How Plants Are Becoming Programmable Assets

Plants are starting to look less like passive agricultural assets and more like programmable biological systems. This landscape looks at the startups building the tools, platforms, and crop products that make it possible to design plants for specific economic use cases: higher yield, climate resilience, better ingredients, new materials, or even molecular production.

How did plant breeding/selection evolve?

• Before the 1960s: Plant breeding had existed for thousands of years, but it was mostly empirical. The core loop was: cross plants, observe the offspring, keep the best ones, and repeat. The process worked, but it was slow and largely based on what breeders could see in the field. The plant was improved through selection, not really designed.

• 1960s to 1970s: the Green Revolution proved that better genetics could reshape agriculture. The Green Revolution showed the power of improved crop varieties at global scale. Semi dwarf, high yielding wheat and rice varieties helped increase food production dramatically, especially in Asia and Latin America. But the model was still mostly “breed and scale”: create better varieties, distribute them widely, and support them with irrigation, fertilizer, and modern agronomic practices.

• 1980s to 1990s: biology became more molecular. The next step was the ability to intervene more directly in plant biology. Tissue culture, transformation methods, and genetic engineering made it possible to introduce specific traits rather than only recombine existing ones. Commercial genetically engineered crops arrived in the 1990s, which changed the field but also created a heavy regulatory and public-acceptance layer around plant biotechnology.

• 2000s to 2010s: breeding became data-driven. As sequencing and genotyping became cheaper, breeders could start looking beyond visible traits and use DNA markers to make selection decisions earlier. Marker assisted selection and then genomic selection turned breeding into more of a prediction problem: instead of waiting for every trait to show up in the field, breeders could use genetic data to estimate which plants were worth advancing.

• 2010s to today: crop development is becoming a design, test, and deployment process. The newest wave combines AI, genomics, gene editing, RNA data, cell-level screening, robotics, and field validation. CRISPR made genome editing easier to engineer in plants, while AI and biological data now help decide which edits, traits, or candidates are worth testing. This is where today’s startups sit: they are not inventing plant breeding from scratch, but compressing the loop between idea, biological test, and commercial crop.

What types of companies are on this landscape?

• I broke down the landscape into three broad types of companies. At a high level, the map follows the different ways plants are becoming programmable. First, the companies that make crop development more predictive. Second, the ones that build the technical layer needed to modify plant biology. The third group turns plants into production systems for valuable molecules.

• Category #1: AI trait design & predictive breeding platforms. These companies try to make breeding less dependent on trial and error. They use biological data and AI models to identify promising plant candidates earlier, before spending years testing weak options in the field.

• Category #2: Genome editing & trait-delivery platforms. These companies focus on the tools that make plant engineering possible. Their role is to help modify plant biology more precisely, whether by editing genes, controlling gene expression, or improving how new traits are introduced and stabilized.

• Category #3: Plant molecular farming & biofactories. These companies use plants less as crops and more as biological production systems. The goal is to make plants produce valuable molecules, such as food proteins, enzymes, or specialty ingredients.

Category #1: How are startups making crop development more predictable?

• The big shift: crop development is becoming a programmable loop. The broad idea behind this category is that startups are trying to make plant breeding less random and much faster. Instead of manually crossing plants, waiting for field results, and repeating the process for years, they use genomics, AI, cell biology, gene expression data, and automation to predict better candidates earlier. The goal is not to remove field trials, but to enter them with a much smaller and better set of plant candidates.

• AI is used to accelerate the search period. A plant genome is huge, and the number of possible trait combinations is almost impossible to test manually. Startups such as Inari, Avalo, Phytoform, Plantik, and TraitSeq use AI to identify which genes, gene expression patterns, or genetic combinations are more likely to create useful traits such as heat tolerance, disease resistance, yield, taste, or input efficiency. Basically, they are using models to prioritize the experiments worth running.

• Some companies tune or edit the plant, others improve the selection process. Not every company is doing the same type of “plant design”. Phytoform, Plantik, and Inari are closer to AI guided genome editing or gene tuning. Amatera uses cell level screening to test plant candidates earlier, before growing full plants. TraitSeq reads RNA and gene expression data to detect whether a plant, crop input, or trait is likely to work. Ohalo changes the breeding process itself by making trait inheritance more predictable. So there are many different approaches.

• The platform matters, but the market validates the crop outcome first. What is interesting is that most of these startups are not just selling a “plant design platform”. ****Most commercialize better seeds or crop varieties as well. For example Avalo is commercializing low input cotton and sugarcane. Inari is moving toward high yield soybean designs. Amatera uses coffee and grapes as early proof points. Ohalo is applying Boosted Breeding to potato, strawberry, and almond. Their platform/software is the engine, but the product is the crop performance and since we’re still early they need to sell the final product as a proof.

