While there is a talk of replacement of lot of jobs by AI (and in some time by robots), agriculture is facing a different challenge. The robots or automated machines are still not good enough, and we don’t have enough people (or enough labor at the right price) to work on the farm in the U.S.

This week I am reposting a guest post by . Austin is a semiconductor analyst, engineer, and entrepreneur. Outside of work, this Iowa dad of four trains for marathons and writes about the semiconductor industry at Chipstrat, and hosts the “Semi-doped” podcast.

This post was originally published in March 2024. Even though Austin no longer works with farmers, this post is very much relevant today as it was in 2024!

Note: I will be at the Ruggedized Conference today in Salinas, CA. If you are going to be there, drop me a note.

The Labor Dilemma

When I talk to farmers, the conversation inevitably turns to the struggle of finding and keeping reliable workers. They're competing for local talent with industries like retail, warehousing, and even work-from-home jobs. Even family farms see their kids choosing other paths. One farmer summed it up: “I have a great team, but am I overwhelmed with job applications? No. Not even in my little agricultural town.”

Government data reveals a shrinking American farm workforce. Surveys by the USDA's National Agricultural Statistics Service show a decline in the number of workers, while the Economic Research Service reports a shift in demographics. Farms are employing fewer family members, with a growing share of the workforce being older and foreign-born. With the world’s population expected to reach 9.7 billion by 2050, farmers need to find ways to produce 50% more food amidst a labor shortage globally.

The core challenge facing global agriculture is this: how to feed a growing population with a shrinking agricultural workforce?

Productivity Must Increase

Let’s dig into the key implication of a shrinking workforce. If we assume a relatively fixed total amount of land in the US, then the decrease in farm workers implies the average worker must cover more acres per season. 

A decreasing labor supply demands an increase in average productivity per worker

The agricultural industry has a history of meeting the challenges of a shrinking workforce by increasing productivity through mechanization – larger tractors, wider implements, and greater speeds.

Productivity has consistently increased by scaling equipment size

This strategy has enhanced productivity for decades, yet there will come a time when relying on scaling alone will not suffice. Fortunately, recent software and hardware innovations have opened the door for new productivity-increasing tools: automation and autonomy. 

Automation and Autonomy will help sustain future productivity increases

Let’s explore how automation and autonomy enhance agricultural labor productivity.

Subscribe now

Automation

Automation is any technology that automates tasks on behalf of the operator. The classic example is automatic steering, where GPS guides the tractor instead of the farmer's hands on the wheel.

Recent advances in computer vision and machine learning give machines the ability to "see".  This unlocks the automation of tasks previously impossible. For example, front-facing cameras on combines can adjust ground speed based on crop density – essentially creating "adaptive cruise control" for harvesting.

Cameras and computers solve agricultural problems in ways never before possible. For example, cameras inside a combine can analyze grain quality in real-time, allowing the machine to automatically adjust its settings for optimal results. This can increase the quality of harvest while simultaneously offloading a task from the combine operator.

Automation Eases Labor Difficulties

The rising adoption of automation in agriculture holds promise due to its potential to boost worker satisfaction and enhance productivity.

First, a great way to hedge against labor issues is to retain employees. Farming involves physically and mentally demanding work, with equipment operators facing long hours and multitasking over extended periods. Automation technology that offloads tasks helps reduce operator fatigue, keeping employees on the farm day after day and season after season. 

Secondly, even the most skilled operators see a decline in focus and efficiency after many hours in the saddle. Automation reverses this trend by maintaining peak productivity throughout the day.

Finally, automation can increase the pool of available labor by lowering the learning curve for inexperienced workers. For example, grain cart automation takes the pressure off new help. The cart automatically matches the combine's speed during unloading, letting the combine operator guide the whole process and creating a smoother, less stressful experience for everyone.

Autonomy

Autonomy removes the operator from the cab entirely. Driverless systems help tackle labor shortages by enabling farmers to reallocate staff and boost the equipment-per-worker ratio.

Autonomy frees operators from the cab, allowing farms to reallocate employees to higher value tasks. For example, tillage autonomy enables customers to address grain hauling bottlenecks during harvest by reassigning operators from tillage tractors to semi trucks and letting the autonomous tractor handle tillage. 

This ability to flex employees to the highest value task across the farm is incredibly valuable, especially for operations with multiple businesses such as livestock or seed dealers. One farmer remarked “Our cows don’t care if it’s the busiest day of harvest – they are still going to give birth. With autonomy we can attend to our livestock and still keep the row crop business moving.”

Freeing an employee from the cab of a tractor with autonomy also enables the farm to scale the output of an employee. Autonomy enables a single operator to manage multiple machines. Take the tillage example from earlier, where a single employee was effectively doing two jobs - tillage (monitoring) and driving a semi. In the future, scaled scenarios could include an operator manually planting while monitoring an adjacent autonomous planter, or a single out-of-cab operator overseeing multiple autonomous grain carts.

From a labor perspective, autonomy also helps operations manage workforce disruptions. Consider the temporal nature of disruptions, which may last for hours (doctor appointment),  days (paternity leave), or weeks (surgery). With autonomy, the tractor can keep running even when down an employee for hours, days, or weeks. 

Finally, autonomy relaxes the one-to-one human to machine constraint, enabling farms to scale or operate in new ways.  For example, an operation may want to acquire more acres, but is currently sized such that they would need another machine and therefore another operator. With autonomy, adding more acres and another machine doesn’t necessarily require another hard-to-find employee. 

Why Not Both?

Automation and autonomy are not mutually exclusive; a farm can employ both productivity tools as they see fit. There’s no one-size-fits-all in farming, and automation and autonomy are simply more tools in the farmer’s toolbelt to increase productivity. 

Here's a simple thought experiment to illustrate. Imagine someone wants to maximize the number of acres they can plant within a limited time. Our first instinct might be to advise a classic scaling approach: a wider planter to cover more rows simultaneously, and a faster planting speed to cover more ground.

For even greater efficiency beyond larger, faster equipment, steering automation offers smaller, yet significant gains by keeping the tractor moving at optimal speed all day.  

Autonomy takes this a step further, allowing a single farmer to manage multiple tractors. For example, the farmer could drive a tractor with steering automation in one field while monitoring a fully driverless unit working the next field over.

Automation and autonomy offer flexible solutions, not rigid "all or nothing" choices. A farmer could strategically use a second autonomous tractor during the peak planting window, especially in large or adjacent fields. However, for smaller fields where a single planter is already efficient, they have the option to operate as usual. Automation and autonomy add valuable tools to the farmer's toolbelt, empowering them to increase productivity in ways that best suit their operation.

Future Labor Dynamics

While automation and autonomy help mitigate today’s labor issues, we’re also beginning to see glimpses of interesting longer-term labor implications.

I’ve talked with farmers who envision a new role on the farm: autonomy manager. This position would optimize the use of autonomous machinery across the farm and ensure maximum uptime. 

Key responsibilities for such a position could include:

• Pre-season Planning: Integrate autonomy logistics into the season's agronomic plan, ensuring machines will be deployed strategically. Develop digital autonomy plans including field boundaries, autonomy preferences, and so on. Address machine maintenance and upgrades based on past autonomy performance.

• In-season Management: Orchestrate autonomous machine deployment, monitor real-time fleet performance, track progress against the plan, and resolve downtime issues swiftly. 

• Post-season Analysis: Conduct deep dives into autonomy performance vs. targets, identifying areas for improvement. Refine strategies for the next season.

The autonomy manager would be a strategic leader on the farm, maximizing the potential of autonomy through a data-driven approach. This role could unlock new levels of productivity and efficiency, ensuring the farm continues to thrive.

This hypothetical autonomy manager role is intriguing because it offers a unique blend: it's tech-driven and analytical, yet firmly rooted in agriculture. This marriage of worlds could be highly attractive, particularly for a younger generation who grew up on the farm and are seeking a tech job, but not a desk job.  Tech-centric and data-driven roles in farming could attract a new generation of talent to the ag industry, offering exciting career opportunities within the local community.

Summary

The shrinking agricultural workforce demands innovative solutions to continue feeding a growing world. While mechanical scaling has been the industry's mainstay, automation and autonomy are additional tools to drive continued productivity growth. Automation reduces operator fatigue, boosts worker satisfaction, and smooths the path for new employees. Autonomy frees workers to tackle higher-value tasks and enables farms to scale without a proportional increase in headcount. These technologies address the immediate labor crisis while laying the groundwork for a future where tech-savvy generations find fulfilling careers in agriculture.

Share

Recently, vertical farming company “80 Acres Farms” announced its closure.

The company has raised more than $350 million, and last year merged with Soli Organic to create what the companies then called “one of the world’s largest and most advanced indoor farming networks.

What follows below is a lightly edited version of what I published last year, when vertical farming company Plenty shut down. I don’t believe the issues faced by 80 Acres Farms were very different than the ones faced by Plenty.

Silo and the Mars series

Apple TV+ is running a mind-blowing show called Silo. (Update: It is in Season 3 now, and episode 5 of the season was the bomb!) It is based on the series of books by Hugh Howey. The premise of the books and the show is as follows: there are a few thousand people who live in a silo that is almost 150 levels deep, because the outside atmosphere is too toxic. The lives of the people living in the Silo are tightly controlled by the powers that be. I won’t give you any spoilers, though the TV version’s storyline is quite different from the books.

How do you feed a few thousand people when you are living in a closed silo which is almost 150 levels deep, with no access to sunlight, a lot of fresh water, nutrients, and other chemicals?

In the Silo world, the people do vertical farming to grow grain, fruits, and vegetables, including animal farming for dairy and meat. (Update: They use bulbs to provide light for photosynthesis, which will be a major plot item in season 3.)

In the fantastic Mars trilogy by Kim Stanley Robinson, when humans first colonize Mars, they construct greenhouses with transparent, radiation-resistant materials and pressurize them to create an Earth-like environment.

Artificial lighting is used to supplement Mars’ weak sunlight during the first few decades of human colonization. As the terraforming progresses, conditions on Mars become more suitable for outdoor farming through the introduction of Earth life like cyanobacteria, lichens, and mosses to initiate soil formation.

These organisms are genetically engineered to survive in Mars' harsh environment and begin the process of creating a fertile layer of soil. After a few decades, the greenhouses transition to fields as temperatures rise on Mars, and the colonists begin to experiment with open-air farming in protected valleys and inner domes.

Plenty takes a massive buzz cut

Vertical farming company Plenty was in the news again a few days ago. Mixing Bowl VC Rob Trice posted on LinkedIn, and the thread has created a big discussion that you should go and read. But based on a news report,

Plenty, which has brought in almost $1 billion from investors like Eric Schmidt and SoftBank Group Corp., is in talks to raise another $125 million as part of the recapitalization, according to people familiar with the matter. The new deal would value the company’s existing shares at less than $15 million, one person said. Previously, investors had valued Plenty at $1.9 billion, according to PitchBook.

The news of a vertical farming company having significant struggles is not new.

The question has been of “when” rather than of “if”.

Previous vertical farming companies which have gone through similar troubles include Aerofarm, Bowering, InFarm and others.

For example, Aerofarm, one of the most prominent players in the vertical farming industry, filed for Chapter 11 bankruptcy and later emerged from it in 2023 by focusing on just one facility for microgreens.

Bowery, another prominent player in the space, shut down completely. New York-based Bowery, which at its peak was valued at $2.3 billion, raised venture capital totaling over $700 million from a cohort of leading firms and A-list celebrities including Justin Timberlake, Natalie Portman and Lewis Hamilton.

Bowery advertised itself as the future of farming powered by 100% renewable energy and requiring 90% less water than traditional agriculture. It also developed proprietary variations on produce like “crispy-leaf” lettuce.

Subscribe now

What problem are you trying to solve and for whom?

One of the biggest challenges with vertical farming has been the lack of clarity in stating the problem being solved and for whom.

For example, a startup called Cloud Produce states this on its website.

In the future, farms will be indoors, allowing nature to regenerate.

It is definitely not clear which problem Cloud Produce is trying to solve, or for whom, by making a definite statement about the future. If anything, barring a nuclear disaster, or we completely f-up climate change, or decide to go to Mars, there is absolutely zero chance of farms being indoors to “allow nature to regenerate.”

When Plenty opened their Compton facility in May 2023, the company news release said the following,

By 2050, demand for food will increase by more than 50%, but, at the current rate of erosion, 90% of all soil will be degraded. At a time when agriculture faces evolving environmental pressures, Plenty has developed a unique farming method that is insulated from those challenges and uses just a fraction of the land and water of conventional farming. The Plenty Compton Farm has the potential to save millions of gallons of water annually compared to field farming.

Companies like Bowery, AppHarvest, and others had a similar problem. They often talked about feeding the planet through higher production (higher yields are needed in traditional agriculture in Asian and African countries, as their yields are very low), using less water, no pesticides, etc.

The attitude was a bit holier-than-thou for problems which cannot be solved by vertical farming. They used a mythology of doing good for the planet, without having a business model which worked or was sustainable. How can you make the planet sustainable if your business model is not sustainable?

If I put on my consumer hat, I want good quality nutritious produce at reasonable prices year-round. If you are going to do it in an environmentally sustainable fashion, you might be able to attract a slight premium from a certain section of the affluent market - think about the millionaire mom in Marin County, or some rich tech person working on the East Coast.

So which of these problems can a vertical farm solve?

Affordability

Vertical farms often struggle to compete with traditional growers because the latter’s produce is often cheaper — in part because open-field agriculture or greenhouses have “sunlight for free” and “use little energy”.

Pricing is a big issue for vertical farms. To make money off their crops given the high production costs, vertical farms need to be able to charge a premium, but it is very difficult to convince consumers to pay a high price for lettuce, just because it is grown indoors. It needs other characteristics, like better taste or longer shelf life, for a consumer to even consider paying a premium. Vertical farms have not delivered on the promise.

Energy costs associated with artificial lighting, temperature control, and ventilation can be substantial for vertical farm companies, limiting profitability. Additionally, the maintenance and replacement of complex vertical farming systems, including hydroponic or aeroponic setups, add to the financial burden.

Construction costs for vertical farms also tend to exceed budgets due to the novel nature of the technology and the limited number of construction experts and contractors experienced in this field.

Energy prices are also very volatile in many markets, which pose additional risks to vertical farming operations.

According to some research and comparison done in 2021 to compare costs between vertical farming (VF) and traditional farming (TF) for lettuce,

The savings are very slim in land and water costs, where vertical farming holds an advantage. Vertical farming does quite poorly when it comes to energy and labor costs, where vertical farming has a disadvantage. The gaps are much bigger for crops like tomatoes or rice.

Farming of any kind is only really profitable at scale. Small farmers who only work it themselves can make some money, but once you start adding labor costs, it eats into your profits quickly. Open agriculture can and has scaled over hundreds of years, while vertical farming is extremely expensive to scale. Traditional agriculture can scale with more land and labor, while energy costs from the sun can stay very low.

One might argue that with automation and energy efficiency, the cost difference between vertical farming and traditional farming could be reduced. This argument might hold true if we assume automation will not address some of the labor cost challenges in traditional farming, which is obviously not a valid assumption.

Subscribe now

Product differentiation and quality

Vertical farms could compete on premium products like strawberries, differentiation through a different taste profile, nutrition, or shelf life for products like leafy greens, etc. Differentiation will be a fairly hard story to tell for vertical farming, unless the products are available year-round, have a very different nutrition profile, and are much more shelf-stable than traditional agriculture products.

This is very challenging, as they would still be competing with traditional agriculture products, which have similar vectors of differentiation, other than shelf life. Year-round availability has been solved through different growing regions and supply chain solutions.

Some vertical farm products are considered better quality because they have different characteristics, like crispness for leafy greens like lettuce. Vertical farms also don’t have to use pesticides, etc., due to a controlled environment.

The lack of pesticide use could be seen as a signal of higher quality by consumers, and they might be willing to pay a premium for certain products, though this is a very small segment of the population.

Consumers have a similar perception about traditional organic products, as they are willing to pay a premium for organic products. Here again, it is not clear whether vertically farmed products have a clear, value-accretive quality advantage over traditional products.

Access

Do vertical farms solve the problem of access? Are they able to provide produce to communities which don’t have easy access to fruits and vegetables? (We are talking only fruits and vegetables, because vertical farms for commodity row crops like corn, rice, wheat, soy have been a non-starter so far). None of the vertical farming facilities are either located near so-called “food deserts” or supply to “food deserts.”

The facility in Compton was shut down in less than 18 months. Walmart and Whole Foods were advertised as buyers of produce from the Compton facility. Walmart and Whole Foods could not be more different in terms of what they would need from a supplier, because their customer base and their expectations are very different.

The location of the facility in Compton is a head-scratcher, as you are near some of the biggest salad bowls in the world in California and Arizona, which can produce crops quite efficiently.

If the vertical farm can produce a crop all year round, then it can create some additional value for the consumer. With crops coming from traditional agriculture, we have a notion of “season” for a crop. The industry has tried to solve this problem through supply chain solutions like storage and moving the supply region for the crop based on season.

Can a vertical farm be set up at a retailer’s distribution center, or better yet, at a grocery store itself? This would provide vertical integration for a grocery retailer for certain crops, and help them eke out a few additional points on margin, given most of them operate on single-digit margins.

These have not been viable options due to the minimum scale requirements to set up a vertical farm to have any hope of profitability (which actually does not exist). There is a reason why you could not make any money by growing some tomatoes in your backyard and selling them at grocery store prices.

None of the major grocery retailers have been strategic investors in any of the major vertical farming companies. This should be a big tell.

We have very efficient supply chains, which can ship products to the right region from the relevant growing region based on the season. A rise in transportation costs due to fuel prices and driver shortages does play in vertical farming’s favor if the majority of the customers are located within a short distance from the vertical farming facility, though vertical farms have not been able to exploit this to their advantage.

Resource utilization

Vertical farms in controlled indoor environments are promoted as a sustainable way to grow fruits and vegetables closer to the point of consumption while using less water and less land.

Vertical farms do use less land on a yield basis due to stacking. But using land as a denominator is not the right metric to measure yields for vertical farming, as at the end of the day, it is the landed cost to the consumer on a per-unit basis that matters.

Also, when you think about fruits and vegetables, traditional agriculture acreage is a very small percentage of total farming acres. According to the 2017 Ag Census, only 10 million acres of farmland were used for fruits and vegetables in the US, compared to hundreds of millions of acres of farmland for commodity row crops like corn, soy, wheat, rice, cotton, etc. So the argument about land, at least in the US, is BS.

As mentioned earlier, vertical farms do use much less water compared to traditional agriculture.

Do vertical farms make more sense in areas where it is difficult to grow produce due to climatic conditions? For example, due to desert-like conditions and a lack of fresh water, it is very difficult to grow produce in the Middle East. For example, countries like Saudi Arabia, UAE, and Qatar import 80% of their food needs, especially for meat, dairy, and produce.

Plenty signed a deal in the middle of 2024 with the UAE to open 5 vertical farming facilities just for strawberries, with a $680 million joint venture.

The oil-rich countries have the money to grow very expensive produce, but they might be better off just solving the problem through better trade deals and terms.

