, and can use the same AI capabilities. Whether AI narrows or widens the competitive gap between cultivated and conventional meat is an open question, with implications for the value of technical acceleration in cultivated meat.

Acknowledgements

This report is a project of Rethink Priorities—a think-and-do tank dedicated to informing decisions made by high-impact organizations and funders across various cause areas. Hannah Moulange conducted the interviews, did the analyses, and wrote the report; William McAuliffe oversaw the project. Thanks to Matt Fay, Jacob Peacock, Rikard Saqe, and Karthik Seker for feedback, Shane Coburn for copyediting, Thais Jacomassi for bibliography support, and Urszula Zarosa and Elisa Autric for assistance with publishing the report online. A special thanks to all of our interviewees for generously sharing their time and expertise.

Appendix

Proposed and Enacted Cultivated Meat Bans

This table only refers to bans or restrictions on the manufacture and sale of cultivated animal products. It does not refer to labeling laws.

* These bans take broader definitions and may apply to any animal-derived cultivated products, like dairy, not only meat (it’s possible other bans would also fall into this category, but it depends on interpretation of the law in practice).

† Applies only to vertebrate animal-derived cultivated products. May not be enforceable due to failure to notify EU.

Source for US bans without specific links provided: The National Agricultural Law Center

Bibliography

Abramson, J., Jonas Adler, Jack Dunger, Richard Evans, Tim Green, Alexander Pritzel, Olaf Ronneberger, Lindsay Willmore, Andrew J. Ballard, Joshua Bambrick, Sebastian W. Bodenstein, David A. Evans, Chia-Chun Hung, Michael O’Neill, David Reiman, Kathryn Tunyasuvunakool, Zachary Wu, & Akvilė Žemgulytė. (2024, May). Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493–500. https://doi.org/10.1038/s41586-024-07487-w

AgFunder. (2022). 2022 AgFunder AgriFoodTech Investment Report. AgFunder. https://agfunder.com/research/2022-agfunder-agrifoodtech-investment-report/ archived at https://perma.cc/DMC3-ZLM7

Asioli, D., Joaquìn Fuentes-Pila, Silverio Alarcón, Jia Han, Jingjing Liu, Jean-Francois Hocquette, & Rodolfo M. Nayga. (2022, October). Consumers’ valuation of cultured beef Burger: A Multi-Country investigation using choice experiments. Food Policy, 112. https://doi.org/10.1016/j.foodpol.2022.102376

Battle, M., Claire Bomkamp, Michael Carter, Jessica Colley Clarke, Lucas Eastham, Liz Fathman, Daniel Gertner, Jody Kirchner, Faraz Harsini, Taylor Leet-Otley, & Elliot Swartz. (2025). 2024 State of the Industry: Cultivated. The Good Food Institute. https://gfi.org/wp-content/uploads/2025/04/2024-State-of-the-Industry-Cultivated-meat-seafood-and-ingredients-GFI.pdf archived at https://perma.cc/AHT3-EZX8

Battle, M., Claire Bomkamp, Michael Carter, Eleni Chalmers, Jessica Colley Clarke, Daniel Gertner, Chelsea Hammersmith, Taylor Leet-Otley, & Elliot Swartz. (2026). 2026 State of the Industry: Cultivated. The Good Food Institute. https://gfi.org/wp-content/uploads/2026/04/GFI-2026-State-of-the-Industry-report-Cultivated-meat-seafood-and-ingredients-1.pdf archived at https://perma.cc/P5G8-JW9G

Bell, E., Carter, D., Cox, B., Hagan, L., Huddleston, D., Stinson, P. M. (2025, January). Georgia House Bill 201. Bill Track 50. https://www.billtrack50.com/billdetail/1812498 archived at https://perma.cc/F3PD-EUCP

Bezos Earth Fund. (2025, May). Cell Meat AI: Lab Grown Protein. https://www.bezosearthfund.org/grants/cell-meat-ai-lab-grown-protein-accelerator archived at https://perma.cc/LW8G-3U5D

Bloch, J. L., Marie Rouault, Cédric Langhi, Malorie Hignard, Victoria Iriantsoa, & Olivier Michelet. (2025, July). The novel food evaluation process delays access to food innovation in the European Union. Science of Food, 9. https://doi.org/10.1038/s41538-025-00492-x

Brachem, J., Henry Krüdewagen, & York Hagmayer. (2019, September). The Limits of Nudging: Can Descriptive Social Norms be Used to Reduce Meat Consumption? It’s Probably Not That Easy. OSF. https://doi.org/10.31234/osf.io/xk58q