Category #2: How are startups building the infrastructure layer for plant engineering?

• This category is less about deciding which trait to build, and more about making the trait technically possible.If the first category is about making crop development more predictive, this one is closer to the infrastructure layer of plant engineering. These companies build the tools, workflows, and biological delivery systems that allow new traits to be introduced, expressed, stabilized, or licensed across different crops.

• The major direction is a shift from internal crop development to platform licensing.What stands out from the events is that the strongest companies in this category are not only using their technology for their own products. Pairwise licenses Fulcrum to partners such as Mars, IRRI, Enza Zaden, CSIRO, Ball, Hudson River Biotechnology, and Wild Bioscience. Tropic follows a similar pattern with GEiGS, through partnerships with Corteva, Syngenta, Bejo, and Genus. The category is becoming less like “we build one better crop” and more like “our editing platform can be used by many crop developers”.

• Some platforms focus on precise genome editing.Pairwise is the clearest example here. Its Fulcrum platform is used to create or modify plant traits across many species, from blackberries and cherries to cacao, rice, vegetables, and ornamentals. The important point is not only that Pairwise can edit plants. It is that the company is turning this capability into a repeatable platform that other organizations can access.

• Some companies combine gene editing with gene-expression control.Tropic’s GEiGS platform is interesting because it is not only about changing DNA once. It combines gene editing and RNAi-like mechanisms to control how genes are expressed. This is useful for traits like disease resistance, where the goal is not always to insert a new trait, but to precisely silence or modulate a biological process. Its partnerships with Syngenta, Bejo, and Corteva suggest that this kind of trait-control layer is valuable for large seed and crop companies.

• Some startups focus on hard biological bottlenecks, not only the edit itself.Hudson River Biotechnology is a good example. Its TiGER workflow is about regenerating full plants from gene-edited single cells, which is one of the difficult parts of crop engineering. In software terms, the edit is only one part of the problem. You also need a reliable deployment pipeline that turns an edited cell into a stable, usable plant.

• Earlier companies are still proving where their platform fits.GeneNeer and Antherium look earlier in the landscape. GeneNeer is building around gene editing and trait-expression control, with a clear focus on hard-to-edit crops and a partnership-led model. Antherium applies programmable genetics to ornamentals, which is more niche but still fits the category. For both, the question is whether the technology becomes a broad platform or remains tied to a narrower crop or use case.

• The product is the platform, but the proof comes from partner adoption.In this category, the strongest signal is not only a scientific claim. It is whether serious crop developers, seed companies, food companies, or research institutions use the platform. That is why the Pairwise and Tropic events matter: they show the technology moving from internal R\&D into external deployment. For this category, partner adoption is the closest thing to product-market validation.

Category #3: How are startups turning plants into biofactories?

• This category is less about improving the plant, and more about turning the plant into a factory. In the first two categories, the plant is usually the final product: a better soybean, tomato, potato, grape, or strawberry. In molecular farming, the plant is more like a production system. The goal is to engineer crops so they produce valuable molecules inside their seeds, leaves, or tubers, then harvest the plant and extract the target ingredient.

• The major direction is a shift from technical proof to ingredient commercialization. What stands out from the landscape is that these companies are no longer only saying “we can make plants produce proteins”. They are moving toward field trials, large scale harvests, commercial agreements, and application specific marketing. Finally Foods moved from lab to field trials for casein producing potatoes. Alpine Bio harvested casein bearing soybeans at scale and is now marketing ingredients around real formulation use cases. Miruku is earlier in the public data, but its safflower crop commercialization story points in the same direction.

• Some companies use familiar crops as low cost protein factories. Finally Foods uses potatoes to produce casein, while Alpine Bio uses soybeans. The logic is simple: instead of producing dairy proteins through cows or fermentation tanks, use crops that already have agricultural supply chains. If it works, the crop becomes a scalable and potentially cheaper biomanufacturing system.

• Field validation is especially important in this category. Molecular farming has to prove two things at the same time. First, the plant must reliably express the target molecule. Second, it must do so at a cost and scale that makes sense. This is why events like Alpine Bio’s large scale soybean harvest or Finally Foods’ field trial matter. They are not just scientific milestones. They are signals that the platform may survive the transition from lab biology to agricultural production.

• The bottleneck moves from biology to supply chain and purification. Once the plant can produce the target protein, the hard part becomes more industrial. Companies need to grow enough crop, isolate the molecule, purify it, meet food-grade standards, and convince ingredient buyers that it works in existing manufacturing processes. That is why Alpine Bio hiring for pilot plant and protein purification roles is an important signal: the company is moving from “can we express this?” to “can we make, process, and sell this reliably?”