Some people have made the argument of national security, which I would label as “salad security.” Grains like rice and wheat, and products like meat and dairy, have a much bigger impact on national security and self-reliance than salad products or strawberries grown in vertical farms.

First principles thinking and implications

Based on the analysis of affordability & quality, access, and resource utilization, vertical farming appears to be a solution for problems that do not exist (also known as the Juicero syndrome). We have seen a lack of first principles thinking within this space over the last few years.

We need to explore better business models to make produce more easily available to a large set of customers at affordable prices, while being profitable for the grower/supplier and the grocery retailer.

We should be thinking about investing in the infrastructure to improve supply chains (for example, better tracking and tracing, better cold storage facilities) and make them more responsive. For example, can we improve the crumbling port infrastructure in the US to be more competitive with ports worldwide?

We should be thinking about energy, which is a much bigger issue than water, when it comes to vertical farming.

We need to move past the storyline of growing more food at a macro level when we are talking about a company, as it is disingenuous. There are many opportunities in traditional agriculture in Asia and Africa to grow yields by 4x to 5x through better seeds, nutrition management, and advisory services, as opposed to trying to do it through vertical farming.

We need to move past resource utilization stories in vertical farming, as there are many opportunities with bigger impact in traditional farming to reduce inputs through precision agriculture, better genetics, and better supply-region selection.

We need to think about growing food as a growing food business, where technology is an enabler. Many of the vertical farming companies have labeled themselves as technology companies that do farming, rather than the other way around. This has created mismatched expectations around value creation, timelines to profitability, multiples, and the types of investors they have attracted.

Vertical farms must strike a balance between leveraging technology and adhering to agricultural best practices to ensure successful operations. Rather than obsessing over becoming the next cutting-edge farming technology stack, vertical farms should prioritize their products, customers, branding, marketing, distribution, and operations. Resolving this identity crisis requires effective messaging from leadership to investors and team members.

At best, vertical indoor farming seems like a debt-financed infrastructure play, rather than a VC play.

As it relates to first principles thinking, I will point you to a LinkedIn post done by Cassidy Johnson, with the title “What AgTech companies will succeed in the US in the 2025 (and beyond?)”

Her first point is right,

They'll be solving an actual, existing problem, not building a product no one wants or needs.

I wish all of us, including the vertical farming industry, would take a lesson from Cassidy.

AgTech Alchemist and rabble rouser Walt Duflock made a similar point in his LinkedIn post.

Startups should be able to tell you the exact problem they solve and the rough economics of the solution (bonus points if they have a customer that's paying them that you can talk to).

With the downturn in VC funding over the last 18-24 months, these stories impact the space in terms of inflows. It puts a damper on the whole food and ag space for VCs who have over-invested in vertical farming.

Subscribe now

So where do we go from here?

Adam Bergman reacted to the Plenty news with his own LinkedIn post, which made some very interesting and valid observations.

Recapitalizations, like Plenty’s, will hurt key stakeholders, particularly employees who will lose jobs and investors who will lose money. My hope, however, is that this will provide an opportunity for Plenty to right-size the business and put its full focus on its Richmond, VA strawberry farm. This could allow Plenty to follow a similar playbook as AeroFarms, which, following its recapitalization, closed most of its operations and now has one farm in Danville, VA focused on microgreens. AeroFarms is showing a path for other vertical farming companies: focusing on product differentiation, optimizing production, and achieving profitability prior to expanding operations.

As I have stated many times before, although I believe there is a short-term opportunity in the U.S. for vertical farming companies to produce traditional leafy greens, in the not-so-distant future, those companies will struggle to compete with greenhouse-grown leafy greens, forcing them to pivot to other crops, including berries, coffee, forestry, microgreens, specialty ingredients, and pharmaceutical products. With all the challenges (climate change, high input costs, increasing labor costs & worker availability, growing regulations) facing the traditional outdoor-grown produce sector, I still believe that indoor farming, including vertical farming, has enormous potential in the coming decades.

Adam is quite right that if vertical indoor farms have to survive to fight another day, they need to focus on product differentiation, optimizing production, and achieving profitability. As we have discussed earlier, vertical farming has struggled and will struggle when it comes to any of these vectors when compared to traditional farming.

The traditional outdoor-grown produce sector does face challenges like climate change, high input costs, increasing labor costs, worker availability, and growing regulations. Traditional agriculture has been working to solve these difficult challenges through tools like precision agriculture, non-chemical operations like mechanical & laser weeding, automation, shifting supply regions from California to Peru & Mexico, and farming systems like greenhouses and horticulture.

Based on science fiction stories like the Mars trilogy by Kim Stanley Robinson, I want to leave you with a simple decision tree to decide whether you should try vertical farming and invest in it or not.

Share

There is a paradox at the heart of agricultural innovation. I attend many university labs and startup demos and see an amazing pipeline of ideas, including gene-edited crops, microbial inputs, autonomous vehicles, sensors, and robotics. And yet only a small fraction of them ever reaches a working farm. The technology gets invented, validated at pilot scale, and then often stalls.

Sarah Garland has spent her career working on that gap. A plant scientist with a PhD in CRISPR gene editing from Cambridge and postdoctoral work at Columbia’s Earth Institute, she founded Triple Helix Institute for Agriculture, Climate, and Society. Triple Helix is a nonprofit designed to serve as the connective tissue among researchers, entrepreneurs, investors, policymakers, and the public. Her premise is that much of what holds agricultural technology back is coordination, capital, and trust.

I spoke with Sarah shortly after Triple Helix launched its newest experiment: an Investment Council that brings together eight investors from across the capital stack, traditional venture firms, corporate venture arms, a mission-driven family-office platform, and a prize foundation, to ask why good agtech stalls before scale, and what could unstick it. Our conversation ranged across the organization’s origins, the economics of farmer adoption, and the design of the council itself.

What follows is a lightly edited version of our conversation.

The Organization That Didn’t Exist

Rhishi Pethe: What is the origin story of Triple Helix?

Sarah Garland: I’m a plant scientist by background, and it’s inspiring to see all of the emerging innovation at the discovery stage. But I became increasingly interested in how innovation makes it from the lab to the field. Or, often in agriculture, it does not make it from the lab to the field. So after my PhD, I stayed in academia but pivoted more into science policy and the role of biotechnology in developing more nutritious and climate-resilient food systems.

I became concerned on two fronts. The first was the lack of productive interaction among stakeholders, including academic researchers, and the manner in which they communicated their cutting-edge science to the rest of the community. The second was the polarizing rhetoric in many high-level climate and environmental circles around technology in agriculture. I was hoping to find an organization to join that shared my evidence-based and nuanced vision for the future of agtech. But I found that the NGO space was similarly polarized. So I created the organization I wished existed at the time, Triple Helix.

As a nonprofit, Triple Helix is dedicated to building cross-sector engagement and awareness around ag tech for climate action. It bridges research, industry, policy, and communication to align stakeholders around credible information and common goals that advance agricultural resilience. The wonderful part of being a nonprofit is the ability to bring together the best of academia, private-sector innovation, and government mechanisms to strengthen the entire ecosystem.

Rhishi Pethe: If you look at the current ecosystem, you have venture capitalists, startups, big business, and academia. What is missing? What other options did you consider when you thought, “This is what I want to do, and it doesn’t exist”?

Sarah Garland: There are great benefits to being in academia, particularly the freedom of thought it brings. But I was discouraged by the lack of emphasis on engaging with practitioners. On the other side, practitioners aren’t always challenged to implement new ideas in the same way that academics are trained to do.

When it comes to science communication, outreach is often unidirectional. “Let me tell you about my science.”

In fact, we really need to listen more as researchers and understand what different types of stakeholders need to make informed, evidence-based decisions, while remaining within frameworks they can actually act on. In my current role, I’m able to bring all these pieces together.

Rhishi Pethe: How does Triple Helix actually work? How do you operationalize the idea of connecting different groups and bringing them together, given the challenges you talked about on both sides?

Sarah Garland: From the very beginning, it was important to involve different perspectives in our work. That means cross-sector and cross-discipline. I was really pleased to build a fantastic advisory team that bridges expertise, whether that’s government policy experience (we have a former Secretary of Agriculture, Dan Glickman, on our advisory team) or deep technical knowledge ranging from AI in agriculture to cutting-edge gene-editing methods. So, already in terms of the advisors we were able to put together, it showed the direction the organization was aiming to go.

Ever since then, whether through tasting events, panel discussions, or co-authored reports, everything we’ve done has brought together different perspectives, all in service of achieving agricultural resilience together, rather than further fanning the flames of division that often get more airtime.

Share

Rhishi Pethe: You mentioned agricultural resilience. How do you define agricultural resilience, and what does that actually mean? How does the industry or a policy organization engage with you? Do they come to you with specific questions, or is it more that you’re acting as a convenor and putting forward a point of view?

Sarah Garland: One half of agricultural resilience is being strong in the face of different types of shocks and threats, whether that’s a climate shock or a supply-chain shock, both of which we’re very much already experiencing. The other half is not contributing to harm. For example, we’ve done a lot of work on innovation for biodiversity conservation.

You’ll see emphasis on both parts of resilience in the topics we cover. We’re looking to make agricultural systems more efficient and productive while also promoting nature conservation and environmental stewardship.

For the second question, it’s definitely more the latter. We have a point of view that comes from what I just said about resilience. From this cross-sector, cross-disciplinary perspective, we ask, “What are the key lever points here that we need to work on?” Whether it involves public-sector mechanisms, investment, or public understanding, the content side always comes from the Triple Helix. Then we identify the right stakeholders and leaders to bring into our work.

Rhishi Pethe: Is your theory of change that, by bringing all these different perspectives together and providing a point of view, you hope to influence some players in the ecosystem to act differently? How do you hope to bring about change within the ecosystem?

Sarah Garland: Triple Helix catalyzes high-impact ag tech innovation through facilitating stakeholder collaboration and strengthening public understanding. Those are the two verticals we work within. Very few organizations focus on the translation layer between researchers, entrepreneurs, policymakers, and the public. By filling that gap, we drive the entire space forward.

In agricultural circles, division often leads to counterproductive outcomes. We’re stalling ourselves when we can’t move past dogmatism about the future. Triple Helix can mend some of that discord. Fostering new types of collaborations, bringing together people who wouldn’t normally talk, and seeing that they actually have a lot in common, is an important role we play in this ecosystem. It’s been really encouraging to see the lightbulb go off in people’s minds: “Oh, we’re working on the same problem. How can we move forward together?”

Rhishi Pethe: I’m assuming you started with a set of assumptions about how this would play out. Looking back, are there certain things that surprised you, “Oh, I didn’t think this was going to happen,” or “This is completely different from what I thought”?

Sarah Garland: When I started Triple Helix, food and agriculture were kind of a footnote in the climate conversation. Part of the organization’s early impetus was to build awareness of the importance of food and agriculture to climate action. That has changed a lot in a short amount of time. There’s more recognition of it now. But unfortunately, many climate efforts still predominantly focus on energy. The amount of investment in food and agriculture, compared with the rest of climate investment, remains small.

So something I didn’t expect was the rapid increase in attention on food and agriculture. It’s good that we have increased attention. But within that, we need to ensure we’re making evidence-based decisions rather than falling prey to whatever narrative is popular at any given moment.

Subscribe now

Why Promising Technology Stalls

Rhishi Pethe: When we last spoke, you said capital constraints shouldn’t stall innovations that are solving critical challenges. It sounds like a conclusion you’ve reached, or something you’ve learned over the years. How did you get to that point?

Sarah Garland: There is currently a mismatch between the urgency of the challenges facing agriculture and the situation we’re in, in which promising technologies aimed at addressing them often stall before they can be deployed at scale. Many factors contribute to that, of course.

Recently, we looked more at the public-sector side. We recognized that historically, the public sector has played a key role in advancing agricultural innovation, particularly in the United States. So we conducted a project to identify opportunities for the federal government to catalyze US ag tech innovation in a modern context. Unsurprisingly, we found that public-sector mechanisms, everything from funding proactive basic research to creating consistent regulatory pathways, are essential for de-risking emerging technology and creating an effective innovation ecosystem. But that’s only one part of the ecosystem.

Through this work in the public sector, the question kept coming up.

“What about private-sector mechanisms?”

Capital allocation is a determining factor in what innovations make it from discovery to impact in practice. That raises the question of what meaningful innovation on the private-investment side would look like, alongside public-sector innovation and the ongoing technical innovation. So that’s how we came to the topic of capital allocation and investment. It’s about understanding the key levers that influence whether high-impact technology actually reaches farmers and, eventually, consumers.

Rhishi Pethe: On the venture side, venture dollars have declined quite a bit over the last few years. But the challenges to scale or adopt also exist at the farmer level, right? When new technology comes along, it has to solve a problem, be economically viable for the farmer, and be easy to use. Multiple factors must be in place before it can scale. So, how are you thinking about that particular challenge? Yes, there’s a promising technology, but that promise must be realized through adoption and scaling.

Sarah Garland: It’s important to frame this by noting that farmers take on considerable risk when they decide to adopt a new tool. Innovators really need to understand that and provide sufficient technical validation for their product when they’re asking farmers to adopt it. So that’s point one for finding early adopters. Building trust by saying, “I understand this is a risk for you, but this is why we think it will benefit your operation.” Respecting the farmer’s perspective is really important for driving both early adoption and, of course, eventual scaled adoption.

Rhishi Pethe: That’s true for any technology. It has to work without posing too much risk to the person adopting it, whether that’s a farmer, an accountant, or anyone else. So, what perspective or point of view is Triple Helix bringing to the discussion to say, “Hey, we need to look at this differently than how it has been done”?

Sarah Garland: There are different stages and transition points a technology goes through: early discovery to first validation, validation to early adoption, and early adoption to eventual scale. So, from validation to early adoption, that framing of the risk farmers are taking will help overall adoption by ensuring end-user consideration is at the forefront from the very beginning of product design.

Then, if a technology has already been validated and farmers are eager to adopt it, it’s a major loss impact-wise if capital constraints impede its deployment at scale. Triple Helix engages with stakeholders across the ecosystem to improve these transitions, enabling the implementation of high-impact technologies to address critical challenges.

Inside the Investment Council

Rhishi Pethe: You recently formed this investment council. It’s a very interesting, diverse group of people you’ve managed to put together, so congratulations on that. Could you go behind the thought process? Why did you feel you needed a group of people to come together to talk about these issues?

Sarah Garland: Through our previous work and understanding the main levers in the space, it became clear that investment and capital allocation were topics to address. No one knows the opportunities and constraints facing ag tech investment better than investors themselves. So it was a no-brainer to collect perspectives from leading practitioners. We’ve demonstrated success with this strategy in our policy work, gathering insights by hosting stakeholder roundtables across the country. This council furthers our model of being a convenor: bringing together the right people who are open to thinking outside the box and generating new ideas. Few forums currently exist where these types of investors can come together, operate outside their silos, and be candid with each other.

This council brings together top investment minds in the space. To have impact, we didn’t just want a group of entirely venture-capital investors or entirely philanthropies, because then you’re only getting insight from one slice of the capital stack. I’m really thrilled by the group of participants we have assembled for the inaugural investment council. They are each extremely impressive in their own lanes, but they’re also committed to strengthening the entire ecosystem. They represent different stages of investment, different types of capital, and different perspectives on the future of agriculture.

Subscribe now

Rhishi Pethe: What types of issues do you want this investment council to discuss? What problem statements are you throwing at them? “We think these are the top three problems, and they’re inhibiting the direction we want to go,” whether it’s resilience, adoption of technology, or a better future for farmers, or the food system.

Sarah Garland: What’s great is that I’ve been able to get the council’s input from the very early stages of this work. A lot has been written about the challenges facing ag tech. The point of the council is not to rehash the challenges. We obviously need to start with a shared understanding of them to move forward with developing solutions. But the key issues we’re beginning to discuss were not prescribed in advance. We began by having the council members themselves identify their top few challenges in this space. There’s a lot of overlap because they are all in the same general sector, but different challenges also emerge depending on the specific part of the pipeline they work in most.

So the entire process is collaborative. It’s less direct than saying, “I want you to answer this question,” but our process is to be a facilitator and convenor, not to dictate the conversation fully. We’ll get more meaningful outcomes by working through the struggle together.

Rhishi Pethe: You have traditional VCs, you have corporate VCs, you have someone with XPRIZE experience. When we talk about agricultural resilience, will you focus more on the problems facing agriculture in the developed world? What are the boundaries of the problems you’re looking at for this investment council that they have to discuss or are willing to discuss?

Sarah Garland: Science, of course, has no boundaries. Innovation is happening all over the world, and that’s really exciting. For the sake of this council and investment mechanisms, we’re primarily focusing on the US ecosystem, and any policy implications would be US-centered. We obviously operate in a global trade environment, so I expect that insights will emerge from comparisons of what different countries are doing better or worse, and how to develop the entire cross-sector implementation ecosystem. But our focus on strengthening the pipeline from the discovery stage through to impact in practice is currently on the US.

Rhishi Pethe: So it’s US-focused; it’s a lot of science; it’s about how you do investments. Will there be discussions around policy and financing models?

Sarah Garland: There are funding mechanisms and external constraints that influence how capital moves through the system. It’s important to consider these factors together, because otherwise you end up putting band-aids on aspects of the problem that aren’t really the root cause. This group will be exciting to work with as we uncover the underlying constraints and identify the highest-impact solutions.

There are also opportunities to guide entrepreneurs, so they have more insight into what investors are seeing. Rhishi, you work with a lot of spinouts from the research-to-early-entrepreneurial side. It’s never too early to think about what a growth-stage investor is thinking and to make sure you’re setting yourself up for success. But at the moment, we’re focused on leverage points that will unlock the entire ecosystem.

Rhishi Pethe: I’m guessing some of the incentives for a corporate VC are a little different from those of a traditional VC, versus someone looking at government- or state-provided funding. So do you let the council figure out those tensions? How do you create an environment where these conflicts of interest don’t degrade the quality of the conversation that needs to happen? What role are you specifically playing to make sure that doesn’t happen?

Sarah Garland: I think the fact that they have different incentives is a strength of the council. If you had everyone coming from a corporate perspective, for example, you wouldn’t see the other actors and the roles they’re playing. So I see those tensions as a benefit rather than a constraint. I invited these individuals specifically because of their willingness and ability to listen to other perspectives and collaborate across their own interests.

Rhishi Pethe: If you hear all these diverse voices represented on the council, are there any voices missing? Do you feel this will come at these issues from all possible angles? Are there any blind spots?

Sarah Garland: Our council members have direct experience in their current roles, but many also have backgrounds in other aspects of ag innovation, including government work. I’m optimistic that we’ll be able to address a pretty wide range of challenges.

But if we reach a point where we realize we need more perspectives, there’s always an opportunity to bring in more people and grow the council in the future.

Rhishi Pethe: Are they also going to talk about different funding models to support innovation?