Brennan, T., Joshua Katz, Yossi Quint, & Boyd Spencer. (2021, June). Cultivated meat: Out of the lab, into the frying pan. McKinsey & Company. https://www.mckinsey.com/~/media/mckinsey/industries/agriculture/our%20insights/cultivated%20meat%20out%20of%20the%20lab%20into%20the%20frying%20pan/cultivated-meat-out-of-the-lab-into-the-frying-pan-vf.pdf?shouldIndex=false archived at https://perma.cc/P2Z5-3H2R

Bryant, C. & Hermes Sanctorum. (2021, June). Alternative proteins, evolving attitudes: Comparing consumer attitudes to plant-based and cultured meat in Belgium in two consecutive years. Appetite, 161. https://doi.org/10.1016/j.appet.2021.105161

Bry-Chevalier, T. (2026, February). Comparing meat alternatives for a sustainable food system. Science of Food, 10. https://doi.org/10.1038/s41538-025-00694-3

CBS News. (2026, March 24). Florida’s ban on lab-grown meat upheld by federal appeals court in “another loss for Frankenmeat.” CBS News. https://www.cbsnews.com/miami/news/federal-appeals-court-upholds-floridas-ban-on-lab-grown-meat-protecting-states-real-food-law/ archived at https://perma.cc/DX5E-LD4D

Cell AgriTech. (n.d.). Cell AgriTech. Retrieved May 21, 2026, from https://cell-agritech.com/#contract archived at https://perma.cc/C8FP-2RRQ

Cellular Agriculture Europe. (2023, October). Cellular Agriculture Europe reacts to legislative move to ban cultivated meat in Romania. https://www.cellularagriculture.eu/news/cellular-agriculture-europe-reacts-to-legislative-move-to-ban-cultivated-meat-in-romania/ archived at https://perma.cc/PE2A-E43W

Centre for Process Innovation. (n.d.). Providing solutions for end-to-end product and process development in cultivated meat. Retrieved May 21, 2026, from https://www.uk-cpi.com/cultivated-meat archived at https://perma.cc/WL25-RXFB

Chelvam, S. P., Alice Jie Ying Ng, Jiayi Huang, Elizabeth Lee, Maciej Baranski, Derrick Yong, Rohan B. H. Williams, Stacy L. Springs, & Rajeev J. Ram. (2025, March). Machine learning aided UV absorbance spectroscopy for microbial contamination in cell therapy products. Scientific Reports, 15. https://doi.org/10.1038/s41598-024-83114-y

Chong, M., Angela Leung, & Tricia Marjorie Fernandez. (2024, June). On-site sensory experience boosts acceptance of cultivated chicken. Future Foods, 9. https://doi.org/10.1016/j.fufo.2024.100326

Council of the EU and the European Council. (2026, February). Qualified majority. https://www.consilium.europa.eu/en/council-eu/how-does-the-council-vote/qualified-majority/ archived at https://perma.cc/3PSM-WMRM

Council of the European Union. (2024, January). The CAP’s role on safeguarding high-quality and primary farm-based food production. https://data.consilium.europa.eu/doc/document/ST-5469-2024-REV-1/en/pdf archived at https://perma.cc/9NMN-5NW7

Cultivated X. (2025, August). Magic Valley Leverages Pythag Tech’s AI to Advance Cost-Effective Cultivated Meat Manufacturing. https://cultivated-x.com/manufacturing-technology/magic-valley-leverages-pythag-techs-ai-advance-cost-effective-cultivated-meat-manufacturing/ archived at https://perma.cc/8LAX-PDMC

Diaz, L. (2026, January). Arizona House Bill 2791. LegiScan. https://legiscan.com/AZ/bill/HB2791/2026 archived at https://perma.cc/BQ9L-YDXT

Dickerson, J. T., Russ Goodman, Drew Echols, Sam Watson, Lee Anderson, Carden Summers, Max Burns, Larry Walker, & Bo Hatchett. (2026, January). Georgia Senate Bill 415. LegiScan. https://legiscan.com/GA/bill/SB415/2025 archived at https://perma.cc/J5SZ-BCJJ

Dodds, W. (2024, February). Italian cultivated meat ban violated EU procedureOpens in new window. Food Manufacture. https://www.foodmanufacture.co.uk/Article/2024/02/07/cultivated-meat-ban-in-italy-violated-eu-rules/ archived at https://perma.cc/5J9H-3GTZ