• The category is still early, but the commercial path is becoming clearer.Compared with the other categories, molecular farming has a very direct value proposition: make high-value proteins or ingredients more cheaply and sustainably. But it also has a harder scaleup question because it touches biology, agriculture, processing, regulation, and food manufacturing at the same time. The most promising companies are the ones moving beyond the science story and showing field production, purification capacity, application fit, and customer pull.

What about regulation?

• Regulation depends mostly on the technique used. The same landscape contains very different regulatory profiles. A company using conventional breeding or cell level selection will not be treated the same as a company using genome editing, RNAi, or molecular farming. So the key question is not only “what does the startup do?”, but “was the plant genetically modified, what trait was added, and where will it be sold?”

• Regulation can be almost invisible, or become a multi-year bottleneck. With conventional breeding, startups usually do not need a full biotech approval. The process is more about field trials, variety registration, and seed market rules, which can still take a few years in Europe. With genome editing, the timeline depends much more on the country and the type of edit. In the US, simple gene edited plants can sometimes move through USDA review in a few months. In Europe, the new NGT framework should make simple edits easier once it applies, but more complex edits will still look much closer to the slower GMO path.

• Europe is moving from a strict GMO framework to a two track NGT framework Historically, gene edited plants were regulated under EU GMO rules. New EU rules adopted on June 17, 2026 create a separate framework for plants made with new genomic techniques. Simpler edits that could also happen naturally or through conventional breeding can be treated more like conventional plants after a verification process. More complex modifications remain under GMO style risk assessment, authorization, and labelling. 

• The US is more product and risk based. In the US, the regulatory path is split across agencies. USDA-APHIS looks mainly at plant pest risk and environmental release. FDA focuses on food and feed safety. EPA gets involved when the plant produces a pesticidal substance, known as a plant-incorporated protectant.

• Molecular farming adds another layer because the crop is only the production system. For molecular farming companies, the regulatory question is not only whether the plant is genetically engineered. It is also whether the extracted protein or ingredient is safe, pure, and properly labelled for food or feed use. If a potato or soybean is engineered to produce casein, the buyer ultimately cares about the ingredient entering cheese, yogurt, drinks, or supplements. So these companies need to prove both agricultural containment and food grade ingredient safety.

What does the funding environment look like?

• This is a slow compounding funding market, not a fast “AI style” category. Looking at the funding rounds in the landscape, the pace feels relatively slow. There are only around 30 funding events across the companies mapped, spread from 2018 to 2026. That makes sense because these companies do not only need to build software, they need biological validation, greenhouse work, field trials, regulatory clarity, seed or ingredient commercialization, and sometimes partnerships with large agricultural or food companies. The funding rhythm follows scientific and commercial proof points, not monthly product iteration.

• The market is very concentrated around a few capital-intensive winners. Inari is the clearest example, with multiple large rounds from Series B to Series G and more than $700M raised across the events captured. Pairwise and Tropic also raised large rounds, especially once their platforms became credible beyond one crop or one internal product. This creates a barbell market: a few companies can raise very large deeptech-style rounds, while many earlier companies are still at seed, grant, or small Series A stage.

• Investors seem to reward platforms, but the proof still needs to be agricultural. The biggest rounds go to companies that can tell a platform story: better seed design, genome editing, trait delivery, or plants as production systems. But unlike pure software platforms, the proof is not only usage or ARR. It is whether the technology can produce better crops, work across species, survive field conditions, and attract serious partners. That is why partnerships with groups like Corteva, Bayer, Temasek-backed investors, Coca-Cola Europacific Partners, and other agri-food strategics matter so much. They are both funding signals and commercialization signals.

• Early-stage funding exists, but it looks selective and milestone-driven. The seed and pre-seed rounds are often modest compared with AI software startups: Amatera raised €1.5M pre-seed then €6M seed, Phytoform raised around $6M across early rounds, GeneNeer raised small pre-seed and seed rounds, and some companies rely partly on grants. This suggests that investors are interested, but they want a clear technical wedge and a credible path to field validation before funding aggressively.

• Non-dilutive funding and grants matter more than in typical software markets. Plantik and TraitSeq both have grant-related events in the landscape. That is not surprising. In a space where the science is risky, timelines are long, and public-good outcomes like climate resilience or food security matter, grants can help bridge the gap between research and venture-scale commercialization.

• The strongest pattern is that funding follows de-risking, not hype alone. The companies that raise the largest rounds tend to have moved beyond “we can edit or design plants” into more concrete evidence: a crop pipeline, a platform that partners can use, field trials, ingredient production, or a clear commercial category. My takeaway is that programmable plants may become a very important market, but it will probably not fund like a fast-moving AI app category. It funds more like a deeptech/agbio market: slower, more concentrated, more milestone-based, and more dependent on strategic partners.