Sarah Garland: It’s all about identifying the core levers. If a different funding mechanism would indeed catalyze the space, then that’s an avenue worth exploring. You mentioned earlier that we have someone from XPRIZE on the council, and I’ve been involved in the XPRIZE orbit for a couple of years. That’s a bold model for incentivizing innovation. When there are critically important areas of innovation that existing types of capital aren’t addressing effectively, it’s time to think outside the box and maybe try something new.

Common Ground, and the Road Ahead

Rhishi Pethe: So, how will you communicate what the council is discussing? Is that going to be in the form of thought papers? How do we access the value of the discussions between you and the council members?

Sarah Garland: We just got the ball rolling at the beginning of June, and we’re working together to pinpoint the actionable areas we want to explore further. After a few more meetings, we’ll have that direction more solidified, and we’ll be able to share some early ideas and insights. From there, the goal is to engage more stakeholders in bringing those concepts to life.

Rhishi Pethe: It’s a very interesting idea to bring this council together to debate and discuss different ways innovation should be funded, financed, rolled out, and adopted. If you look further out and you’re looking back: “Hey, this was a great idea. Sarah, great job, you put this council together, and it worked out fantastically well.” What are some things you would see changed for you to say, “This worked really well”?

Sarah Garland: I expect that insights and strategies derived from this council will lead to better coordination across capital types at key transition points, so that no high-impact technology is left behind. That will come from breaking down silos, developing a common understanding, and aligning toward a shared vision, thereby facilitating the flow of capital more effectively.

I also think it’s important that this initiative builds on our previous work on public-sector mechanisms. We need to prioritize both private- and public-sector innovation to implement high-impact agtech in practice. The more we can coordinate all the levers at our disposal, the better.

Rhishi Pethe: Let’s say a few years down the line, it’s, “It wasn’t a bad idea, but it just didn’t work out the way I thought it would.” If you’re doing a premortem, what are some reasons this wouldn’t work?

Sarah Garland: Bringing people together to tackle hard questions is never a bad idea, so everything we learn from this experience will be a meaningful contribution. In science, you often publish only when you get exciting results, but negative results are also really important for building our knowledge base. So everything we learn from this experience will have an impact.

There are inherent challenges in agriculture that are daunting, and coordinating actors across sectors and across incentives is easier said than done. But it’s important to foster that type of vision. There are eight members of the council, but the aim is for this to have an impact far beyond those eight members and to inspire more thought about how we can be creative about agtech investment, rather than just feeling stuck where we are right now. The more we can provide optimism and solutions, the better.

Subscribe now

Conversations about the bioeconomy in San Francisco or Boston tend to focus on invention.

The next clever microbe, the next breakthrough protein.

Eshan Samaranayake came to food and agriculture by an unlikely route. He trained as a molecular biologist in Hong Kong and was drawn to the commercial side of biotech. After cutting his teeth at a food-biotech startup, he began writing a newsletter that helped him enter the venture capital industry. Today, he invests across Asia-Pacific in upstream technologies, including ingredients, fermentation, and ag inputs.

From Asia, the bioeconomy looks like a problem of scale, economics, and coordination. Over the course of our SFTW convo, we discussed what outsiders misread, why venture capital is an awkward fit for parts of food and ag, why TAM can be a trap, and why cultivated meat’s timelines slipped by decades.

We discussed Eshan’s path to VC, what working in a lab teaches you about scaling biology, and the economics of feedstock. Eshan finally reframed the discussion, noting that the binding constraint today is no longer science but the ecosystem around it.

The actual conversation has been lightly edited for clarity.

An outsider’s path in, and the venture-capital question

Rhishi Pethe: Give us a little bit of your background. How did you get to the point and the role that you’re in right now?

Eshan Samaranayake: I’m Sri Lankan, born and raised in Sri Lanka. I moved to Hong Kong and studied molecular biology and biotechnology. I’ve always been fascinated with biotech. Since I was 12, I knew I wanted to do it. When I first heard that scientists could create glow-in-the-dark carrots by inserting a firefly gene into carrots, I was fascinated.

I thought I would get into academia and do a PhD in gene editing. I still love biotech, but I didn’t want to go into academia, mainly because, for me, the value of biotech lies in commercialization. Academia is the backbone of biotech. You need the research before you commercialize. I want to help get that research to the end user. That’s why I decided to stay in biotech, but on the commercial side.

I graduated from my bachelor’s program in 2021, in the middle of the pandemic. At that time, there was a lot of talk about food insecurity, and that’s really when I started reading about the food system. There’s so much waste and inefficiency in the food system. Chicken is probably a bit of a graphic example, but with chicken, you feed in 9 calories and get 1 out.1 That’s pretty wasteful. So I saw an opportunity for biotech to improve the food system.

My first full-time job was for a food biotech startup in Hong Kong. I was very keen on joining a startup because I wanted firsthand experience of what it takes to build something in this field. I was at the startup for almost three years. I was wearing multiple hats. Even though my title was product manager, I was doing product development, R&D, fundraising, marketing, and business development.

I was so interested in learning everything I could about the food and ag tech space that I was consuming all this content, great articles from folks like Green Queen and AgFunder, and listening to all these podcasts. Because I’m based in Hong Kong, which isn’t an ag-innovation hub, I had limited opportunities to discuss this with peers. But I knew there were people around the world who were really passionate about this, and I wanted to reach them.

I’m already consuming this content. I’ll share my learnings in a newsletter, upload it, and let the internet help me find my people. The newsletter did two things. It was an accountability tool for me to learn about the sector, but also a touchpoint and a way to reach out to people in the agrifood and ag tech space globally. I was able to get in touch with more founders and investors globally. It opened my mind to the fact that startups are the best tools we have to address the major challenges in the food system.

How can I help multiple startups? That led me to the VC side, and I joined Better Bite Ventures two years ago.

A bit about Better Bite: we invest in early-stage food and ag startups based in APAC. We typically invest in upstream technologies. We don’t do much CPG or consumer-facing work. We’ve made over 30 investments across eight countries in the region2, so we’re pretty active in the sector.

Rhishi Pethe: That’s a really interesting journey. Here in the US, there’s a prevailing narrative that the VC model isn’t the right way to fund ag startups. Is it true, and if it’s true, under what conditions is it true?

Eshan Samaranayake: There is definitely some truth to that statement, and it is good that more people are saying it openly, including some of the most experienced GPs in the space. Honesty is good for the ecosystem.

I think it comes down to managing expectations with LPs. If you pitch it to LPs as a high-velocity technology sector and then deliver the returns of a patient capital strategy, you have a structural mismatch between what you promised and what the sector can deliver. So LP selection and expectation management from the start matter a lot. Food and ag can generate real returns, and it can attract capital that cares about more than pure financial return, impact-aligned LPs, development finance institutions, and corporate strategics with long horizons. Those are often better matched to what the sector actually needs.

So, it’s not that VC should not exist in food and ag, but that the VC model imported from software, large funds, short timelines, and high velocity, is structurally mismatched to agricultural transformation timelines. The sector needs a mix. So things like early-stage risk capital that is honest about its role and return profile, patient growth capital that can hold through long commercialization cycles, and development finance that absorbs systemic risk private capital rationally avoids. None of those is sufficient alone.

On whether VC is the right model at all, the answer depends on the stage. At the pre-seed and seed stages, VC is often the only viable source of funding for startups. No bank funds a high-risk startup with no revenue and no hard assets. Grants come with criteria so specific that they fit a narrow slice of what founders actually need. Private equity comes in far too late. So if early-stage VC does not exist in this sector, a large number of technologies never reach the stage that would allow them to raise a larger round from a more patient capital source. That enabling function is real, and I feel it is under-discussed in the critique of VC in agrifood.

The whole sector generally needs more exits. It’s about managing expectations while determining which technologies align with the 10-year horizon.

Going upstream, and the hard reality of scaling biology

Rhishi Pethe: You mentioned you typically invest in upstream technologies. What are some of the sectors or types of technologies you’ve invested in?

Eshan Samaranayake: On the food side, ingredient manufacturing, whether it’s proteins, lipids, coffee, or palm oil. Typically, they’re selling to other businesses, not to consumers. On the ag side, it would be about ag inputs: fertilizers, biostimulants, and seed treatment. We have a portfolio company called Rainstick that’s treating seeds with electricity to improve germination rates. And then the platform plays. Another portfolio company, Algenie, is developing novel photobioreactors for microalgae cultivation.

We realized that, especially in APAC, investors tend to be more risk-averse. The minute they see a technology that’s very deep tech or very science-heavy, they tend not to get involved. But we like those very science-based technologies, because that’s how you get the kind of breakthroughs you need to address some of the major challenges the sector is facing. You don’t want technologies that are making something incrementally better. That’s just not going to cut it for the type of returns you want a VC to have. So we’re looking for technologies that are very science-driven and aim to do something new rather than something incremental.

If you do something consumer-facing, your multiples are typically capped at, let’s say, 2-6x revenue for a CPG deal. But if it’s something very tech-driven, you have the premium for the technology. And these tend to be platform plays that could eld expand into other verticals beyond food, like chemicals and materials. So that also makes more sense from a venture standpoint, to command those premiums that you’re looking for, that optionality.

Rhishi Pethe: You started your career at a startup focused on microalgae production, and then moved to VC. What did you learn that you think you’re able to apply now in your current role as you think about these different investments?

Eshan Samaranayake: The first part is probably kind of boring, because it’s a soft skill. It’s just having more empathy for founders. I feel like more VCs should have experience working at a startup before they write a check or even analyze these startups, especially early-stage ones, where it’s all about the team. There’s so much risk, so many things that could go wrong, and that affects people’s personal lives. You tend to have a lot more empathy for founders who are taking these big risks to tackle these major challenges.

On the tech side, what’s kind of undercovered in most VC due diligence for these early-stage deals is the focus on upstream economics. You talk about yield and titer, but depending on the product, 50% of the ingredient cost can be just downstream processing. So, dewatering, drying it if it’s microalgae, and, if it’s a precision-fermented product, filtering the product from the microorganism. That cost adds up, and I feel it doesn’t get enough coverage in due diligence and in the media.

Whatever works at a two-liter scale, once you get to production scale, once the physics of biology hit, there’s really little you can do about it other than going back to the drawing board and doing more iterations. There’s no easy way to get around the scaling challenge.

One other thing that probably doesn’t get appreciated enough, especially if you’re cultivating a microorganism, is the effort it takes to avoid contaminating the samples. Whenever you open any culture containing microorganisms, sugar, and all the feedstocks, there are always microorganisms around us. They can come in and ruin everything. That is not a trivial thing to solve. It takes effort to control contamination because if you contaminate the sample, your whole batch is ruined. You have to start from the top, and cleaning is a pain in the ass.

Microorganisms are lazy. Energy-efficient is a better way to put it. If you’re using precision fermentation, you’re engineering it to make something it usually won’t make. They are not incentivized to make it because it takes a lot of energy to produce this dairy protein. Over time, the microorganisms that don’t make the whey protein make up most of the culture, and the ones that actually make the dairy protein are very few, because from an evolutionary standpoint, they’re incentivized not to make the dairy product. You just realize how energy-efficient these tiny creatures are, for better or worse.

You end up having these non-producers that make up most of the population. There’s this interesting company called Enduro Genetics3. Their fix for this global problem is to tie the microorganism’s survival to protein production. So if the microorganism doesn’t make the protein, it’s going to die. That’s the genetic switch that Enduro Genetics is making.

What the West misreads about Asia’s bioeconomy

Rhishi Pethe: You mentioned that investors in Hong Kong, or LPs who work in APAC, are more risk-averse compared to other places. What are some of the things that people like me, sitting in San Francisco, don’t understand about what’s happening there?

Eshan Samaranayake: I think Western media typically tends to overweigh invention but doesn’t cover scale as much. We’re at this stage in bio-manufacturing where technology is not the bottleneck. You have all these cool technologies, but how do you scale them and commercialize them?

Asia is the bio-manufacturing hub in terms of scale. China and India have already done that with biopharma. They’re the manufacturing hubs for biopharma. And now we’re getting to non-biopharma manufacturing, because India, with the BioE3 policy4, is investing heavily in bio-manufacturing. We definitely need more capacity for scaling up. There’s some cool stuff happening with BioMADE5, which is great. But generally, Asia tends to be the manufacturing hub, and I think that’s going to be the case for the bioeconomy as well.

When international investors look at Asia as an investable region for ag, they get carried away by the “TAM trap”. There are a few billion people here, so why not expand to this region and capture these markets? But the thing is, agriculture in Southeast Asia is run by smallholder farmers who don’t really have the money to spend on these innovations. So the TAM isn’t telling you the whole story. But who’s actually paying for these innovations? That’s still much smaller than what the TAM implies.

They say Southeast Asia is a big market, but that’s many different countries with their own distinct, localized markets, policies, value chain dynamics, and consumer patterns. So it’s very dangerous to bundle all these countries together. And we’re seeing local startups run into that same issue. There’s a great report by Beanstalk AgTech on local startups. Say there’s a startup in Thailand, they do well in Thailand, they’re gaining traction, and because VCs back them, they feel the need to expand into other regions, get that TAM. But what you do in Thailand you can’t replicate one-to-one in Vietnam or the Philippines. You have to build from the top again. So that’s probably less appreciated, even by local VCs. They tend to underestimate the complexity of these local markets.

One country I’m looking at closely is Thailand. They’ve had a national bioeconomy agenda since 20216, and not many people are talking about it. Thailand is very gifted from a natural-resource standpoint. It’s one of the largest sugar producers globally. And as you know, sugar is the feedstock of the bioeconomy, so if you can get the cost of sugar down, that’s already a big part of the COGS you’re addressing.

They also have government incentives to attract international companies to establish in Thailand, especially in biotech and biomanufacturing. They have these income tax exemptions, import duty relief for machinery and R&D equipment, and infrastructure cost write-offs for these designated biotech zones. So it’s not just a general investment promotion regime. It focuses on biomanufacturing in these biotech hubs.

That said, Thailand still has a way to go. One area where Thailand is falling short in bio-manufacturing compared to other countries in APAC is the regulatory side. There’s no novel ingredient that’s been approved there, unlike Singapore and Australia, where the regulation is robust, but they don’t really have cheap feedstock, definitely not Singapore.

Australia does have cheaper feedstock, but it doesn’t have these policies as favorably as Thailand does. So Thailand still has a way to go in terms of figuring out the regulations. They don’t have the playbook for incorporating these novel ingredients, but they definitely have the infrastructure in place to reach that stage.

Rhishi Pethe: You talked about feedstock and sugar, and whether it’s soy or glycerol, there are a bunch of other feedstocks that go into it. You mentioned that sugar prices are really important because they’re an input. So how do the unit economics of a fermentation startup actually work? What’s the incremental value they’re adding to make the unit economics work, if they’re very dependent on the price of feedstock?

Eshan Samaranayake: In terms of how much of the cost of production is influenced by feedstock, again, depending on the product, it can go up to 50%. They must bring down the cost of feedstock, whether that’s by using different feedstocks or by establishing operations near a region where they can easily access sugar.

You need the feedstock to make anything in the bioeconomy, and right now, sugar is the dominant feedstock. I don’t see that changing in the near future. Glycerol, or mycelium substrate, as you mentioned, we’re seeing early traction of companies trying out these different feedstocks to reduce cost, like food waste. Food waste as a feedstock is another big one. But at the end of the day, sugar is still the biggest input because it’s much more stable and not as crazy expensive as some others that are less available.

Rhishi Pethe: When you’re making an investment decision in a bioeconomy company, the feedstock price dictates a lot of their unit economics and profitability. What gives them alpha, or what gives them a competitive advantage compared to someone else?

Eshan Samaranayake: Feedstock is just part of the equation. But the more important end of the equation, for me, is how much of this feedstock you can convert into the end product. That comes down to strain optimization, mostly. If you have a microorganism that’s not very good at converting feedstock into the end product, that’s not very helpful. You use all the feedstock, but you aren’t getting the product you need. That’s the biggest lever startups are pulling.

The second lever is the scale. You can make this at a 2-liter scale, but can you achieve the same yield and productivity at a 10,000-liter scale? The yield and titer numbers at scale determine unit economics, not bench results. So when we evaluate a fermentation company, we spend most of our time on the biology. What is the conversion efficiency? What does the titer look like at relevant volumes, and how much does performance degrade as you scale up? And if they’re looking at using food waste as a feedstock, which is potentially cheaper than what you can get from commodity ingredients, that’s great. But we spend most of our time looking at the biology and how they’re utilizing the feedstock.

A third thing people underweigh, as I mentioned previously, is downstream processing. You can have excellent strain performance and still have poor unit economics if the separation, dewatering, or purification step is inefficient. For some products, downstream costs rival or exceed the fermentation cost itself. That is another source of real differentiation that does not show up in feedstock price comparisons.

Two companies using the same sugar at the same scale pay roughly the same price. What is proprietary is the biology, the strain library, the years of optimization work that gets you to a conversion efficiency your competitor cannot easily replicate.

Rhishi Pethe: There’s a commodity trader or large food company in the US who comes to you and says, “Hey, what should I be doing five years from now, given how the bioeconomy trend is going?”

Eshan Samaranayake: I think commodity traders have a real opportunity to be the feedstock providers for the bioeconomy. There’s an opportunity for them to transition from a commodity trader to a feedstock provider for industrial bioprocesses. And what used to be useless, some byproduct from farming that you couldn’t really sell for anything, all of a sudden, you now have microorganisms that could utilize this. So there’s also an opportunity to make good use of the waste products you previously discarded.

I’m seeing more companies in the bioeconomy trying to establish their bioreactors right next to farms or mills to have closer access to sugar. The thing about sugar and feedstock is that you have to look at the total cost of sugar. Transporting sugar is a big part of that total cost. So if they’re near a company like Cargill, they can reduce their feedstock costs. There is a mycelium-based company that Cargill has invested in, which has established its facilities next to a Cargill facility7, giving it near-term access to feedstock and exclusive agreements to supply the startup, allowing it to lock in that supply chain.

There are opportunities for them to move up the value chain rather than just be a commodity trader.

Cultivated meat’s reckoning

Rhishi Pethe: There have been a lot of cultivated meat startups and unicorns, and a lot of them have struggled to scale for various reasons. Singapore was among the first countries to approve cultivated meat almost five or six years ago. What has that experience taught the ecosystem about how much capital is needed and what kind of capital intensity is required? What has Singapore actually managed to prove, and what has it not proven?

Eshan Samaranayake: The first thing is that more people have more respect for the challenge of scaling any kind of bioprocess. Because people looked at the early numbers, maybe at a one-liter scale, and thought, oh, this will neatly translate into a production-scale bioreactor, but that clearly didn’t happen. They realized it wasn’t as straightforward as they’d thought.

And then you had these unicorn-cultivated-meat startups. They invested in all these big facilities even before they’d proven that their process could actually scale. They raised all this money and then felt the need to invest it. The way they invested was by buying all these heavy-capex facilities even before they’d proven out the fundamentals of their scaling. So that was a big miss on the technology side.