Dodds, W. (2025, May). Cultivated meat firm successfully trials low-cost bioreactor. Food Manufacture. https://www.foodmanufacture.co.uk/Article/2025/05/27/cultivated-meat-firm-successfully-trials-low-cost-bioreactor/ archived at https://perma.cc/4HVM-JL2R

European Parliament. (2015, November). Regulation (Eu) 2015/2283 Of The European Parliament And Of The Council. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A02015R2283-20210327 archived at https://perma.cc/67FV-4VUN

eurostat. (2026, April). Total population, Candidate countries and potential candidates. eurostat. https://ec.europa.eu/eurostat/databrowser/view/tgs00027/default/table?lang=en archived at https://perma.cc/ETM2-RVNK

Executive Office of Governor Ron DeSantis. (2024, May). Governor DeSantis Signs Legislation to Keep Lab-Grown Meat Out of Florida. https://www.flgov.com/eog/news/press/2024/governor-desantis-signs-legislation-keep-lab-grown-meat-out-florida archived at https://perma.cc/7HBS-QRY9

FDA. (2025, December). FDA Expands Artificial Intelligence Capabilities with Agentic AI Deployment. https://www.fda.gov/news-events/press-announcements/fda-expands-artificial-intelligence-capabilities-agentic-ai-deployment archived at https://perma.cc/584L-BFLR

Flynn, D. (2024, May). Arizona food bill died on a tie vote. Food Safety News. https://www.foodsafetynews.com/2024/05/arizona-food-bill-died-on-a-tie-vote/ archived at https://perma.cc/5RP3-XKVE

Food Standards Agency. (2022, January). Survey of consumer perceptions of alternative, or novel, sources of protein. https://doi.org/10.46756/sci.fsa.ncn554

Food Standards Agency. (2026a, January). Consumer Insights Tracker (October 2025 – December 2025). FSA Research and Evidence. https://doi.org/10.46756/001c.155397

Food Standards Agency. (2026b, March). Thematic report on emerging food innovations in the UK. https://www.food.gov.uk/our-work/thematic-report-on-emerging-food-innovations-in-the-uk#tier-1-high-impact-and-near-term-feasibility-uk-scaling archived at https://perma.cc/JF65-CXK3

Food System Innovations. (2026, May). Food Intelligence Lab. https://www.fsi.org/food-intel-lab archived at https://perma.cc/LL6C-4WER

Good Food Institute. (2021, February). Online course: The science behind alternative proteins. https://gfi.org/resource/plant-based-and-cultivated-meat-online-course/ archived at https://perma.cc/BG3S-KQBJ

Good Food Institute. (2024a). Consumer snapshot: Cultivated meat in the U.S. https://gfi.org/wp-content/uploads/2025/01/Consumer-snapshot-cultivated-meat-in-the-US.pdf archived at https://perma.cc/PA6W-PGU6

Good Food Institute. (2024b). Explaining the EU regulatory framework for cultivated meat. https://gfieurope.org/wp-content/uploads/2024/07/Explaining-the-EU-regulatory-framework-for-cultivated-meat.docx.pdf archived at https://perma.cc/65NQ-JAPW

Good Food Institute. (2024c, October). Deep dive: Cultivated meat cell lines. https://gfi.org/science/the-science-of-cultivated-meat/deep-dive-cultivated-meat-cell-lines/ archived at https://perma.cc/E583-XNPN

Good Food Institute. (2025a, February). The science of cultivated meat. https://gfi.org/science/the-science-of-cultivated-meat/ archived at https://perma.cc/Q42E-PEJC

Good Food Institute. (2025b, April). Serum-free media for cultivated meat. https://gfi.org/wp-content/uploads/2025/04/Serum-free-media-for-cultivated-meat.pdf archived at https://perma.cc/C42E-MAYF

Good Food Institute. (2025c, October). Cultivated meat cell lines freed up for use by entire sector, enabling the field to leapfrog years of R&D. https://gfi.org/press/cultivated-meat-cell-lines-freed-up-for-use-by-entire-sector-enabling-field-to-leapfrog-years-of-research-and-development/ archived at https://perma.cc/EU4R-9N2Z

Gourmey. (2025, June). DeepLife and Gourmey Partner to Digitally Engineer the Future of Cultivated Meat. https://gourmey.com/blogs/deeplife-and-gourmey-partner-to-digitally-engineer-the-future-of-cultivated-meat archived at https://perma.cc/FHT8-RVZL

Gu, H., Yan Kong, Dejian Huang, Youfa Wang, Vijaya Raghavan, & Jin Wang. (2025, July). Scaling Cultured Meat: Challenges and Solutions for Affordable Mass Production. Comprehensive Reviews in Food Science and Food Safety, 24(4). https://doi.org/10.1111/1541-4337.70221

Hao, Z., Haisheng Li, Jianhua Guo, & Yong Xu. (2025, July). Advances in artificial intelligence for olfaction and gustation: A comprehensive review. Artificial Intelligence Review, 58. https://doi.org/10.1007/s10462-025-11309-4

Hart, C. (2026, January). West Virginia Senate Bill 225. LegiScan. https://legiscan.com/WV/bill/SB225/2026 archived at https://perma.cc/QGA9-HMCK

Hart, C., Randy Smith, Laura Wakim Chapman, Chris Rose, & Patricia Rucker. (2026, February). West Virginia Senate Bill 932. LegiScan. https://legiscan.com/WV/bill/SB932/2026?guid=vvzDcYaWxE8sawyr8MuaX archived at https://perma.cc/HX5B-64YK

Helleckes, L. M., Sebastian Putz, Kshitiz Gupta, & Matthias Franzreb. (2026, February). Perspectives for artificial intelligence in bioprocess automation. Current Opinion in Biotechnology, 97. http://doi.org/10.1016/j.copbio.2025.103392

Hoxton Farms. (2025, December). Meet Hoxton Farms’ latest tools for AI image analysis: BrightQuant and LipiQuant. https://hoxtonfarms.com/ai archived at https://perma.cc/439Y-NKD3

Humbird, D. (2021, June). Scale-up economics for cultured meat. Biotechnology and Bioengineering, 118(8). https://doi.org/10.1002/bit.27848

Imperial College. (2026, March). Bezos Centre for Sustainable Protein Annual Conference 2026. https://www.imperial.ac.uk/events/204897/bezos-centre-for-sustainable-protein-annual-conference-2026/ archived at https://perma.cc/NMM7-RKXJ

Johnson, H. & Alessandro Monaco. (2025, June). Global developments in the regulation of cultivated meat: A comparative study of the EU, Singapore, US and Australia and New Zealand. Review of European, Comparative & International Environmental Law, 34(2). https://doi.org/10.1111/reel.70007

Kantor, B. N. & Jonathan Kantor. (2021, February). Public Attitudes and Willingness to Pay for Cultured Meat: A Cross-Sectional Experimental Study. Frontiers in Sustainable Food Systems, 5. https://doi.org/10.3389/fsufs.2021.594650

Kern, S., Ryan Eaves, & David Bullard. (2026, February). Oklahoma Senate Bill 2121. LegiScan. https://legiscan.com/OK/bill/SB2121/2026 archived at https://perma.cc/M4Z3-DSWA

Lee, D. Y., Seung Hyeon Yun, Seung Yun Lee, Juhyun Lee, Ermie Jr. Mariano, Seon‑Tea Joo, Inho Choi, Jung Seok Choi, Gap‑Don Kim, Jihyun Lee, Seong-Ho Choi, & Sun Jin Hur. (2023, December). Analysis of commercial fetal bovine serum (FBS) and its substitutes in the development of cultured meat. Food Research International, 174. https://doi.org/10.1016/j.foodres.2023.113617

Lever VC. (2025, March). A Second Generation of Cultivated Meat Companies Breaks Through Projected Cost Barriers. 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 archived at https://perma.cc/FM6V-4NTP

Mainali, P., Melvin Shen-Wei Chua, Ding-Jie Tan, Bernard Liat-Wen Loo, & Dave Siak-Wei Ow. (2025, February). Enhancing recombinant growth factor and serum protein production for cultivated meat manufacturing. Microbial Cell Factories, 24. https://doi.org/10.1186/s12934-025-02670-8

Marston, J. (2024, March). New report: Agrifoodtech startup investment drops 50%, accounts for just 5.5% of global VC dollars. AgFunderNews. https://agfundernews.com/agrifoodtech-startup-investment-drops-50-now-accounts-for-just-5-5-of-global-vc-dollars archived at https://perma.cc/3DUZ-88K6

Marston, J. (2025, July). Agrifoodtech VC funding drops to $5.1bn, but ‘these are typical bottom signals, indicating real opportunities.’ AgFunderNews. https://agfundernews.com/agrifoodtech-vc-funding-drops-again-but-these-are-typical-bottom-signals-indicating-real-opportunities archived at https://perma.cc/8UB4-BS6V