The other big issue is that most of these cultivated meat and seafood companies were targeting chicken and pork, not really high-value products. And if you’re competing with chicken producers, you have to bring your costs down to crazy amounts, which is impossible at the current scale they’re producing8. So the first issue is that they picked the wrong end product to target. Something like a high-value product, bluefin tuna, would have been a much better starting position than chicken. So, targeting the wrong thing at the start, and the physics of scaling up.

When these companies started, they were pitching 100% cultivated meat products. But when you look at what’s in the market right now, it’s mostly blended products, mostly plant-based, and less than 10% cultivated ingredients. It’s just the reality of the economics. There were clearly different expectations about where it would be in 5 or 10 years. It’s just about being more cautious about how hard it is to scale biology.

Singapore is the first country to approve cultivated meat for sale. What Singapore has proven is that you can eat these products. They’ve shown that with competent food safety regulators, you can approve these products in a reasonable time, so it’s not too long, where startups burn all their funding while waiting for approval. So they can do robust safety regulation in a reasonable time and show that these products are safe to eat. So they proved that these products are edible and real, not just a science project.

What Singapore didn’t prove is the commercial scale. There were a bunch of companies that sold the products in one store, but we didn’t really gain traction beyond that. So Singapore didn’t really prove that you can sell these products on a large commercial scale.

And then, maybe the strongest evidence of Singapore’s conviction in cultivated meat comes from their most recent food safety strategy in 2019: a “30 by 30” strategy, which is why they invested heavily in these alternative ingredient technologies. And now they’ve said that this is no longer part of their “30 by 30” strategy9, because it didn’t meet the consumer acceptance and cost targets they had in mind back in 2019. So they clearly shifted their focus. In my view, Singapore is no longer the cultivated meat hub in APAC. I think South Korea is taking that throne.

Most of the cultivated meat stars are now pivoting into cosmetics, which makes sense from a business standpoint, because you’re looking at a few grams and not the tons and kilograms you have to manufacture. Your unit economics are already attractive even at the earlier stages. So that’s much more reasonable.

So it’s just about being open about cultivated meat as a mainstream consumer product, being at least a decade out from what was pitched.

Rhishi Pethe: In a lot of biotech or robotics, proving out your concept and proving out the technology, maybe that’s a VC-funded model. But when you’re ready to scale, either you need distribution, or you need more manufacturing, a lot of times you can go to strategic capital from incumbents, instead of VCs continuing to put money in. Is that a realistic path to scale?

Eshan Samaranayake: I think that is the path. I don’t think it’s a good idea to use equity to buy capex. That’s not a good way to build a startup. But that’s what some of the earlier cultivated meat companies did. They raised a lot of money and then bought stainless steel with that capital, which is not the best use of equity. So you have to look at other sources of funding for that, like strategic capital.

When these cultivated-meat unicorn startups launched, they didn’t have the infrastructure to support them. We now have more CDMOs to help startups scale, but when they started, there was really nothing available, so they had no choice but to buy facilities and build bioreactors themselves. Now the ecosystem is much better because you have more enabling players, and these big manufacturing folks who have the bioreactors ready, and startups can come in and use them. So you have to be smarter about how you scale up, whether it’s partnering with CDMOs or using non-equity-based funding methods to get there.

When investors look at a cultivated meat or seafood company, they want to see how it’s going to get funded, not just in the next two years, but where the funding is coming from until they get to scale and start selling to consumers. Because you can no longer bank on the fact that we’ll raise funding now and then, 18 months later, we’ll raise again, it’s not as simple anymore. So, given this amount of funding, what reasonable milestones can you hit? And if you do hit those milestones, which investors would open up now that you’ve hit them? It’s not going to be those generalist investors. Probably a strategic or another sector-specific investor, and for them, you need to hit those milestones. So we’re seeing that investors are more cautious with capital, which is good, but they’re also more gated in terms of capital. We’ll give you the funding once you hit this milestone, and I’m not just going to take your word for it.

Rhishi Pethe: As a venture fund in 2026, what are some of the things you look at that have become more salient today compared to, say, three or four years prior?

Eshan Samaranayake: Increasingly, we have to identify the company’s acquirers. In 2021, people were at least considering IPOs for these companies. But when you look at the recent exit data for food and ag globally, it’s very clear that it’s almost all M&A, whether it’s a trade sale or a good exit. So we have to look at who will acquire this technology. What is the startup solving right now that an incumbent is having issues with, and does the startup have any contact with this incumbent? We’d like to see them having these early conversations. It doesn’t need to be a full LOI; let’s do a pilot together. Are they aware of the challenges an incumbent in their space is having? Coffee, for example, if you’re investing in an alternative coffee startup, are they aware of what UCC and the other coffee brands are facing? Because those are the folks who are going to buy your business. So we’re more and more cognizant about where the exits are going to come from.

We’re more and more cognizant about how the unit economics would work today, not once you get to scale. Once we get to scale, we’ll be price-competitive, but how about right now? What are you competitive with? You don’t have to compete on the commodity ingredient, but there could be a low-inclusion, high-value product you can still target and generate revenue today. So that’s why we invest in B2B platform technologies: they offer more optionality.

From a technology problem to an ecosystem problem

Rhishi Pethe: You talked about how you want the unit economics to work at each stage of scale. Let’s pick whatever commodity you want, sugar or soy. What happens to the demand curve of that particular commodity if the bioeconomy hits even a moderate scale? How does the demand curve change for that underlying ingredient or feedstock as these things scale?

Eshan Samaranayake: The example I keep coming back to is coffee. Coffee supply is at real risk of halving by 205010.

Maybe it’s halving, or 30% of the coffee supply is lost by 2050. These are structural issues that won’t be addressed soon, even as bio-manufactured ingredients reach scale and become cost-competitive. They’re going to be the commodity, and conventional coffee is going to be the specialty, where you have it very rarely, or it’s really expensive, because there’s less of it now than two decades ago. So that’s how that balance is going to shift. Of course, you have to get to scale, and you need all this capex to get there.

The roles are going to shift. The bio-manufactured thing is going to be the commodity. The conventional coffee is going to be the premium one, just because there’s less of it. Supply and demand, it’s going to be more expensive because of it.

But increasingly, we’re seeing more blended offerings. Coffee is another good example: one of our portfolio companies, Prefer11, is working with coffee companies to blend their coffee. So 50% Prefer’s dry coffee powder versus 50% conventional coffee powder, and apparently, consumers can’t tell the difference between real coffee and this blended product. So that helps address some of the supply chain volatility for the coffee companies while satisfying consumers. The blended route is another product line that will gain traction in the coming decades. It’s going to be a blend. It’s not going to be black and white.

Rhishi Pethe: When you start writing, you go in with a set of assumptions or ideas of what this is going to be and what you’re going to learn from it. Are there certain beliefs or ideas you had when you started that you feel are no longer true? What have you learned through that process?

Eshan Samaranayake: I started writing in 2023 as a learning experiment, and it still is. So I’m not saying I’ve reached a certain destination. I think it’s always going to be a process.

When I started, I was more naive. Because I come from a biotech background, I was maybe overly bullish about the role technology plays in a product’s success. So, my underlying belief when I started was that great technology and product-market fit are sufficient for a product’s success. What I think now is that, even more than the technology, does the value chain surrounding your target customer have the economic incentive to accommodate your technology? I know I said a lot of words there, but the point is, your technology in isolation doesn’t matter. The ecosystem you’re selling into also has to support your product for it to be adopted at scale. And if you don’t have that ecosystem supporting you, or that ecosystem isn’t ready, then a great product alone is not going to cut it. And if you’re trying to build the ecosystem while you’re selling the product, it requires an entirely different capital structure that VC won’t support. It’s just going to take a lot longer than you think.

In 2023, three years ago, when I started, I thought the sector had a technology problem. But in 2026, I don’t think we have a technology problem. I think we have an ecosystem problem. The science is working. What’s broken is this coordination architecture that helps products reach scale.

Preethy and I discussed the concept of community gardens which connect women farmers with consumers in the Bangalore area in India through a collective, how elephants are a big threat to the farms in certain areas, how women farmers are very different than male farmers, how the next generation does not want to farm, and how Preethy and team are working on climate resilience, while providing high quality vegetables, and greens to their customers.

Get more from Rhishi Pethe in the Substack app

Available for iOS and Android

Get the app

Substack Live with Mike Lee, Food Futurist, today at 10 AM Pacific

What the hell is a Food Futurist???

This was my first reaction when I stumbled upon Mike Lee’s writings. I especially loved his essays on how food systems are both fragile and anti-fragile by design. When I connected 1-on-1 with Mike Lee, I realized we had been thinking about similar system design problems in food and agriculture for some time now.

So we decided to discuss these topics live on Substack Live. Neither Mike nor I have done a Substack Live before, so it should be a fun experience.

Mike is the founder of The Future Market, former head of Innovation & New Ventures at Chobani, and author of Mise: On the Future of Food. He grew up inside a multi-generational restaurant family in Detroit and has spent two decades watching the gap between the pitch deck and the dinner table.

We will explore the following questions in this conversation.

• Software’s real footprint in food. The gap between the pitch deck and the dinner table. Which “concept cars” actually reached the production line, and where the disruption was theater.

• How food companies innovate and whether the signal reaches the farm.

• MAHA & GLP-1s. A policy-and-culture movement and a class of drugs are both reshaping what America eats. It is rippling all the way upstream to what gets grown.

• The supply chain underneath it. Can a system built to push volume out even sense and act on a pull toward quality?

I hope many of you will join us today.

Substack Live with Preethy Iyer, Advocate of Climate Resilient Agriculture, and Women Farmers in India on June 29, 2026, at 8 AM Pacific

I am spending a lot of time thinking about the challenges facing Indian farmers and the role of technology through my AI advisory work with the Gates Foundation in India and Africa.

The first time I met Preethy was when we appeared on Venky Ramachandran’s podcast. Preethy has a very interesting story about how she went from being in tech to helping women farmers. I was really surprised by her observations about the differences between women and men farmers in India. (Men, we don’t show up very well, which is not very surprising!)

So I decided to have her 1x1 on another Substack Live, scheduled for 8 AM Pacific on June 29, 2026. It will be 8:30 PM India time on June 29, 2026, which is supposedly the best time for folks in India.

We will explore the following topics in this conversation.

• What is a Kai Thota, and why did she choose this framing for her work?

• What is the big gap in understanding about farming between AgTech and farmers, and what are its implications?

• How and why do women farm differently from men, and why did Preethy decide to do collective farming?

• Why is entry into and exit out of farming in India so difficult, and what are its implications?

• Technology adoption, the role of AI, and what it misses.

• What is the role of Digital Public Infrastructure, and what does she think the future of Indian agriculture will be?

I hope many of you will join us on Monday.

Subscribe now

Two SFTW Convos coming up

An SFTW Convo is a slightly different format from an SFTW Live. It is presented in written form and is a lightly edited version of a conversation I had with an agrifoodtech leader. I typically enrich the transcript with additional links and context.

SFTW Convo with Eshan Samaranayake of Better Bite Ventures

Eshan is a VC at Better Bite Ventures in Hong Kong. He writes the Better Bioeconomy newsletter and has a very good handle on what’s happening in APAC.

Eshan Samaranayake came to food and agriculture by an unlikely route. He trained as a molecular biologist in Hong Kong and was drawn to the commercial side of biotech. After cutting his teeth at a food-biotech startup, he began writing a newsletter that helped him enter the venture capital industry. Today, he invests across Asia-Pacific in upstream technologies, including ingredients, fermentation, and ag inputs.

This SFTW Convo will be published on July 1, 2026. You can read previous SFTW Convos here.

SFTW Convo with Sarah Garland of Triple Helix

Sarah Garland has spent her career working on that gap. A plant scientist with a PhD in CRISPR gene editing from Cambridge and postdoctoral work at Columbia’s Earth Institute, she founded Triple Helix Institute for Agriculture, Climate, and Society. Triple Helix is a nonprofit designed to serve as the connective tissue among researchers, entrepreneurs, investors, policymakers, and the public. Her premise is that much of what holds agricultural technology back is coordination, capital, and trust.

I spoke with Sarah shortly after Triple Helix launched its newest experiment, an Investment Council that brings together eight investors from across the capital stack. Our conversation ranged across the origins of the organization, the economics of farmer adoption, the design of the council itself, and how a self-consciously nonpartisan group navigates a politically charged moment in food and agriculture.

This SFTW Convo will be published on July 15, 2026. You can read previous SFTW Convos here.

I am excited to restart the SFTW Convo series with the first conversation of 2026.

SVG Ventures | THRIVE is a Venture & Innovation Platform advancing the Food and Agriculture industries.

They work with leading corporations, startups, universities, and growers to solve the biggest challenges facing the food and agriculture sector.

SVG|THRIVE has worked with more than 15,000 startups across more than 100 countries.

Just before World Agritech 2026, I had the opportunity to sit down with John Hartnett, CEO and Founder of SVG|THRIVE, and Marion Meyer, Managing Director of Corporate Innovation, to discuss the state of innovation in the agrifood ecosystem.

Our conversation touched on the history and philosophy of SVG|THRIVE, the nature of innovation, how corporates can access outside innovation by working with startups, and the challenges and opportunities in that process, the big trends in agrifood as seen by SVG|THRIVE how the ecosystem is thinking about the adoption of AI, how corporates are ready (or not ready) for AI, and how SVG|THRIVE is reformulating its approach with a change in the pace of innovation.

We talked about whether agrifood is like a Bermuda Triangle, how an ecosystem model is like Tinder, and how international understanding and connections are critical for SVG|THRIVE.

I hope you enjoy this conversation as much as I did having it.

History of SVG Ventures|THRIVE

Rhishi Pethe: John, you founded SVG Ventures. What are some of the core principles that you follow? Initially, you said, “I want to connect Silicon Valley to Salinas Valley”. It’s 2026. Could you speak to your mission and how things are going?

John Hartnett: I’ve been in Silicon Valley for almost 30 years. I’ve had a front-row seat to companies like Google walking in as startups and watching them scale. I spent a decade building a smartphone company and interacting with the network here, the VCs, the corporates, the startups. Much of what became THRIVE came from that experience and what I see as the recipe for success.

When we founded THRIVE, we didn’t want to build a traditional fund that simply invests and takes board seats. We recognized that successful companies need more support in scaling, access to customers, capital, international markets, and the right talent and experience. Our goal was to build a platform that could deliver all of that.

Of course, the tip of the spear is investment. That’s our main pillar. Ultimately, everything we do leads to investment outcomes, and today we’re investing through two funds: our Pioneer Fund and our Sunrise Fund.

Our second pillar is venture development. We help startups grow at scale through programs that meet them at different stages, from our venture studio at the earliest phase to our academy and through to our flagship accelerator. We work with startups from idea to Series A. Entrepreneurs learn best from other entrepreneurs. It’s not about lectures or theory. It’s about lived experience and navigating the journey firsthand. Our venture development platform is built around that principle.

The third pillar is the ecosystem and network. It’s connecting key people who will be influential in the scaling and success of companies, including board members, customers, investors, growers, and farmers. We’ve found this to be particularly critical in the food supply chain. It’s one of the oldest industries in the world, yet it remains highly fragmented and difficult to navigate.

One of the things we joked about when we were looking at how to connect the dots between Salinas and Silicon Valley was that entrepreneurs and farmers aren’t hanging out in San Francisco, drinking cappuccinos together. They operate in very different worlds, so the question becomes, how do you connect them? In the Bay Area, the power of networks is evident: look at the founders of Google, the connections at Kleiner Perkins with John Doerr, the connections with Bill Campbell and Eric Schmidt, and the relationships among all the players there. People call it an ecosystem; you can call it a network. We use both. It’s about connecting key people who will be influential in the scaling and success of companies. That’s a network in action. We wanted to create one for food and ag.

Our fourth pillar is engaging with corporates. In food and ag, significant exits tend to be trade sales, rather than IPOs, making corporates crucial, whether as customers or strategic investors. The most forward-thinking corporates recognize that innovation doesn’t just happen within their own walls. They actively look outside for disruption and new ideas. That includes start-ups, as well as industry peer connections within or beyond their own sector.

Those four pillars are what hold up our platform.

I live in Silicon Valley, an hour from fields, farmers, and some of the biggest players in specialty crops, but we still weren’t connected. And at the time, the food industry continues to face challenges, whether known or not, related to water, labor, climate adaptation, and climate change.

So I spent time early on with Bruce Taylor, chairman and founder of Taylor Farms, and Dennis Donohue, then mayor of Salinas, and we looked at how to bridge the gap. And we really took the approach of calling this where we are, like the Golden Triangle: Silicon Valley technology and investment to the west; the capital of fresh food and wine to the south and north; UC Davis, one of the world’s top agricultural universities, to the east. That’s where we wanted to bring it all together. And the formula we developed to address it is those four key areas: startups, ecosystem, corporations, and investment.

Ecosystem dynamics

Rhishi Pethe: You mentioned the ecosystem, innovation, how startups learn from others, and how entrepreneurs learn from other entrepreneurs. You mentioned the golden triangle. Right now, AgTech feels more like the Bermuda Triangle than the Golden Triangle.

Marion, you worked at corporates. You are physically thousands of miles away from Silicon Valley in Germany. When you think about ecosystems and innovation, could you characterize what you consider good innovation? What are some of the friction points that you see between startups and corporates?

Marion Meyer: Good innovation always starts with the same questions: What are you trying to achieve with innovation, and what are you solving for your customers? What does success look like?

Corporates have ongoing operations to sustain, but they also need to drive into the future. Those who have managed both areas have lasted for decades. But the speed of innovation has changed dramatically. What took 20 to 30 years in the last century now happens in three to five years or faster. So one of the most important things a corporation can do is ask: what are the topics that – even if they seem far away from core – could have an influential impact on our operations and value proposition? Where is the threat, and where is the opportunity?

Historically, a lot of innovation happened inside corporations, inventing new products and processes by their own teams –and that was sufficient. Today, a lot of great innovation is also coming from entrepreneurs building something new outside established companies and working across multiple technologies, which often aren’t part of the corporation’s DNA. The question becomes: how do we identify what is strategically relevant, gain access to it, and engage in ways that meaningfully impact the business?

For startups, the challenge is similar. They might have a great idea, passion, and the right moat to develop it and bring it to the market – but one can very rarely solve all challenges and access all growth steps alone. They could benefit enormously from other players, such as corporates, for deep market knowledge, existing infrastructure, and decades of customer access. But they often don’t know which door to knock on, or who to talk to.

Getting your own agenda clear, as a startup, a corporate, or an investor, is a prerequisite. And having someone who can enable and facilitate the collaboration, who understands both perspectives and can spot where the fit is, is invaluable.

Let me give an example. In one of my former companies, we experienced great excitement in the team about innovation and new technology. But often, innovators landed at various points of contact within the company, and employers were interested but had no real mandate (nor the time or process) to act on them. They’d do nothing, reject engagement, or pass it along inside the company, say “this is interesting,” and nothing would happen. We then launched an open innovation initiative and made ourselves visible and accessible to the outside world. Suddenly, startups had a real door to knock on. It is not just an email address, but someone who would look at it and respond. And we had a funnel and process created to ensure we were seeing, filtering, and reviewing the interesting ones for us. What is crucial for success is identifying the topics and partners you really want to engage with. Finding the path from idea to objective to action is everything.