McKay, H. & Sagar Shah. (2025, November). Mapping salmon welfare: A global overview. Rethink Priorities. https://rethinkpriorities.org/wp-content/uploads/2025/11/Mapping-salmon-welfare_-a-global-overview.pdf

Monaco, A. (2025, February). A perspective on the regulation of cultivated meat in the European Union. Science of Food, 9. https://doi.org/10.1038/s41538-025-00384-0

Mridul, A. (2025a, January). Dutch Govt Cements Sustainable Protein Leadership with €25M Investment in Scale-Up Facilities. Green Queen. https://www.greenqueen.com.hk/cellular-agriculture-netherlands-scale-up-lab-grown-meat-facility/ archived at https://perma.cc/ES37-4PVD

Mridul, A. (2025b, November). Hungarian Parliament Votes to Ban Cultivated Meat, Despite EU Criticism & Legal Uncertainties. Green Queen. https://www.greenqueen.com.hk/lab-grown-meat-ban-hungary-cultivated-eu/ archived at https://perma.cc/T9Z5-DJEQ

Mridul, A. (2026a, February). Cultivated Meat Outlook: 10 Leading CEOs Lay Out Their Challenges & Plans for 2026. Green Queen. https://www.greenqueen.com.hk/lab-grown-cultivated-meat-outlook-funding-companies-challenges/ archived at https://perma.cc/23VK-GLGY

Mridul, A. (2026b, March). EU Bans 31 Meat Names on Plant-Based & Cultivated Products, But ‘Veggie Burger’ Saved. Green Queen. https://www.greenqueen.com.hk/eu-veggie-burger-ban-plant-based-meat-names-vegan-labelling/ archived at https://perma.cc/P79W-ZQJ6

Multus. (2025, December). Our Technology. https://www.multus.bio/our-technology archived at https://perma.cc/A289-SNPN

Nectar. (n.d.). Accelerating the protein transition with taste. Retrieved May 21, 2026, from https://www.nectar.org/

Nielsen, M. (2026, March 23). I’ve been leading a “vibecoding” initiative at UPSIDE Foods over the last few weeks […]. LinkedIn. https://www.linkedin.com/posts/milla-nielsen_ive-been-leading-a-vibecoding-initiative-share-7441662267433332738-5inO/?utm_source=share&utm_medium=member_desktop&rcm=ACoAADKYn_IBlqM1TKebnfQmcXLVdLZQNG_KqNc archived at https://perma.cc/XG3A-4TBC

Nunes, C. A., Michele N Ribeiro, Thais CL de Carvalho, Danton D Ferreira, Luciana L de Oliveira, & Ana CM Pinheiro. (2023, February). Artificial intelligence in sensory and consumer studies of food products. Current Opinion in Food Science, 50. https://doi.org/10.1016/j.cofs.2023.101002

Onwezen, M. C., E.P. Bouwman, M.J. Reinders, & H. Dagevos. (2021, April). A systematic review on consumer acceptance of alternative proteins: Pulses, algae, insects, plant-based meat alternatives, and cultured meat. Appetite, 159. https://doi.org/10.1016/j.appet.2020.105058

Pasitka, L., M. Cohen, A. Ehrlich, B. Gildor, E. Reuveni, M. Ayyash, G. Wissotsky, A. Herscovici, R. Kaminker, A. Niv, R. Bitcover, & O. Dadia. (2022, December). Spontaneous immortalization of chicken fibroblasts generates stable, high-yield cell lines for serum-free production of cultured meat. Nature Food, 4, 35–50. https://doi.org/10.1038/s43016-022-00658-w

Pasitka, L., Guy Wissotsky, Muneef Ayyash, Nir Yarza, Gal Rosoff, Revital Kaminker, & Yaakov Nahmias. (2024, August). Empirical economic analysis shows cost-effective continuous manufacturing of cultivated chicken using animal-free medium. Nature Food, 5, 693–702. https://doi.org/10.1038/s43016-024-01022-w

Peacock, J. R. (2026, January). Price-, taste-, and convenience-competitive plant-based meat analogues would not currently replace the majority of meat consumption: A narrative review. Appetite, 216. https://doi.org/10.1016/j.appet.2025.108301

Pigott (2025, March). Mississippi House Bill 1006. LegiScan. https://legiscan.com/MS/bill/HB1006/2025 archived at https://perma.cc/PE2Y-X3UH