Corporates, Corporate Venture, and Innovation

Rhishi Pethe: John, when you talked about access, it’s not just access to corporates for startups, it’s access to startups for a big company. Marion, you’re seeing the pace of innovation accelerate. John, how are corporates actually keeping an eye on what’s happening within the ecosystem, which startups are working on what, and what are the trends? Setting up a corporate venture capital fund was one mechanism. But we are seeing a pullback. FMC shut down its corporate venture fund. Other corporate venture funds have not made many investments. Do you see that their current approach is reasonable?

John Hartnett: Venture capital has been one of the first avenues many corporates have explored, often through setting up their own CVC. They gain tremendous insights by doing this, but alternatively, they can invest in independent VC funds.

The CVC Avenue is expensive and risky because corporates aren’t just looking for financial returns. They are seeking strategic value, including deeper market understanding, early visibility into disruptions and emerging trends, key areas to address proactively, and potential acquisition opportunities. They also want access to information, to the hottest startups, and financial upside.

Finding the right startup is like finding a needle in a haystack.

Corporates often don’t have the luxury to dig through that haystack. That’s why they’re leaning into organizations like ours. Over the past decade, we’ve built a network of 15,000 startups across more than 100 countries. That didn’t happen overnight. But we’ve also built an intelligence layer on top of it. We don’t just know who these companies are; we’ve engaged with them. We’ve accelerated over 300 of them. When you evaluate a company, technology is essential, but you’re also looking at the team, the traction, the problem they’re solving. And not every corporate partner wants to invest or acquire. Driscoll’s, for instance, isn’t looking to buy companies. Instead, they want to solve real-world problems around water, irrigation, and labor.

Rhishi Pethe: So you are doing matchmaking between corporates and startups. It’s like Tinder for innovation, I guess?

John Hartnett: If we take the Tinder analogy, before you get married, you date, you get engaged, then you commit. Nobody gets married on day one. Our accelerator programs de-risk the engagement cycle. We’re working with these startups for months, sometimes years. You get to “date” the people and the companies for a time before committing to anything.

Rhishi Pethe: You guys mentioned that you help corporate executives identify which startups to engage with, depending on their needs, thanks to your strong network. If you flip the question, not every startup can work with every corporate client. What is the process for saying, “You should work with these one or two corporates because they’re a better fit for the startup”? How does the reverse matchmaking work?

Marion Meyer: The better we know a startup, the better we can judge whether there’s a genuine fit with a particular corporate, because they are solving or supporting exactly what this corporate needs or is looking for.

We start by getting to know the startup deeply, what they’re doing, and how it fits a real market need on the customer’s or corporate side. We have different programs depending on the stage of the start-up, which allow us to do this, especially our accelerator, challenges, and immersion programs. We also recently launched a Scale-Up Assessment. It is a structured deep dive that examines a startup’s setup, organization, solution, and go-to-market strategy and identifies gaps, risks, and opportunities for the next phase of growth. And enablers and potential partners for commercialization and growth, such as Corporates.

On the other side, we spend a significant amount of time with each corporate partner to understand their strategic priorities and innovation needs. We listen carefully but also offer an outside-in perspective. For corporates who are still early in their innovation journey or want to review their existing activities, we offer an innovation assessment to help build the process, set priorities, and sharpen scouting areas.

For more advanced organizations, we execute through structured briefings and ongoing reviews to ensure continuity across innovation pathways while allowing for flexibility and adaptation.

Sometimes a fit between start-up and Corporate might be seen via a pre-identified scouting area; sometimes a fit appears as an unforeseen solution, for instance, when it hasn’t been on the radar of a particular technology search. Still, it provides a desired solution from an unexpected track.

Trendspotting

Rhishi Pethe: John, you have worked with 15,000 startups over the last 10-15 years and have talked with many corporates. Your team will have a front-row seat to current trends and what these amazing entrepreneurs and corporates are thinking.

What are some of the two or three trends that everybody is always curious about? Here in San Francisco, if you don’t mention AI, people think you are a psychopath.

Could you speak to the broader macro trends in the sector and then explain how AI fits into them?

John Hartnett: Technology follows big problems. That’s why we remain passionate and excited about food, agriculture, and climate. The big problems out there today will unlock the big technology solutions.

On one side, you have resource scarcity, arable land, water, energy, and labor, which significantly affect production. At the same time, we’re seeing the cumulative effects of fires, droughts, and extreme weather on production. We’ve also seen our supply chains be very fragile. We went through COVID, showing us how quickly empty shelves can appear. We’ve seen the effects of geopolitical conflicts, which have increased food costs significantly in many areas. On the other side, health and nutrition have become central to how people think about what they eat. Yet the US still has some of the highest obesity rates in the world. These are the big challenges to tackle.

For us, breaking it down into actionable innovation themes means looking at it from a deep-tech perspective.

The first major trend we’re watching is the intersection of biology and AI. Companies like Heritable, a spin-out from Google, are working on programming plants to address resource scarcity, reduce climate effects, and deliver more nutritious food. That deep-tech intersection feels like one of the biggest opportunities of our lifetime.

The second big thing is physical AI. AI applied to robotics and automation. Physical AI will unlock solutions to real-world agricultural challenges much faster than many expect. And it will attract major tech players to this space in ways we haven’t seen yet. Software eats the world, right? But when that software evolves beyond digital confines and enters the physical world, it will be eating faster.

Then, I think this is always up for debate, and I’m still in the camp that CEA farming is going to get to the other side of the bridge here. I know the timing hasn’t been right. The economics of running large-scale facilities haven’t been penciled out yet. But I keep coming back to smartphones. Everyone thinks that the story started in 2007 with the iPhone. It started a decade earlier. The timing and technology just weren’t aligned. You needed compute power that fit in your hand, and networks fast enough to support it. None of that happened overnight. Apply that same lens to where we are with agrifood and AI, and I think the timing is now to support it.

Big tech’s engagement in AI applied to life sciences, agrifood, and climate has been tentative so far. Google’s Mineral was a compelling experiment, but it didn’t represent a full commitment to the sector. Microsoft has dipped a toe in, but none of the major AI players, such as NVIDIA and AMD, have truly engaged. I think that changes with physical AI, and when it does, you’ll see these sectors light up in a significant way in a two-to-five-year window. The more big tech gets into this sector, the bigger the opportunity, the faster it’s going to happen.

There’s been a lot of hype in AgriFoodTech, but that hasn’t always translated into strong outcomes. We saw a similar dynamic during the early internet era: from 1998 to 2001, all you needed was “dotcom” in your name for valuations to skyrocket. But out of that came Google and Yahoo. I joke a little bit that after COVID hit in 2022, investors acted like drunken sailors, investing from home because they thought agrifood was a hot area, but they didn’t fully understand it. Yet I do believe that out of the current hype cycle will come some of the biggest companies in agrifood. And those companies that come out the other side will be the ones built to last.

Corporate Innovation

Rhishi Pethe: It’s a good parallel you drew to the internet or the smartphone. We’ll see much more movement and development in this area over the next 2-5 years, especially in AI. If I’m a corporate, I need to be ready to take advantage of it. Marion, when you talk to different corporates, how much of the amazing innovation coming through AI are they able to take advantage of? And how ready do you think they are when some of the things that John is talking about will start to pick up the pace even more? And if they’re not ready, what are the gaps? How do you help them figure that out?

Marion Meyer: Every corporate I speak to is experimenting, and that’s exactly right. But it’s important to understand that “AI” isn’t one single topic for a corporation. It operates across multiple layers, and you have to manage them deliberately.

The first layer is AI for organization and process, tools that apply broadly, regardless of sector, to how you work internally. This is where adoption is the hardest challenge. AI isn’t a typical system rollout. You don’t implement a tool and tick the box. You have to change behavior, culture, and processes, and the way people think about and use these tools. The journey has started, but it’s far from over.

The second layer is AI applied to differentiating processes. Your R&D, your time-to-market, your cost of production. These have real competitive implications. And the third layer is AI in the products and services you build for customers.

There’s still a meaningful opportunity, particularly in internal R&D and operational functions that aren’t traditionally IT-focused. Most organizations are still in the early stages of mapping out what AI can and should actually do for those teams. Two practical things can help. First, set priorities rather than trying to do everything at once. Identify the core layers, make tool choices, and bring your teams along properly. Second, be deliberate about data. Know what’s proprietary, what you’re willing to share, and in what context. If you try to protect everything, you limit your ability to use the most powerful tools. If you carelessly give away your proprietary data, you may put an asset at risk or create a dependency. Getting that clarity is foundational.

We support Corporates and Start-ups also in the arena of AI-driven innovation. One element is curated networking for our partners – in dedicated roundtables, we bring industry players together to exchange insights on initiatives, challenges, and best practices, as well as perspectives on the most impactful developments. We add external speakers to provide insights and inspiration, and we bring the tech industry into that conversation. In 2025, we initiated an AI & Automation roundtable, and we will have the next one alongside the Global Impact Summit 2026. In addition, we are launching complementary AI-focused innovation services. With those services, we can help companies find and understand the right starting point, put curiosity to use in use cases, and drive value.

EU vs. the US

Rhishi Pethe: Because you’re from Europe, I have to ask you this question. Sitting here in Silicon Valley, some people look down on Europe for its regulation and how it deals with innovation.

Does that question come up in your conversations with corporates or startups? You’re in Germany, you shut down a bunch of nuclear plants, and people criticize it over here. What can the US, or US ag tech or agri-food tech, learn from the European approach to innovation? What are your inputs on that when it comes to SVG|THRIVE?

Marion Meyer: First of all, many ideas, challenges, and approaches around innovation we talked about are shared within corporates and start-ups, independent from the individual markets and regions they are operating in. Most companies already operate globally or plan to do so. That’s why support is not limited to regional experience but could also benefit from additional lenses and insights into other markets.

We do see specific regulation and sustainability requirements in Europe. They do slow down innovation in some respects. There’s no doubt about that. But they can also be a trigger for it. When a regulation effectively says “you can’t do this anymore,” it creates real pressure to find an alternative. Business finds a way, like water finding its path. Frankly, that kind of pressure often drives more durable innovation than subsidies do.

The second thing Europe offers is deep industrial and engineering competence. When we talk about physical AI and automation, Europe and Germany in particular have decades of robotics and manufacturing expertise embedded in both corporations and universities. When that hardware and process knowledge combine with software capabilities, much of which comes from the US, the results can be very powerful. We’re already seeing that with unmanned aviation and geospatial data. In the Munich ecosystem, for example, we are seeing a wave of drone and related technology startups driven by dual-use applications, and the underlying knowledge transfer may directly benefit agriculture.

Third, Europe’s diversity can actually be seen as an asset. Europe is facing pricing challenges as dozens of distinct markets operate within a shared economic framework. A company that learns to scale across Europe has learned something genuinely hard. That capability translates globally.

Silicon Valley is SVG’s headquarters, but it’s a global team, and team members are also deeply embedded in the Canadian market, in Ireland, across Europe. We’re connected to ecosystems around the world, which enables us to understand different markets, bridge regions, and match technology to local use cases.

John Hartnett: Global reach is a non-negotiable for us. Seventy percent of the 15,000 startups we engage with are outside the US. We’ve taken what we call the A7 approach, focusing on the seven most important agrifood markets globally.

Evolution of the SVG Model

Rhishi Pethe: John. You’ve been in this space for more than 10 years. People talk about how one-person teams or two-person teams can scale ARR very quickly. Do some of those principles apply to the sector that you’re working in? If yes, how do you see the SVG|THRIVE model evolving over the next five years?

John Hartnett: The model is already evolving. When we started, we were very focused upstream. As we’ve looked at where technology has the greatest impact across the food system, the frame has expanded. It’s now genuinely farm-to-fork. We’ve broadened our focus over the past few years to reflect that, and now we include downstream, supply chain, and delivery technologies.

We’ve also leaned much more heavily into climate and sustainability. We launched our Global Impact Initiative in 2020, specifically to help corporates deliver on the ambitious 2030 and 2050 sustainability targets they’ve committed to. That’s a growing part of our work.

Another key area is deep tech, given its ability to address challenges across ag, food, life sciences, supply chain, and related areas. It’s largely sector-agnostic. Something that solves a problem in the food supply chain may also solve a problem in construction or another industry. We’re continuing to refine our investment thesis, putting deep tech in focus as a solution for planetary health, human health, and supply chain resilience.

In parallel, our focus is to scale our investment platform. To date, we’ve primarily invested at the Seed and Series A stages, deploying approximately $100 million. We see a significant opportunity to expand this further. The market opportunity clearly supports it.

Rhishi Pethe: Thank you, John and Marion, for your time today and your work for making it easier to bring innovations to market at scale.

One of my favorite breakfast items is avocado on sourdough toast, topped with apricot-jalapeño jelly, salt, and pepper. The beauty of this breakfast is that you can pretty much have it year-round and pair it with a mimosa on the weekend!

The miracle of being able to have avocado toast almost anywhere and anytime in the United States didn’t happen overnight. It is due to a combination of deliberate choices and technology development over the last 30 years.

The avocado is one of the silent logistics triumphs of the last three decades. The fruit that was kept out of the United States until the late 1990s is now a year-round phenomenon, and its future seems brighter. OECD-FAO predicts that avocados will become the most-traded major tropical fruit by 2030, overtaking pineapples and mangoes in quantity.

No visa for Mexican Avocados

The original ban on importing Mexican avocados to the United States dates back to 1914. USDA’s APHIS imposed a quarantine on Mexican avocados after the detection of avocado seed weevil. The pest risk posed by seed weevils, stem weevils, seed moths, and fruit flies was the stated reason to maintain the ban for the next few decades.

An avocado seed weevil is a destructive beetle (family Curculionidae) native to Mexico and Central America that attacks avocado crops. Avocado seed weevils are considered major agricultural pests and are subject to quarantine in the U.S. to prevent them from decimating domestic avocado crops. Because the larvae live deep inside the fruit, they are extremely difficult to control with traditional insecticides.

It is debatable whether the pest-based ban was protectionist toward local California growers. As late as the mid-1990s, Southern California produced 90% of the avocados Americans ate. California growers had every incentive to keep the competition out.

“Bigger than the Alamo”

The NAFTA accord brought a massive step change in the availability and U.S. domestic consumption of avocados. The U.S. wanted to send corn and other agricultural products to Mexico, and Mexico demanded a return favor, and avocados were part of that agreement.

NAFTA opened the door to Mexican avocado imports, though initially they were allowed only in 13 states and for only four months of the year.

Mexico put pest worries to rest by saying it would export the thick-skinned Michoacán Hass, instead of the pest-vulnerable thin-skinned variety. It is not surprising that California excluded itself from the list of states importing avocados.

The state’s avocado industry had opposed the APHIS import rules, and the California Avocado Commission challenged the 1997 and 2001 rules in court, arguing that APHIS violated the National Environmental Policy Act by failing to prepare an Environmental Impact Statement (For all you housing and energy YIMBYs and NIMBYs, this might all seem quite familiar!)

Almost 10 years after NAFTA opened up the American market, most U.S. states (including California) allowed the import of Hass avocados.

In the early 2000s, Congress passed the Hass Avocado Promotion, Research, and Information Act and created the Hass Avocado Board with the explicit purpose to “increase the consumption of Hass Avocados in the United States,” as the domestic supply couldn’t meet demand.

The Mexican exporters and the US importers came together and formed the Avocados from Mexico (AFM) marketing subsidiary to promote the sale and consumption of avocados in the United States. But before the Avocados from Mexico (AFM) subsidiary could convince customers, it had to convince grocery retailers to carry the product year-round.

The Super Bowl was chosen as an ideal time to promote avocado consumption. Super Bowl Sunday is the third-largest party day in the United States, after New Year’s and Halloween. The Super Bowl was also chosen because it conveniently coincided with Mexico’s production season and its peak supply window. The state of Michoacán in Mexico produces almost 80% of Mexico’s avocados.

When the US lifted the ban on importing avocados to California in 2007, the chairman of the California Avocado Commission said,

Mexican avocados into California is like they won the Alamo1. It is bigger than you and I can ever imagine.

In 2026, 300 million lbs of avocados were imported 4 weeks before the 2026 Super Bowl (20% above the historical average), with 20% of annual sales occurring around Super Bowl Weekend. (300 million lbs of avocados is about 7,000 semi-truck loads!!)

Cinco de Mayo2 is another manufactured celebration in the United States that also spikes avocado sales and consumption, as Americans eat copious amounts of guac, chips, and cerveza! Avocados first appeared on the U.S. food pyramid as early as 1992, and they have remained there even in the latest 2026 food pyramid.

Subscribe now

Avocados are the John Stockton of grocery stores

For non-basketball fans, John Stockton is the career assist leader with almost 30% more assists than the second player on the list.

Avocados have a halo effect in the grocery store. Avocado consumption increased from about 1 pound per person per year to almost 9 pounds per person per year in 30 years. But another interesting thing happens in grocery stores, where smart merchandising boosts sales of other commodities when avocados are present.

According to a report from the Super Market Perimeter,

Produce displays that put complementary items close to each other encourage cross purchases, and classic pairings of contrasting colors, such as avocados and tomatoes, help draw the shopper’s attention.

Based on the study commissioned by Avocados from Mexico,

AFM-branded displays placed throughout the store helped boost sales for the meat department by 19%, the packaged bakery by 5%, and the deli by 4%.

The following video shows the journey of an avocado from the field to the chip (assuming you scoop up your guac with a chip!)

Multi-origin and counter-seasonal sourcing

Let us now try to answer the question of how avocados are available year-round, even as avocado consumption has increased 9-fold in the United States over the last 30 years.

Over the last 15 years, worldwide avocado production has nearly doubled, with Mexico, Peru, and Colombia leading the increase. Michoacán produces over 80% of Mexico’s avocados, with an annual export value of $2.4 billion. The value of Mexico’s avocado industry roughly quadrupled over the last decade.

Just like any other fresh perishable product, demand has to match supply over some time horizon. Just as we saw with blueberries, the year-long availability of avocados worldwide is made possible through overlapping avocado production seasons combined with some really cool logistics and supply chain technology.

Counter-seasonal sourcing fills supply gaps when the primary supply region slows or when supply is unavailable. It helps smooth out supply to match year-round demand. As a consumer, you might not even realize that the supply source for your avocados has changed if you are not paying attention to the little stickers on your avocados.

Avocados from Mexico can provide products almost year-round in the US, with Peru and California filling any supply gaps. Peru’s production has roughly doubled over the last few years.

The counter-seasonal calendar in the United States is anchored around Mexico, which meets 80% of the US demand. California’s peak window is in spring and summer, meeting seasonal domestic demand. Peru meets the summer demand, though it ships to Europe first, and can now satisfy the excess summer demand. Chile fills the gaps during August through December, while locations like Colombia are Europe-leaning and seeing rising supply.

Multi-origin sourcing provides a smooth supply to different parts of the world. Still, for a perishable product like an avocado, this is possible thanks to innovative logistics and technology.