Pigott (2026, March). Mississippi House Bill 1153. LegiScan. https://legiscan.com/MS/bill/HB1153/2026 archived at https://perma.cc/7SYV-2ART

Protein Design and Wet Lab Teams. (2025, September). AlphaProteo generates novel proteins for biology and health research. Google DeepMind. https://deepmind.google/blog/alphaproteo-generates-novel-proteins-for-biology-and-health-research/ archived at https://perma.cc/8H4V-5PM5

Rajasekhar, N., Thota Karunakaran Radhakrishnan, & Samsudeen Naina Mohamed. (2024, June). Reinforcement learning based temperature control of a fermentation bioreactor for ethanol production. Biotechnology and Bioengineering, 121(10). https://doi.org/10.1002/bit.28784

Riquelme-Guzmán, C., Andrew J. Stout, David L. Kaplan, & Joshua E. Flack. (2024, September). Unlocking the potential of cultivated meat through cell line engineering. iScience, 27(10). https://doi.org/10.1016/j.isci.2024.110877

Rolland, N. C. M., C. Rob Markus, & Mark J. Post. (2020, April). The effect of information content on acceptance of cultured meat in a tasting context. Plos One, 15(10). https://doi.org/10.1371/journal.pone.0231176

Saad, M. K., John S. K. Yuen Jr, Connor M. Joyce, Xinxin Li, Taehwan Lim, Talia L. Wolfson, Justin Wu, Jason Laird, Sanjana Vissapragada, & Olivia P. Calkins. (2023, March). Continuous fish muscle cell line with capacity for myogenic and adipogenic-like phenotypes. Scientific Reports, 13. https://doi.org/10.1038/s41598-023-31822-2

Savage, S. (2026, March). Cargill uses AI to get more meat from the bone as beef prices soar. Finantial Times. https://www.ft.com/content/9089e369-92f4-48dc-ac09-46b6a62035a6?syn-25a6b1a6=1 archived at https://perma.cc/YW55-AXSL

Siegrist, M. & Christina Hartmann. (2020, December). Perceived naturalness, disgust, trust and food neophobia as predictors of cultured meat acceptance in ten countries. Appetite, 155. https://doi.org/10.1016/j.appet.2020.104814

Slade, P. (2018, June). If you build it, will they eat it? Consumer preferences for plant-based and cultured meat burgers. Appetite, 125. http://doi.org/10.1016/j.appet.2018.02.030

Specht, L. (2020, February). An analysis of culture medium costs and production volumes for cultivated meat. The Good Food Institute. https://gfi.org/wp-content/uploads/2021/01/clean-meat-production-volume-and-medium-cost.pdf archived at https://perma.cc/S8V8-DURD

Squires, N. (2023, November). Italy could face EU fines over ban on lab-grown meat. The Telegraph. https://www.telegraph.co.uk/world-news/2023/11/17/italy-could-eu-fine-lab-grown-meat-ban/ archived at https://perma.cc/GSB5-QED6

Stone, E., Jakob Christopherson, & Emily Whitest. (2026, April). Alternative Protein Laws State Compilation. The National Agricultural Law Center. https://nationalaglawcenter.org/state-compilations/alternative-protein-laws-state-compilation/ archived at https://perma.cc/J77V-W4S8

Stout, A. J., Miles J. Arnett, Kristin Chai, Tina Guo, Lishu Liao, Addison B. Mirliani, Miriam L. Rittenberg, Michelle Shub, & Eugene C. White. (2023, May). Immortalized Bovine Satellite Cells for Cultured Meat Applications. ACS Publications, 12. https://doi.org/10.1021/acssynbio.3c00216

Strock, T., Gary Brown, & Lynn Hutchings. (2026, February). Wyoming House Bill 61. Bill Track 50. https://www.billtrack50.com/billdetail/1950131/63183 archived at https://perma.cc/SAE4-Y2XJ

Swartz, E. (2023, January). Anticipated growth factor and recombinant protein costs and volumes necessary for cost-competitive cultivated meat. The Good Food Institute. https://gfi.org/wp-content/uploads/2023/01/GFI-report_Anticipated-growth-factor-and-recombinant-protein-costs-and-volumes-necessary-for-cost-competitive-cultivated-meat_2023-1.pdf archived at https://perma.cc/XMM4-BLAH

Szakály, Z., Bence Kovács, Mihály Soós, Marietta Kiss, & Nikolett Balsa-Budai. (2021, July). Adaptation and Validation of the Food Neophobia Scale: The Case of Hungary. Foods, 10(8). https://doi.org/10.3390/foods10081766

Teare, G. (2026, January). Global Venture Funding In 2025 Surged As Startup Deals And Valuations Set All-Time Records. Crunchbase News. https://news.crunchbase.com/venture/funding-data-third-largest-year-2025/ archived at https://perma.cc/8HSQ-USMX

The Cultured Meat Safety Initiative. (2025, January). Creating a Protein Thermostability Dataset to Accelerate Cultured Meat and Seafood Development. https://www.culturedmeatsafety.com/thermostability archived at https://perma.cc/H5JG-VB9V

Thomas, A. T., Adam Yee, Andrew Mayne, Maya B. Mathur, Dan Jurafsky, & Kristina Gligorić. (2025, February). What can large language models do for sustainable food? arXiv. https://doi.org/10.48550/arXiv.2503.04734

Todhunter, M. E., Sheikh Jubair, Ruchika Verma, Rikard Saqe, Kevin Shen, & Breanna Duffy. (2024, September). Artificial intelligence and machine learning applications for cultured meat. Frontiers in Artificial Intelligence, 7. https://doi.org/10.3389/frai.2024.1424012

Tufts University. (2025, October). Open Cell Bank. https://cellularagriculture.tufts.edu/resources/open-cell-bank archived at https://perma.cc/TJA7-LYV4

van der Heijden, M. (2026). Regulatory frameworks for cell-based food and precision fermentation-derived food products. Food and Agriculture Organization of the United Nations. https://doi.org/10.4060/cd8165en

Van Loo, E. J., Vincenzina Caputo, & Jayson L. Lusk. (2020, August). Consumer preferences for farm-raised meat, lab-grown meat, and plant-based meat alternatives: Does information or brand matter? Food Policy, 95. https://doi.org/10.1016/j.foodpol.2020.101931

Vegconomist. (2023, December). French Parliamentarians Propose Law to Ban Cultivated Meat, Calling it “Junk Food.” https://vegconomist.com/politics-law/french-parliamentarians-propose-law-ban-cultivated-meat-junk-food/ archived at https://perma.cc/7XYF-EP4A

Venkatesan, M., Cameron Semper, Stig Skrivergaard, Margrethe Therkildsen, Peter J. Stogios, & Alexei Savchenko. (2022, October). Recombinant production of growth factors for application in cell culture. iScience, 25(10). http://doi.org/10.1016/j.isci.2022.105054

Watson, E. (2025a, April). The death of cultivated meat has been greatly exaggerated, says report as Vow predicts it will soon be ‘unit margin positive.’ AgFunderNews. https://agfundernews.com/humbird-was-spectacularly-wrong-on-cultivated-meat-economics-says-report-as-vow-predicts-it-will-soon-be-unit-margin-positive archived at https://perma.cc/YGQ8-284S

Watson, E. (2025b, June). Clever Carnivore’s pragmatic path to cultivated meat profitability, starting with $0.07/liter media. AgFunderNews. https://agfundernews.com/clever-carnivores-pragmatic-path-to-cultivated-meat-profitability-starting-with-0-07-liter-media archived at https://perma.cc/NTH5-4842

Watson, E. (2025c, December). Breaking: Believer Meats ceases operations, but ‘setback’ does not mean cultivated meat sector is doomed, insists AMPS. AgFunderNews. https://agfundernews.com/breaking-believer-meats-ceases-operations-but-setback-does-not-mean-cultivated-meat-sector-is-doomed-insists-amps archived at https://perma.cc/3C7S-FDNM

Watson, E. (2025d, December). More grim news in cultivated meat sector as Meatable folds: ‘We are now entering the second phase of this industry.’ AgFunderNews. https://agfundernews.com/more-grim-news-in-cultivated-meat-sector-as-meatable-folds-we-are-now-entering-the-second-phase-of-this-industry#:~:text=AgFunder,2025 archived at https://perma.cc/6DMB-3E6L

Weinstock, I. (2026, April). AI cannot taste things. Frontier Animals. https://substack.com/home/post/p-194587426 archived at https://perma.cc/34SQ-GL3B

Williamson, S., William McAuliffe, & Hannah McKay. (2025, December). How AI is affecting farmed aquatic animals Part 1: Innovation. Rethink Priorities. https://rethinkpriorities.org/wp-content/uploads/2025/12/How-AI-is-affecting-farmed-aquatic-animals.-Part-1-Innovation.pdf