Subscribe now

Sleeping on a long voyage

Just like a perishable item, avocados are climacteric3.

Avocados are hard and green when picked. You could hurl a recently harvested avocado at someone and do some serious damage. But avocados ripen on a clock. The best part is that this clock can be stopped and restarted with the right temperature and the right type of gas.

Avocados naturally produce ethylene gas, which ripens the avocado.

To keep avocados “asleep” (preventing them from ripening) during long-distance transportation, the industry uses Controlled Atmosphere (CA) in refrigerated shipping containers. The shipping container stays at 5 to 7 degrees Celsius (41 to 45 degrees Fahrenheit). The CA containers lower oxygen levels and increase carbon dioxide levels through specialized gas management.

To further manage the effects of ethylene, ethylene inhibitors, such as 1-methylcyclopropene (1–MCP)4, are used in climacteric fruits like avocados to delay ripening and irreversibly bind to ethylene receptors on cell membranes. It extends the post-harvest life of avocados, slows softening, reduces color change, and reduces susceptibility to physiological disorders. The recommended MCP application protocol can extend the avocado’s shelf life to between one and two months.

These interventions ensure that when the avocado reaches the destination market, it is in the same hard, green condition as when harvested. The temperature and the MCP intervention help decouple the supply from the demand. Grocery retailers can then release them just-in-time to match the avocado demand at the consumer level.

But how do you match the ripeness of a sleeping avocado to the consumer expectations, depending on when the consumer demand is there?

Just-in-time avocados

Matching demand with the right avocado ripeness for consumers is achieved through ripening. You thought that auto and other industrial manufacturing industries are the heavy users of the just-in-time manufacturing philosophy. Well, the avocado supply chain also uses just-in-time principles.

Our big container of avocados has made it very close to its market. Those avocados came from Mexico, Peru, or one of the producing regions. The avocados are still asleep and not really ripening. The avocado supplier or the grocery retailer now needs to wake up the avocados and control the ripening process.

Timing the ripening process perfectly is important so that when you go to your grocery store, the avocados will be ready to eat in a few days. The supply of avocados has to match the demand for avocados. You ship too few ripening avocados, and you leave money on the table. You send more avocados than there is demand for, and you have to throw them in the trash, as nothing is uglier than an overripe avocado.

This process occurs in large ripening rooms. Bananas use a similar process, though avocados are a bit more complicated. As I have experienced during my many trips to banana ripening rooms, the process is both an art and a science.

Ripening fruit is a dying art, typically done by men who have been doing it for 30-40 years. The newer ripening rooms use a combination of sensors, IoT devices, and sophisticated AI algorithms to control the ripening process very carefully. The ripening process is matched with demand forecasting to ripen the right quantity and quality of avocados at the right time.

Controlled ripening ensures that the avocados you get at the store provide the best possible experience for your avocado toast, guac, or whatever else you use them for. (Smoothies, avocado ice cream???)

Subscribe now

Will the Avocado Rush continue?

The big question is whether we will continue to get an uninterrupted supply of avocados throughout the year. Will we be able to enjoy our avocado toast and guac in the future?

NAFTA/USMCA and demand generation around the Super Bowl kicked off the avocado rush. Growers, vertically integrated suppliers such as Mission Produce and Calavo Growers (which merged as of May 2026), the supply chain, and technology responded to meet that demand.

The tariff wars of the last few years make the avocado rush question relevant, though Mexican avocados have largely escaped tariffs. As long as avocados meet USMCA’s rules of origin, they stay on the 0% tariff. Suppliers from Peru, Chile, and Colombia got hit with the 10% tariff, making them less competitive vis-à-vis Mexico for the US market5.

The tariff is not the near-term risk, but the whole USMCA structure itself is. The USMCA replaced NAFTA on July 1, 2020, but it included a sunset clause. (article 34.7) The three countries have to hold a mandatory review on the sixth anniversary (July 1, 2026 - just two weeks from now).

At the review, they can agree to extend the USMCA for another 16 years, or, failing agreement, fall back to annual reviews, with the USMCA slated for termination on July 1, 2036. Any party can walk away with a six-month notice during this period.

The non-existence of USMCA is one of the biggest risks (low probability) to your avocado toast.

There is enough diversification of supply across countries to cushion short- and medium-term hiccups by drawing on supply from elsewhere. A major supply shock to Mexico could have a larger impact, given that it accounts for 80% of the US supply. The sleeping and ripening technology helps cushion demand and supply shocks of a certain magnitude.

The drug cartels of Mexico have gotten involved in the avocado business, as it is “green gold”. Cartels seem to be charging growers a protection tax of a certain number of cents per unit weight of avocados. A further deterioration in the cartel situation in Mexico could adversely affect avocado supply. During a 2024 episode, assailants assaulted USDA inspectors at a Michoacán roadblock. It paused imports for a week, and carton prices shot up by 40%. A major escalation could pause imports for a longer period.

As a safety measure, in 2024, the USDA transferred the pest inspection of Mexican avocado orchards to the Mexican government, a move that raised significant concerns in the United States. The biological risks posed by pests can escalate quickly6 (as we are seeing with the screwworm episode right now), especially given the limited number of avocado varieties.

Remember that the original pest risk kept Mexico’s avocados out of US markets for almost 83 years.

On the surface, it might seem that avocados carry a similar monoculture exposure risk to bananas, given the dominance of Hass avocados, much like Cavendish bananas7.

Avocados have far better monoculture insurance, though. Avocados have seeds, so the plants that emerge during breeding exhibit genetic variability. Breeders can breed a resistant successor to a Hass avocado, but it is very hard to do with bananas.

One of the biggest long-term risks to avocado supply from Mexico is water availability, as avocados require substantial water to grow. Access to water is not a risk only for avocados but for crops in general. A peer-reviewed Michoacán study8 found that avocado farms can consume up to 120% of a municipality’s water concession.

Mexico will have to figure out a way to manage its water.

Trade agreements, demand for avocado products, global supply expansion, and technological sophistication will likely help protect our avocado toast breakfast in the future!

The biggest risk could be just be you and me, who cannot manage avocado ripeness at home, as was very delightfully described in this comedy bit by Josh Johnson!

I want to thank , , , and for their feedback on this piece. Any errors are purely my own.

Share

The Bollywood movie “Lagaan” (Land Tax)1, set in the 1890s, tells the story of how the population of a small farming village in India is driven into bankruptcy and ruin when rainfall fails, and a drought sets in. The local colonial British tax authorities do not provide any relief. The villagers either pay the tax or find another way out.

Some things about this situation have not changed in 2026.

Even though the Britishers are long gone, and the government does not impose direct punishing taxes on Indian farmers, millions of farmers in India are at the mercy of rainfall. They are a season or two away from being pushed into extremely challenging economic situations.

Let us take the example of a farmer in a village in India. He farms seven-tenths of a hectare (the average farm size in India, smaller than a soccer field). The monsoon (rainy season) has been irregular, and he has had to make tough decisions about which crop to plant as input prices, such as fertilizer, have shot up over the last few years.

He has inherited this piece of land, which was twice the size, when his grandfather farmed. This farmer’s average monthly income is about ₹5,300.

That is roughly fifty-five dollars.

He can barely feed his family and relies on off-farm work to augment his income, which is often extremely difficult to find. He is one of nearly 700 million people in India who live off the land, and his life is the statistical center of Indian agriculture.

There is a word for this type of farming. It is called “subsistence” farming.

It is a strange way to define a profession.

We don’t say “subsistence plumbing,” or “subsistence manufacturing,” or “subsistence consulting.”

The use of the word “subsistence” is a quiet admission that hundreds of millions of people in South Asia and Africa produce food that is barely enough to stay alive to do it again tomorrow. And unfortunately, they are trapped in this situation2.

And AI alone will not help extract them from the situation.

The idea of putting an AI agronomist and an AI advisor in the pocket of every smallholder, which helps with on-farm management decisions and access to important information about resources and markets, is one of the most promising developments in the sector.

The AI-powered capability is real, and some of the early gains seen by organizations like Digital Green, KissanAI, and others are real. For example, even with the video intervention from Digital Green3,

Researchers found that the video increased farmers’ agricultural productivity and profits. After one year, farmers who were offered the videos increased yields by 12-18 percent and estimated profits by 9-24 percent, relative to farmers who received conventional extension, with smaller effects after two years.

An AI-powered agronomist (when farmers have very limited access to this kind of expertise in a place like India) can help lift farmers’ income, reduce their risks, and keep their children in school longer, at one-tenth the cost of previous video-based interventions.

Unfortunately, it does not solve some of the other structural problems that plague Indian agriculture.

And given India’s agricultural workforce size, it is a huge problem!

The scale of the problem

In a huge country like India, with a population of 1.5 billion, 43% of the working population is engaged in farming!

As et al said in their 2024 paper4,

Indians make up roughly 17% of the world’s population, living on only 2.5% of the global landmass, representing one of the lowest levels of land area per capita worldwide. Agriculture’s share of GDP has fallen to roughly 16% for FY 2024, yet it employs nearly half the workforce. This suggests profound spatial and economic misallocation. Workers remain trapped in low-productivity rural areas because land markets function poorly. Fragmented holdings, unclear titles, and restrictions on land sales keep agricultural land locked in place.

There are a few countries, such as the United States and the Netherlands, that have pushed the frontier on this metric more than most. In the United States and the Netherlands, about 1-3% of the population is involved in agriculture, and agriculture contributes roughly the same percentage to each country’s GDP.

If you look at these numbers and the tough life of farmers, it feels like we should have fewer people involved with farming in Asia and Africa. The life of a farmer is tough for most of the hundreds of millions of farmers, especially in the Global South. They work long hours, are at the mercy of the market, and with small plots, can barely feed their families.

For example, the reality of most Indian farmers is not romantic. Nearly 400,000 farmers took their own lives in India between 1995 and 2018 (48 farmer suicides per day). The suicide rate among Indian farmers is now 47% higher than the national average.

Crushing indebtedness (not GMOs) is one of the leading causes of stress, stemming from insufficient agricultural investment and reliance on non-institutional credit sources. In India, even though the share of the working population engaged in farming has declined over the last few decades, the mismatch between its contribution to GDP and its share has widened.

Sood (2018) reported that 76% of farmers wanted to give up farming. Agriculture is seen as a non-remunerative livelihood activity with unstable returns due to rising production costs, climate risks, market failures, and a lack of social respectability.

India is an also outlier in terms of average farm size. In almost all countries (except India), less efficient farmers (or farmers who wanted to exit farming) either leased or sold their land to move to urban centers, thereby driving land consolidation.

The average size of operational holdings decreased from 2.28 hectares in 1970-71 to 1.84 hectares in 1980-81, to 1.41 hectares in 1995-96, and to 1.08 hectares in 2015-16. Average landholding size decreased further from 1.08 hectares in 2016-17 to just 0.74 hectares in 2021-22.

Very small landholdings make the application and adoption of any new technology extremely challenging. Any individual farmer has limited financial resources or cash flow to invest in farm improvements.

When you live in the SF Bay Area, you start to believe AI will solve all the problems in the world.

AI is a powerful force slated to transform many sectors in the coming years, but I believe AI alone will not solve India’s farming problems.

What structural change is actually required

AI advisory is very good for individual farmers’ welfare, but struggles to address major structural issues such as land-market lock-in, the subsidy system, and the absence of a non-farm economy.

Locked Land Markets

The existing agricultural land markets in India are locked up. India needs to make land leasing for agriculture much easier and implement land ceiling reforms that limit how much land one can own, with limits varying by state.

For example, in the western state of Gujarat (where I grew up), you must be a farmer or have an agricultural status to buy agricultural land. According to the land ceiling laws, an individual or family cannot own more than 54 acres of dry cropland. The land ceiling for perennially or seasonally irrigated land is from 10 to 27 acres!

Historically, land leasing in India has been heavily restricted and regulated by the state governments. In some states, there are strict prohibitions on leasing out agricultural land.

On the other end of the spectrum, some states do not prohibit leasing but strongly regulate it, and if a tenant works the land for a specified period, they acquire the right to purchase that land from the owner.

Some recent reforms pushed by the federal (central) government’s think tank have encouraged states to legalize and formalize land leasing.

Research has also shown that distortions arising from leasing and land ceiling regulations prevent land from flowing to the highest and most productive users5. The potential impact on agricultural productivity through efficient reallocation of land could be staggering!

An efficient reallocation of land in India increases agricultural productivity by 65 percent and by more than 100 percent in some states, with more than 50% of these effects attributed to state-level rental barriers. Distortions associated with land-market participation contribute substantially to agricultural productivity differences across Indian states.

A distorting subsidy system

India needs to reform the subsidy scheme, which addresses both the input side (fertilizer, power, irrigation, credit, crop insurance) and the output side through market price support. The subsidy schemes, especially on the input side, are a genuine safety net. But it creates significant structural challenges that trap farmers in a cycle of subsistence.

The massive amounts of money spent on the subsidy crowd out funding for other projects like irrigation, R&D, roads, and cold-chain infrastructure. For example, my friend just pointed out in his latest newsletter edition the ridiculousness of the fertilizer subsidy for fiscal year 2027, and these numbers are from before the Strait of Hormuz fiasco!

India’s fertilizer subsidy for FY27 is budgeted at about ₹1.71 lakh crore6, more than the entire ₹1.33 lakh crore the Union Budget set aside for agriculture and farmers’ welfare that same year. The instrument built to support the farm sector now costs more than the ministry built to support it. How the hell did we get into this mess?

The subsidies have environmental damage baked into the incentives. As you just saw, the fertilizer subsidies for urea (nitrogen) are the largest, and the farmers over-apply nitrogen.

India’s N:P:K7 application ratio is roughly 7:2.7:1, compared with an ideal ratio of 4:2:1.

This leads to runoff and soil toxicity, which are harmful to soil health in the long run. The availability of free power to farmers creates an incentive for farmers to pump groundwater, which is costless. It has drained the aquifers in the rice- and wheat-growing regions in the north.

The excessive subsidies create a massive distortion in what gets grown. Price support through minimum-sales-price procurement focuses on rice and wheat. As a result, farmers continue to grow water-intensive crops in water-scarce regions instead of diversifying into pulses and oilseeds, which the country actually has to import.

It traps a large segment of the farming population in ever-shrinking plots, holds them hostage by input and output price volatility, keeps them entrenched in a cereal-heavy, groundwater-depleting status quo, and leaves them riding a subsidy tiger, with no way for the farmer to get off.

And the tiger is getting hungrier by the day.

Missing non-farm economy

The lack of creation of jobs outside farming through off-farm work, manufacturing capacity, and infrastructure investments does not provide exits for farmers who might want to get out.

Archaic land regulation makes both entry and exit difficult. But the reforms that could help are politically more or less frozen, as evidenced by the 20218 farming law battles. Many changes are occurring only at the margins, but many more structural reforms are needed.

The lack of substantive movement on the political front makes AI advisory attractive, as at least there is a clear path to individual farmer welfare with AI.

AI-driven services for knowledge access and dissemination

For farmers to be effective and efficient in growing crops, they often need help with when to plant, how to address specific crop threats, and product selection and timing. This is not a problem unique to any particular part of the world.

Africa and Asia also have systems of agricultural experts provided by their respective states (and private enterprises) to help farmers with their knowledge needs. Unfortunately, the ratio of farmers to agricultural experts in these regions can range from 3,000 to 1 to 10,000 to 1. Effectively, the farmers are farming without any expert help, other than what they can get from their neighbors and friends.

AI agronomists can deliver context-specific information to millions of farmers at a low cost. The cost of delivering these services is sometimes 100 times lower than that of existing non-tech methods. Data infrastructure and phone penetration are quite high in these countries, so there is an existing distribution infrastructure capable of delivering context-aware services to millions of farmers at very low cost.

Suppose you take a representative small farmer with 0.7 hectares, who grows rice and wheat in a semi-irrigated area. The national average monthly farm income is ₹5,300.

The AI advisory helps the farmer optimize timing, quantities, and products, resulting in a 25% increase in yield and a 20% reduction in costs. (World Economic Forum data9)

For a 0.7-hectare plot, with rice prices at ₹2,300/quintal (a quintal is 100 kg, or roughly 220 lbs), this translates into a monthly income increase of roughly ₹1,200-₹2,000.

This farmer goes from ₹5,300/month to perhaps ₹6,500-7,300/month. It will have a material impact on their condition as it might mean better food, children staying in school instead of getting pulled to work on the farm, or reduced borrowing from the informal money-lending economy.

But, they still earn far below the urban average. Without the savings to transition out of farming, they are still just one bad monsoon away from debt.

So, AI advisory has some meaningful impact, without structural change.

1. AI advisory improves individual farmers’ welfare. It does not help with the challenge of too many people on too little land producing poverty-level income on a 0.7-hectare plot. It does not help them escape the structural conditions of Indian agriculture, which include fragmented landholdings, a lack of non-farm jobs, frozen land markets, and inadequate social security.

2. The pathway to a structural change is indirect and generational. If AI-driven yield improvements give a family extra money to invest in a child’s secondary education or vocational training, then the child may be better positioned to exit. It is typically education that enables non-farm transition, not farm income.

3. One of the biggest economic threats to an Indian farmer is catastrophic variance. The reduction in risk from using AI advisory could be even more important than low average yields.

For the reality of Indian agriculture to change, the Indian economy needs to create non-farm jobs, build manufacturing capacity, and expand social security. It needs to reform its land tenancy laws around the land ceiling acts, and entry and exit requirements.

AI advisory services for Indian farmers are a valuable and scalable intervention. But it is important to keep in mind that they are a treatment for symptoms within a system that structurally produces those symptoms.

Technology helps farmers farm better, but only structural change can make farming populations smaller and more prosperous. India needs both, and one cannot substitute for the other.

Grants from institutions such as the Gates Foundation and the work of organizations such as Digital Green are funding an AI-powered advisory service to reach a growing share of farmers. There is real potential (and early evidence) for higher yields, lower risk, and for the AI advisory to nudge away from subsistence farming.

Conclusion

Food production (farming) is Lindy. As long as there are humans who must eat to survive and thrive, the profession will continue, though what it means can, will, and should change dramatically.

But at the same time, we need to move beyond the romanticism of farming (which is true for a percentage of people) and look at the reality of farming. We need to ask the question as to why we have and need so many farmers and what we can do to help them make a choice that is best for them, without being on the subsistence hamster wheel.

For a country like India, AI advisory services will have a meaningful impact, but for true structural change to happen, we will have to look at other vectors beyond and alongside AI. Given the lack of political progress, organizations need to continue developing AI services that can help millions of farmers increase their annual incomes and move further along the path to prosperity and out of subsistence farming.

The tax collector from the movie “Lagaan” is gone. But the Indian state has continued colonialism by finding subtler ways to tax the farmer. The village in Lagaan won by changing the rules of the game. Indian farmers also need a different set of rules for open land markets, rational subsidies, and an economy with somewhere to go.

It would be great to release hundreds of millions of people from the drudgery of farming and subsistence. AI will play a significant role in farmers’ individual welfare, but it will also need support from other structural reforms.

I want to thank , , , and for their feedback on this piece. Any errors are purely my own.

“Software is Feeding the World” is a newsletter from “Metal Dog Labs.”

Water access might be an easier challenge for AI than almonds.

I am at a Blue Bottle Coffee shop in San Francisco, getting a latte with almond milk. The common discussion is all about AI, with San Francisco being the AI capital of the world. Along with crazy talk about a permanent underclass, the future of society, and dramatic GDP growth, the amount of water used by AI data centers is a topic of discussion in certain circles.

If you go 80 miles east of San Francisco to Ripon, CA, water is a very big and even more consequential topic of discussion. Ripon is the self-proclaimed “capital of the world for almonds.”

Almonds and water are a big discussion in California, as the state produces about 70-80% of the world’s production of almonds.

I am sure you have heard stats like you need more than a gallon of water to grow a single almond. It takes hundreds of gallons of water to produce a gallon of almond milk.

If you think these ratios are skewed, they were much more skewed about 30 years ago. The California almond industry and the ecosystem around it, along with regulatory changes, have increased water-use efficiency for almonds by 30-50% over the last four decades. Most of the changes in almond growing in California have been driven by the availability (or lack thereof) of water.

The big picture is that California’s average annual water supply has not cratered over the last forty years. The big changes have been in the volatility, timing, and type of water available, prompting changes in cropping practices.

Doing shots vs. sipping wine

The snowpack in the Sierras is declining and shifting earlier. The April snowpack across the Western US has declined by 24%1, with Northern California among the most affected in the last fifty years. Historically, the snowmelt from the Sierra Nevada provides roughly 75% of California’s agricultural water.

The snowpack acted as a slow-release medication, melting gradually, but rising temperatures have shifted the timing to a flash-flood profile. The state’s surface water reservoir infrastructure, designed in the mid-20th century to receive snowmelt steadily, cannot handle water that overflows and is lost to the sea.

At my age, it is like what happens if I do Tequila shots instead of sipping wine. I cannot handle it, either things overflow or I get dehydrated.2

The volatility in the water available from the snowpack has prompted significant groundwater withdrawal from the region3.

According to the State Water Resources Control Board,

Groundwater is one of California’s greatest natural resources. In dry years, it makes up to 60 percent of the state’s total water supply, serving as a buffer against the impacts of drought and climate change. Even when California has adequate precipitation, groundwater still makes up to 40 percent of the state’s total water supply. Groundwater is vitally important to California’s agricultural industry and is a major source of the state’s drinking water.

The California State Legislature adopted the Sustainable Groundwater Management Act (SGMA) in 2014 to protect and regenerate California’s groundwater, but there has been some sharp criticism around SGMA due to its economic impact.

Five five-year plans with volatile budgets

Imagine you are a mandarin in the CCCP or during Soviet Russia. You are responsible for creating the next five-year plan. Your job is extremely difficult because you don’t have visibility into inflows and outflows over the next five years.

You try your best and bake in a set of assumptions, based on bad data. Now multiply that by five times and imagine you have to do a 25-year plan. This is what it feels like when you are planning an almond orchard, when it comes to water availability during an almond orchard’s life span.

An almond tree’s productive, commercial lifespan is about 25-30 years. While trees can live for 40–60 years under ideal conditions, commercial orchards are typically removed after 25 years, when yields begin to decline. Peak production occurs between years 5 and 15.4

It is extremely difficult to model the volatility in water availability in California over a 25-year period. Old assumptions around water storage, the timing and quantity of water flow, and most importantly, the expected price of water do not hold true. Also, a water deficit in one year cannot be caught up with a water excess in the next year, as both less water and more water are not beneficial for the tree’s fruit potential.

Growers have had to plan on the assumption that they might not get all the water they need every year for twenty-five years in a row. The agricultural ecosystem has responded proactively to manage water issues associated with almond cultivation. Growers have had to manage so that the almond trees and orchards need much less water, allowing them to ride out the low end of water availability over a 25-year cycle.

For example, the Almond Board has cited a one-third reduction in water per pound for many years.

Improving Water-Use Efficiency

Changing the irrigation method

In the 1980s and 1990s, most almond growers followed flood-and-furrow irrigation. Flood and furrow irrigation in almond orchards is a surface, gravity-based method for delivering water. Flood irrigation covers the entire orchard floor, while furrow irrigation uses small trenches along tree rows, applying water only to specific areas to reduce usage. This method of irrigation was used to recharge groundwater, or when there was not enough capital to invest in micro-irrigation.

Most growers have moved from flood-and-furrow irrigation to micro-irrigation. Micro-irrigation, including drip and micros-prinkler systems, is used on over 80% of California almond farms to increase water-use efficiency by targeting water directly to the root zone. This type of irrigation ensures that 90-95% of the water is available to the almond tree roots, compared to much lower percentages with flood and furrow irrigation. This approach minimizes evaporation and runoff, supporting sustainable production while reducing water consumption per pound of almonds by 33% over the past two decades.

The move to drip and micro-sprinkler systems required capital investment, but it protected growers from the downside of water availability as long as water availability didn’t drop by more than a third.

A shift from calendar-based to evapotranspiration-based irrigation occurred in the 2000s and 2010s. Evapotranspiration (ET) based irrigation in almonds is a precision water management strategy that calculates the exact amount of water lost from the soil (evaporation) and the trees (transpiration) to determine how much irrigation water needs to be replaced. It ensures trees receive optimal water throughout the season, maximizing yield while minimizing waste.

Almond orchard evapotranspiration (ETc) is calculated by multiplying the weather-based reference crop (ETo) by a canopy-based crop coefficient (Kc). The weather-based reference crop (ETo) is available primarily from the California Department of Water Resources’ CIMIS network of nearly 145 weather stations, which provide daily reference evapotranspiration values.

Calendar scheduling treats the orchard water demand as constant. ETo scheduling treats it as a daily-varying physical quantity. The same orchard, on the same soil, with the same trees, can use 20–30% less water under ETo scheduling than under calendar scheduling, simply because the calendar approach systematically over-irrigates during cool weeks and shoulder seasons.

Precision Irrigation

Modern California almond orchards have largely converted from flood-and-furrow irrigation to micro-irrigation. These are drip and micro-sprinkler systems that deliver water to the root zone with minimal evaporation, runoff, or deep percolation losses. The Almond Board reports that micro-irrigation adoption accounts for roughly 85% of California almond acreage. This shift happened between the late 1980s and the early 2010s.

The precision-irrigation layer that sits on top of micro-irrigation and ET scheduling. This includes soil moisture sensors at multiple depths (tensio-meters, capacitance probes, neutron probes), stem water potential measurement using pressure chambers, and an emerging set of remote-sensing tools. It adds an additional 10-20% efficiency on top of what is already shaped by upstream decisions.

Genetics

The genetic decisions that govern an almond orchard’s water use are made in two places. The rootstock (the underground tree onto which the productive variety is grafted) and the scion (the productive variety itself). Both are bred over decade-long timelines by a small community of geneticists.

The rootstock determines how the tree extracts water from the soil. The root architecture, its tolerance to salinity, its drought response, and its ability to handle waterlogged or compacted soil play a role. Different rootstocks are different machines with different water curves.

For example, UC Davis breeder Tom Gradziel’s program has produced 14,000-plus seedling progeny in recent cycles, screened for a long list of traits (yield, kernel quality, harvest timing, pollinizer compatibility, disease resistance). Drought tolerance is on the list, but is not the top selection criterion.

Improvements in genetics take more than 10 years to come to fruition, and often the water-efficiency improvements will be gradual rather than a step change.

Reducing overall water usage

Throwing good water after bad

One of the largest determinants of an almond orchard’s lifetime water use is where it was planted. California’s almond geography is based on land availability, water-rights inheritance, and federal infrastructure. Soil conditions and water infrastructure affect water-use efficiency for almonds.

Picking the right location to grow and cultivate an almond orchard is an important decision, and it cannot be overturned for 25 years, unless one decides to discontinue the orchard before its useful life is over.

Picking the right location does not directly improve the water-use efficiency of almonds, but fallowing land unsuitable for growing almonds improves overall water-use efficiency at the aggregate level.

For example, the east side San Joaquin Valley orchards used 3.0 to 3.5 acre-feet per acre of water, with yields of 2800-3200 lbs per acre on mature, well-managed orchards, implying a water use of 350 to 400 gallons per pound of almonds. The west side of San Joaquin Valley orchards typically use 3.5 to 4.5 acre-feet per acre, with yields of 1800 to 2400 lbs per acre and an implied water use of 480 to 680 gallons per pound of almonds.

The per-pound water-use efficiency gap between these two categories of orchards is roughly 40–70%. The gap is due to factors such as water reliability stemming from senior surface water rights and stable groundwater basins, and soil quality, with the east side ground dominated by alluvial soils, whereas the west side soil has marine sediments with high salt concentrations.

The combination of sediments and poor drainage creates a salinity feedback loop, which accumulates salt in the root zone. This suppresses tree health and yield. The East side also benefits from lower summer temperatures and higher humidity, which reduces the evapotranspiration demand by 5-10%. Many farmers have fallowed acres on the Western side and moved them to the eastern side.

Many growers have stopped planting water-intensive crops that are not high-value on a per-acre basis. For example, the amount of cotton acreage5 has dropped significantly in California. This does not reduce the amount of water needed to grow almonds, but it reduces overall agricultural water consumption.

Water allocations

The Sustainable Groundwater Management Act (SGMA) is California’s plan to help communities manage their groundwater for the long term.

SGMA was enacted to halt the unsustainable use of groundwater and bring groundwater basins into balanced levels of pumping and recharge. Overuse and excessive groundwater pumping can overdraft aquifers, emptying them faster than natural systems can recharge them, leading to: lowering of groundwater levels, permanent loss of storage in aquifers, land surface sinking (subsidence), degradation of water quality, seawater mixing with fresh groundwater in coastal areas (seawater intrusion), and diminished surface water supplies where surface water and groundwater are interconnected.

The Sustainable Groundwater Management Act (SGMA) gave local groundwater sustainability agencies 26 years (until 2040, with a 2042 deadline for some basins) to bring critically overdrafted basins into balance. The first major implementation milestone was 2020, when local Groundwater Sustainability Agencies (GSAs) were required to file groundwater sustainability plans for the most overdrafted basins.

SGMA requires limiting groundwater extraction, which is forcing farmers to fallow land, reducing farm productivity, lowering land values, and causing financial distress. Estimates suggest significant job losses and economic decline in agricultural communities.

For example, during the 2021–2022 drought, surface water spot prices in the San Joaquin Valley exceeded $2,000 per acre-foot. Mature almonds use 3.5 to 4.5 acre-feet per acre per year, resulting in water costs of $7,000 to $9,000 per acre. It can exceed the total cost of all other production costs.

Nasdaq launched a California water futures contract in 2020. Districts have institutionalized fallowing-for-water programs in which growers with senior rights and flexible (annual) crops sell their allocation rather than plant. Farmers cannot fallow without killing the orchard, so they are price-takers. The SGMA restrictions are effectively shifting almond acreage to areas with lower groundwater dependence.

Geography, Infrastructure, and Policy story for the future

The one-third efficiency improvement in water use over the last 30 years has mostly come through changes in irrigation methods and evapotranspiration-based irrigation scheduling, with another layer of precision irrigation on top.

Since 1982, California has supported these efforts by creating a large weather station network, mentioned earlier as CIMIS, in partnership with the University of California at Davis. The precision irrigation layer has the potential to add a few more percentage points of improvement, but it does operate within the existing set of conditions. The combination of updated irrigation methods and new technologies, supported by infrastructure, has significantly boosted water efficiency, but the benefits of the technology curve might be approaching an asymptote.

The next curve will have to be a geography and policy story, with the removal of acres from marginal land, and effective management of water resources through SGMA.

I am a grocery-list-captured male shopper. If something is not on the grocery list, it most likely does not go into the shopping cart. There is one item, though, for which I make an exception. Whether it is on the list or not, I always get a pack of mushrooms because I love them. I love mushrooms in my soup, in my burgers, on my toast, or just sauteed with garlic1.

Given the short shelf life of mushrooms and their fragility, I had always assumed that most of the mushrooms I buy must be coming from some nearby place in California.

I recently learned that Canada’s mushroom production has been growing over the last 20 years, and much of it is exported to the United States, while production in the United States has declined. Differences in policy toward migrant workers between the United States and Canada, and differences in investments in new technology may explain the divergence in mushroom production between the two countries.

But before we get into the details, it is important to understand where and how mushrooms are grown, harvested, and shipped.

US mushroom production

You would be surprised to learn that almost 69% of the US mushroom production occurs in the borough of Kennett Square, Pennsylvania. It is a small town of about 6000 people, but mushroom-growing facilities around town produce almost 451 million pounds of mushrooms annually (2024). 451 million pounds of mushrooms would occupy about 45 American football fields or 35 soccer fields. The dollar value of mushroom production in the US is roughly $ 1 billion per year.

China is the undisputed leader in mushroom production. China accounts for 93% of the world’s global mushroom production.

The history of mushroom farming in Kennett Square dates back to 1885, when a grower obtained mushroom spores from Europe and began growing mushrooms. This concentration of mushroom farming in Kennett Square, Pennsylvania, is due to historical immigration patterns, primarily of Italians in the 1950s or 1960s, easy availability of horse manure for the mushroom substrate, an easy access to the Philadelphia and New York markets.

Growers use an old system called “Pennsylvania doubles” to grow mushrooms. Specialized, two-story cinderblock buildings with wooden shelves and intensive manual picking characterize the system. The system is designed with the assumption of cheap labor.

The growing houses provide a strictly controlled environment for growing white button, cremini, and portobello mushrooms on stacked beds, producing approximately 400 to 500 million pounds of mushrooms annually. Growers can manage the temperature, humidity, and airflow to create optimal conditions for mushroom mycelium to grow and fruit.

The houses are equipped with vertical wooden or aluminum shelves, which maximize growing space. The shelves house pasteurized compost (often made from hay, straw, poultry litter, and cocoa shells) used to grow mushrooms year-round.

As you can see in the video below, the conditions inside the mushroom-growing facilities are hot, humid, and stinky! The process of harvesting mushrooms is fairly manual, unless the grower has invested in a robotic harvesting system from companies like 4AG Robotics. Most US production facilities lack an automation design.

Mushroom shelf life dictates the supply chain

Mushrooms are a type of fungus. If you have the right spawn available and can control the environment economically, you can grow mushrooms year-round. Mushrooms have a short shelf life. Mushrooms are 92% water. A mushroom starts losing water as soon as it is harvested. Anyone who has seen a fresh mushroom that has begun to dehydrate knows how unappetizing it can look.

The dominant mushroom variety grown in both the U.S. and Canada is called the Agaricus. Your most common mushroom variety in your grocery store, the white and brown button mushrooms, cremini, baby bellas, and portobello, all belong to the Agaricus family. The Agaricus family accounts for more than 90% of mushroom production and sales in North America. Shiitake (my favorite or oyster mushrooms) do not belong to the Agaricus family.

Canada and the United States grow mushrooms year-round in climate-controlled, indoor warehouses. Readers of this newsletter are aware of my massive skepticism about the economic viability of vertical farming, but mushrooms provide a counterexample in which vertical farming actually works. The main difference is that mushrooms are fungi and do not need sunlight for photosynthesis.

Due to supply chain constraints and the limited shelf life of mushrooms post-harvest, most fresh mushroom consumption occurs within a few days of production. For example, mushroom production from Pennsylvania mostly stays within a few days of transit.

Most mushroom production facilities in Canada are located in British Columbia and Ontario, close to the US border, and deliver their products to the northern United States within 36-48 hours of harvest. Production geography relative to population is the structural constraint that neither shelf-life extension nor improved cold chains can fully overcome.

The shelf life of mushrooms is the hard constraint. Fresh button mushrooms have a 7-10 day usable shelf life under an optimal cold chain, beginning at harvest. Mushrooms are a high-respiration-rate product. They consume oxygen, produce CO2, and generate moisture. Every degree of temperature above the ideal range of 34-38 degrees F doubles the respiration rate and halves the effective shelf life. Continuity in temperature from the moment of harvest to when the customer picks it up is a critical supply chain variable.

Mushrooms are fragile and bruise under their own weight. Vibration and pressure can cause bruising in transit. Each bruise initiates a localized decay, which accelerates from the point. It limits the number of handoffs or transfer events, since each event is a risk. Mushrooms can lose quality if they dehydrate or become too heavy, and they require an ideal relative humidity of 90-95%.

The same MAP technology used for packaged salads is also used for mushrooms and can extend their shelf life, though it cannot do much for the product’s fragility or the minimum handling requirements.

So, why is US production dropping while production in Canada is rising, even though 99.6% of Canada’s exports go to the US? A big part of the answer to this question lies in how the United States and Canada provide support to migrant workers who come over to pick mushrooms.

Seasonal vs. Annual

Both countries grow mushrooms year-round in climate-controlled indoor warehouses. Mushrooms are not a seasonal crop. But the labor policies of the U.S. and Canada treat it very differently.

The primary flaw in the U.S. is a mismatch between the nature of mushroom farming and the design of the guest worker visa system.

The H-2A visa program, introduced by the Immigration Reform and Control Act of 1986, provides foreign workers with legal status for seasonal agricultural labor. Unlike most agricultural products, however, mushrooms can be grown and harvested year-round. Accordingly, the U.S. government provides no clear pathway to legal status for noncitizen workers in the mushroom industry.2

As mushroom farming is a year-round specialty crop, it doesn’t qualify for the H-2A temporary guest worker program. Without the H-2A labor option, growers concentrated in the Pennsylvania region are competing fiercely for local workers in a tight labor market.

The industry has used a mix of undocumented workers, immigrant workers with temporary protections, and labor contractors. The termination of TPS (Temporary Protected Status) for Venezuelan migrants, combined with broader immigration enforcement, will likely shrink the pool of available workers further.

Canada has taken a more flexible and stable approach. The Seasonal Agricultural Worker Program (SAWP) of Canada includes mushrooms. The national commodity list for SAWP workers includes mushrooms alongside fruits, vegetables, grains, dairy, and other sectors.

Foreign workers can be in Canada for up to two years at a time under the Temporary Foreign Worker Program (TFWP), and SAWP workers for up to eight months. They can return each season, creating a reliable, trained, recurring workforce. Many workers return to the same Canadian farm year after year.

Canada also offers a pathway forward, through the Agri-Food Immigration Pilot, where experienced workers in mushroom cultivation can qualify for permanent residency, either directly or through a Provincial Nominee Program.

H-2A is not a mid-century modern problem

The H-2A program3 was designed for seasonal crop agriculture in the mid-20th century. It did not account for year-round agriculture across multiple sectors; despite efforts by the US legislature, the seasonal restriction has not changed.

The consequences in the U.S. could be tangible if the U.S. allows year-round migrant workers under existing visa schemes.

The American mushroom industry is estimated to be missing more than a million pounds of production per week due to labor shortages, with some companies reporting 10–20% of their crop going unharvested. Some Pennsylvania farms have slowed operations by as much as half, with growers saying they cannot harvest all the mushrooms they’re growing and are struggling to pay the bills.