Wolf, M. (2025, November). We Talked With Nectar About Their Plans to Build an AI for Better Tasting Alt Proteins. The Spoon. https://thespoon.tech/we-talked-with-nectar-about-their-plans-to-build-an-ai-for-better-tasting-alt-proteins/ archived at https://perma.cc/2L2Z-HPEE

World Population Review. (n.d.-a). Europe. Retrieved May 5, 2026, from https://worldpopulationreview.com/continents/europe archived at https://perma.cc/R5SY-GFP2

World Population Review. (n.d.-b). United States. Retrieved May 5, 2026, from https://worldpopulationreview.com/countries/united-states archived at https://perma.cc/H9FQ-89QQ

World Population Review. (2026). US States—Ranking by Population 2026. Retrieved May 5, 2026, from https://worldpopulationreview.com/states archived at https://perma.cc/7M3V-7E2R

Wynroe, K., Atkinson, D., & Sevilla, J. (2023). Literature review of transformative artificial intelligence timelines. Epoch AI. https://epoch.ai/blog/literature-review-of-transformative-artificial-intelligence-timelines archived at https://perma.cc/K5JH-EYR8

Yu, Y., Bianca Wassmann, Madeleine Lanz, & Michael Siegrist. (2025, October). Willingness to consume cultured meat: A meta-analysis. Trends in Food Science & Technology, 164. https://doi.org/10.1016/j.tifs.2025.105226

1. Suspension adaptation refers to engineering or training cells to grow floating freely in liquid medium—rather than attached to a surface. Suspension cultures can more easily be scaled up in large stirred-tank bioreactors. ↑
2. A related, non-technical challenge remains around whether native or genetically modified cells are more acceptable to consumers and regulators (though note that UPSIDE foods has already received regulatory approval in the US with a genetically engineered cell line). ↑
3. Note that a lack of published applications is not necessarily indicative of low actual application of AI tools. Companies may keep models and AI use as proprietary information. ↑
4. Dr. Lucy Cadwell from Google DeepMind did give a presentation on AlphaProteo at the Bezos Centre for Sustainable Protein Conference 2026. ↑
5. Experts told us that there is disagreement in the field about how sensory parity can be achieved. Some argue that the flavour of cultivated meat will largely emerge from the biology of the cells themselves. Others argue that sensory quality that these attributes must be actively engineered from the outset through choices about cell differentiation, scaffolding, and formulation. ↑
6. The wide range reflects different assumptions about whether pharma-grade or food-grade inputs are used, which growth factors are required, and whether the media includes serum. We note that these figures are based on expert estimates and techno-economic analysis projections rather than published data from commercial-scale facilities, which we could not find. ↑
7. Believer Meats has since ceased operations ↑
8. Note that where genetic modification is used, a product may be subject to more stringent regulatory approval processes. ↑
9. An analysis of 292 EU novel food applications found that an application receives an average of 2.7 requests for additional data by the European Food Safety Authority (EFSA) to the applicant, and that answering these constituted 47% (± 25%) of the total evaluation time (Le Bloch et al., 2025). These figures are for EU novel food applications generally, not cultivated meat specifically, but we expect the pattern to hold. ↑
10. Based on World Population Review state and country data. ↑
11. A federal court also recently upheld Florida’s ban, ruling against a suit by Upside Foods. ↑
12. Based on World Population Review and eurostat data ↑
13. These bans may not be enforceable due to failure to notify the EU. ↑
14. Awareness also varies by terminology used (e.g., see GFI, 2024a, p.4) ↑
15. The “cultivated” burger was served in smaller pieces than the conventional one, which the authors note could have played a role in consumer perception, potentially reflecting a scarcity effect. ↑
16. One study surveyed 100 consumers in Singapore who had purchased a cultivated chicken product available there and found that the average willingness to eat it again was 4.41 out 5. However, since participants had to preregister to purchase the product, the results may not generalize to consumers not already aware of cultivated meat, or with a low opinion of it. Moreover, the study does not address whether consumers would stop eating conventional chicken in favor of cultivated chicken. ↑
17. One potential example of this comes from Cargill, which is using a computer vision system in processing lines to identify meat that is still left on the bone of the animal. ↑
18. Cultivated meat science may also be more likely to benefit from spillover of discoveries in adjacent fields (e.g., pharmaceuticals, protein engineering) than conventional agriculture is. ↑