The impact of seasonal H2A workers is not limited to mushrooms only.

It creates a legal labor vacuum across all commodities that operate year-round. It includes year-round industries such as dairy, beef cattle and other livestock operations, hog and poultry production, mushrooms, nursery and greenhouse operations, some tree fruit operations, aquaculture, and forestry. These sectors are a significant portion of the US agriculture GDP and the food system.

According to the National Milk Producers’ Union,

Immigrant labor accounts for 51% of all dairy labor. Dairies that employ immigrant labor produce 79% of the US milk supply. If the US dairy industry lost its foreign-born workforce, it would nearly double retail milk prices and cost the total US economy more than $32 billion.

The H2A seasonal restriction makes this a structural and permanent risk.

For seasonal crops like fruits and vegetables, where H-2A technically does apply, the program is increasingly accessible but still deeply dysfunctional in practice. For fruit and vegetable production, labor accounts for 38.5% and 28.8% of input costs, respectively, as I highlighted in my blueberry and lettuce pieces a few weeks ago.

This cost gap is directly driving a structural shift in where specialty crop production happens. For fiscal year 2024, USDA’s Economic Research Service estimated a record $32 billion agricultural trade deficit, with the largest increases in imports driven by labor-intensive products such as fresh produce and horticultural goods.4

The biggest impact will be on year-round industries such as dairy, livestock and pork operations, nursery and greenhouse operations, and sugarcane and sugar beet operations. As I said before, the dairy sector faces a $32 billion tail risk amid ongoing operational stress from labor issues. The challenge is similar to that of the beef cattle and meatpacking industry, which is worth north of $100 billion and relies on an undocumented workforce with limited legal alternatives. The poultry and hog industry also poses a $ 50-$ 60 billion tail risk.

Mushrooms are a microcosm of larger issues

US mushroom production is declining, while Canadian production is rising. All of Canada’s mushroom exports come to the US due to higher quality and greater economic competitiveness than its US counterparts. At this point, we do have to admit that Canadian mushroom producers have some real structural cost advantages over US producers. Canadian producers have vertical integration, newer facilities, larger average farm size, proximity to large US markets, and access to SAWP labor.

Regulatory changes have been challenging to get through. US producers have been unwilling to invest in new technologies, such as steel trays for holding substrates and growing mushrooms, and in automation to alleviate high labor costs. Mushroom growers in the Netherlands have overcome the labor issue through investments in the latest growing equipment and robotic picking solutions. Canadian growers have similarly invested in newer technologies, including robotic picking solutions, and have the advantage of year-round labor.

The US mushroom industry is stuck in last-century methods and has been hesitant to invest in new technologies. Competition from Canada and pressure from grocery retailers are forcing some growers to upgrade their equipment.

Based on my conversation with an industry insider, US grocery retailers say they get very good mushrooms at reasonable prices in the northern US (for example, Minneapolis, Chicago, etc.) that come over the border from Canada. They are asking their US growers to do the same for the geographies closest to their operations, forcing some of the old US-based mushroom growers to update their setups and consider adopting robotic mushroom-harvesting solutions.

Without regulatory changes and some unwillingness among US producers to invest in new technology, the US risks the mushroom industry shrinking, maybe disappearing, and moving to Canada. The impact of H2A on mushrooms is clear and fairly easy to study and understand, given the industry’s smaller size.

The U.S. Commerce Department has initiated dumping and countervailing investigations, but it is more of a tell about what’s really happening on the ground. We have a shrinkage in crop acres across multiple commodities, such as berries moving from the US to Peru, asparagus moving out of California, and sweet beetroot moving out of California.

A common criticism of allowing year-round guest workers in agriculture is that it will depress wages for local workers and might slow the adoption of automation and robotics. The practical reality is that many American workers are not interested in seasonal farm work. If anything, the adverse effect wage rate (AEWR) requirements for the H-2A program (which have grown 185% in the past 10 years) have put a floor on farm wages and raised average farm wages.

The bigger challenge with automation has been the availability of harvesting technology and the unwillingness of mushroom growers to upgrade their technology to make it suitable for automation. Automation still requires labor for other tasks, which is a challenge due to the seasonal workforce requirements within U.S. law.

Mushrooms might be a case of a canary in a coal mine or a fungus in a growing house. US producers, the agricultural technology ecosystem, and policymakers need to pay attention if we Americans want mushroom independence.

Dr. Peter Norvig, eminent AI scientist and professor at Stanford, Google researcher, and the author of one of the most widely read textbooks on AI, Artificial Intelligence: A Modern Approach, and I have something in common. If you see Dr. Norvig’s photos online, he is always wearing a colorful shirt. I do have a small collection of themed shirts for agrifood events, with the tractor-themed one2 being among the most popular!

Dr. Norvig was the opening speaker at the AI and Agriculture Immersion Day at Stanford’s Computer Science and Data department. Stanford’s Center for Human-Centered Artificial Intelligence (HAI) and McKinsey organized the session. So they had much better food for lunch than what I experienced at World Agritech 2026 in San Francisco on Tuesday and Wednesday.

The audience included a bunch of food and agriculture executives, McKinsey partners, Stanford researchers, and a random lowly newsletter writer!

I had gone to the session with two specific questions in my mind.

1. What is the shape of problems that are most tractable for AI?

2. If the cost of experimentation is low, should we focus on use cases or just brute-force-apply AI and reimagine workflows?

I’m in love with the shape of you

My first question was answered right off the bat by Dr. Norvig’s presentation. He highlighted a class of non-physical problems: data-rich, with messy, distributed data, executable or simulatable, easy to verify, and semi-personalized and custom.

He highlighted sectors ripe for transformation based on his criteria. These sectors include wealth management, banking, paralegal, compliance, customer support, healthcare administration, real estate and insurance, drug discovery, and supply chain management. Software engineering fits in very well in this class of problems. There are tons of code samples (data), and the most important part is the ease of verification. Your code either works or it doesn’t.

AI adoption in agriculture faces challenges in verification and attribution. You provide an AI recommendation to plant a particular variety of seed or to apply the right timing for spraying your fungicide. The farmer often has to wait for a few weeks or months to see whether the recommendation worked. Even if the result is good, it is difficult to attribute the outcome to that specific recommendation because of numerous confounding factors, such as weather, soil conditions, and management practices. (I wrote a whitepaper with Tenacious Ventures on this topic in 2024)

I asked Dr. Norvig about Moravec’s Paradox and the challenge with physical tasks. Humans often do these tasks. For example, labor for harvesting specialty crops accounts for 60% of total production costs in California. Dr. Norvig was cautiously optimistic but felt we might be a few years away from robots picking apples or strawberries.

Dr. Norvig also presented a list of how not to fail with AI adoption (this bulleted list is taken directly from his slide)

• Al is not regular IT!

• Go bottom-up

• Have a clear use case

• Provide training and support

• Make relevant data available

• Optimize processes, don’t necessarily look for revenue

• Problem-solving is a contact sport

• There will be uncertainty

• Make it everyone’s business

• Encourage communication

As you can see from this list, most AI adoption challenges have less to do with technology and more with human factors. According to Dr. Norvig, we have done a great job of focusing on the easy parts of the job.

Dr. Norvig presented Amdahl’s law, which helps us consider how certain jobs can be sped up or automated in any field, and its implications for the food and agriculture system.

In 1967, Gene Amdahl3 made a simple observation about parallel computing. The speedup of a program is limited by the fraction of the computation that must run sequentially. Amdahl’s Law highlights that the maximum performance improvement in a system is constrained by its most significant bottleneck, making it essential to focus optimization efforts on the components that contribute the most to inefficiencies. (I had highlighted how bottlenecks and shifting bottlenecks have a huge impact on the overall process performance in my cotton article.)

Amdahl’s law is very important to think in terms of its impact on automation and which jobs are under threat. Dr. Alex Imas broke this down in a fantastic Substack article by looking at the dimensionality of a job.

On the other hand, if AI automates all of the tasks—let’s say your job only involves two tasks, and they both get automated—then yes, human labor will get displaced. Importantly, the fewer the number of tasks (what we call the dimensionality of a job), the greater the company’s incentive to automate it in the first place. This is the part much of the analysis on automation misses: adopting AI into an existing organization is costly, so the firm will be more likely to invest if it can automate the job, not just the task.

If you think about the activity of harvesting a specialty crop like lettuce or strawberries by a human, it requires a large series of tasks. Many tasks require experience and decision-making within a split second. For example, is this strawberry ripe enough for picking? Where should I cut the lettuce stem to get the product off the ground without damaging the lettuce? How many lettuce leaves should I take off, and which ones, before packaging the head of lettuce? People who have studied this more closely have told me that a farm worker has to make more than 20 decisions and take actions to get a head of lettuce from the ground and into a plastic bag, while maintaining the quality expected by the end consumer.

Even if we can automate some tasks in this process, it will be challenging to eliminate human labor unless we are willing to compromise on other dimensions, such as quality, food waste, or other aspects.

When in doubt, use brute force

My second question for the evening was about how indiscriminately we should try to apply AI in our day-to-day lives and businesses. Should we apply AI to as many use cases as possible?

Dr. Norvig had already mentioned the idea of having a clear use case for AI. He also presented the concept of Exaptation. Exaptation is a concept from evolutionary biology, but it could be very useful for AI in general and for AI in agriculture in particular.

Exaptation4 is a feature that performs a function, but that was not produced by natural selection for its current use.

The direct implication of exaptation is that we should have ideas floating around within the organization. We should think of AI as a general-purpose technology (GPT) rather than a specific tool. A 2023 research paper, “GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models.”

In real life and especially in the online writer community, many people have also called it “increasing the surface area of serendipity.” The basic premise is that if you put your ideas out in public, you increase the chances that interesting people will engage with them. The paper referred to GPTs (LLMs) as GPTs (General Purpose Technologies).

The concept of exaptation should not be unique to AI. Still, given the innumerable possibilities of AI, it makes sense to have ideas about AI floating around within your organization, so that it increases the chances of someone else finding novel, not-designed ways to solve existing problems or solve new problems.

The presentation also alluded to Ken Thompson’s (inventor of the UNIX operating system) epigram, “When in doubt, use brute force.” Ken Thompson was implicitly saying that simple solutions are often better than complex solutions. If a problem is not excessively large, a simple, brute-force approach can be written and verified quickly, rather than wasting time on a “clever” algorithm that may be harder to maintain or more prone to bugs. It often works better when human time (programmer time) is much more expensive than machine time. (highlight by me)

What is the direct implication of this principle? I am not 100% sure.

When we find a domain or a set of problems that fit Dr. Norvig’s criteria, should we apply AI indiscriminately if running an experiment is not very expensive? We can potentially afford to do this because of AI’s general-purpose nature, provided the cost of experimentation is low. A great example of this approach went viral on X/Twitter when someone used ChatGPT to find treatment for their dog’s tumors. (This is not generally applicable or available.)

Please let me know if you have any thoughts on when and how AI should be applied.

Touch the grass

The second part of the presentation included a talk by Dr. Riitta Katila. Dr. Katila is a very energetic and engaging presenter, and the

Professor of Management Science, Faculty Director of the Stanford Technology Ventures Program, and HAI Sabbatical Scholar at Stanford Institute for Human-Centered Artificial Intelligence at Stanford University. Her research is in the intersection of technology strategy and organizational learning, using machine learning, statistical analysis, and mixed methods.

Dr. Katila’s presentation went deeper into the shape of problems, where AI is well-suited. Given the jagged nature of AI5 (and humans as well), it is a highly context-specific question, and AI can often be more creative because it might not consider constraints, whereas humans can implicitly consider implementation constraints.

Dr. Katila shared data from a paper by Brynjolfsson et al. (2025), which showed that using GenAI in a call-center setting improved the performance of the lowest-performing humans by more than 15% when the AI model was trained on inputs from the highest performers.

Even though AI in the call center boosted efficiency and customer satisfaction, highly skilled workers follow AI suggestions more often, even when those suggestions slightly lower the quality of the conversation. Because top workers are contributing fewer original, high-quality solutions, the AI has no “new” expert data to learn from. There is a risk that future AI iterations may be less effective at solving new or complex problems because their own average output is replacing the human expertise they mimic. There couldn’t be a more classic definition of “regression to the mean”!

In this use case, if you optimize AI agents for growth, you may underinvest in novelty and exploration. Dr. Katila recommended protecting innovation through human “exploration lanes”, incentives for original contributions, and deliberate non-AI inputs (real users, field experiments).

Touch the grass!

Where are you drawing the shape?

Discussing different contexts and problem shapes, she also presented data from Uber’s Q4 2025 earnings, which showed why San Francisco is a uniquely favorable market for AV (autonomous vehicle) deployments.

And how Uber’s network of human drivers allows Uber to have the highest AV utilization, when AV demand is highly variable based on the day of the week and the time of the day.

It highlights that not only is it important to consider the shape of problems for AI to tackle, but also to deeply understand the conditions in which they exist to understand adoption challenges and the economics behind them.

Is it time to cancel my medical insurance and pay for ChatGPT?

She presented data showing that an AI model specifically tuned on medical research data significantly underperformed a frontier model like ChatGPT-4o (We are on ChatGPT-5.4 now) in a clinical setting. OpenEvidence, the fastest-adopted software product in hospitals, uses a fine-tuned model to assist doctors with diagnosis and decision-making. I have to admit, I had never heard of OpenEvidence before. OpenEvidence’s performance is significantly worse than ChatGPT’s.

The research paper indicates that,

A notable strength of OpenEvidence is its integration of citation-linked, up-to-date references, an asset for clinicians involved in academic writing, literature reviews, or manuscript preparation. This feature makes OpenEvidence particularly well-suited for research-oriented tasks. However, its reliance on well-structured and precise queries may limit its practical utility in real-time clinical scenarios. In contrast, ChatGPT-4o exhibited strong contextual awareness, consistently interpreting vague, incomplete, or grammatically flawed questions with ease. This advantage enhances its usability in fast-paced, real-world settings where flexibility and immediacy are essential. ChatGPT-4o demonstrated higher accuracy, significantly faster response times, and greater interreviewer consistency, suggesting it may serve as a more reliable and efficient tool for point-of-care education and clinical decision support.

Once again, the problem’s shape, the data used to train your model, and the context become extremely important. Also, note to self. During my next doctor’s visit, ask the doctor to turn off OpenEvidence!

Understand shapes, contexts, and make choices

The AI immersion day6 highlighted the importance of understanding the shape of problems when applying AI, recognizing its general-purpose nature, practicing exaptation, recognizing human creativity, and thinking from first principles.

As humans have always had to do throughout history, we have to make conscious choices.

AI is absolutely creating anxiety throughout the economy. Parents like me with high-school and middle-school-aged children are thinking and worrying about what an AI-powered future will look like.

I want to leave you with some sane words from Ada Palmer, a University of Chicago historian,

The dangers of ChatGPT and its successors do not lie in the technologies themselves. They lie in the fact that we must now make choices, good or bad, about how we help those navigate the rollout effects.

Knowledge empowers; experience empowers; expression empowers; each of these twenty generations has been more powerful than the last.

That’s why, as AI stuns us with its possibilities, we need to take a deep breath, zoom out, and remember that information revolutions are the normal state of human life.

There is no better illustration of second-order effects than the evolution of America’s cotton industry.

Sea Island Cotton was the Lamborghini of cotton in the late 1700s and early 1800s. Prior to the 1790s, it was the only commercial cotton variety grown in the U.S.

“The Sea Island cotton, with its long, silky fibers, extraordinarily fine texture, and luxurious softness, could produce the very finest thread count, unparalleled by any other fabric in the world.”

Sea Island produced long staple, silky fibres that were highly valued by the British “upper class”. Think upstairs in Downton Abbey, or the “Ton” of Bridgerton.

The term “staple” refers to the length of cotton plant fibers used to weave a piece of cotton fabric. Shorter cotton fibers (“short staple”) are quicker to grow and therefore cheaper to produce and sell. When short-staple cotton fibers are used to make cotton fabric, the fabric is rougher, stiffer, and less durable.

Cotton goes through multiple steps from the field to the gin. Cotton is picked in the field during harvesting, and it is then transported to the gin, where it is dried, cleaned, separated into lint and seeds, and prepared for further storage.

Cleaning was the first bottleneck

Harvesting Sea Island cotton was relatively easy. It had smooth, black seeds. If you pinched the lint, the seed would pop out easily. It was highly profitable because it was easy to clean by hand or with a simple, ancient hand-cranked roller gin.

The Sea Island variety was very fragile to grow. It required a very long, frost-free growing season and high humidity. It could only be grown on the barrier islands and immediate coastlines of South Carolina, Georgia, and Florida. If you planted it inland, the frost would kill it.

Farmers of the time were aware of another variety called Upland cotton, which had a short staple. Upland was a tough, robust plant that could survive the colder winters and drier soils of the inland American South. Unlike Sea Island cotton, Upland had fuzzy, green seeds that stuck to the fiber like superglue.

It was extremely difficult to separate the lint from the seed. It took an enslaved worker an entire, grueling day to clean just one pound of short-staple cotton by hand. Because the labor cost of cleaning it vastly outweighed the price it could fetch in England, Upland cotton was considered a worthless weed for commercial export.

No one had grown Upland cotton commercially before 1793.

But Eli Whitney invented the cotton gin to specifically solve the “sticky seed” problem of short staple cotton. Whitney’s gin, instead of rollers, used a cylinder with wire teeth (subsequent versions used circular saws) to pull the cotton through a narrow wire grid. The lint passed through, but the sticky seeds were too big and were stripped away.

A single hand-cranked Whitney gin could clean 50 pounds of short-staple cotton like Upland per day. Now it was feasible to grow cotton inland.

The gin made short-staple cotton profitable by “easing the removal of seeds from the cotton boll” problem. The planters were no longer confined to the coast, as Upland cotton could grow anywhere in the South.

The Whitney gin shifted the bottleneck from cleaning cotton to planting and harvesting cotton inland.

This completely inverted the economics of cotton production. It triggered a massive, violent land rush and opened up millions of acres of newly-prime inland real estate specifically for short-staple cotton plantations. The planters required significant labor to plant, cultivate, and harvest large quantities of cotton.

The United States was already exporting cotton to England. The British textile industry mechanized in the early 19th century. Power looms and spinning jennies in Manchester created an insatiable, bottomless demand for raw fiber. Britain didn’t care that short-staple wasn’t as luxurious as Sea Island; they just needed massive volume to feed their factories.

It massively expanded and entrenched slavery as a profitable labor system. The enslaved population in the South grew from 700,000 in 1790 to nearly 4 million by 1860, with cotton becoming “King Cotton” and accounting for 50–60% of U.S. exports.

Today, the Upland variety accounts for 97% of US production, with Pima (a long-staple variety) accounting for the remaining 3%. The states that produce the most Upland cotton in America include Texas, Georgia, and Arkansas. California is the top producer of the Pima variety.

The next obstacle was picking

Whitney always intended the gin to reduce reliance on labor. But it would take another innovation in the following century to reduce the reliance on any kind of human labor. (including the abolition of slavery)

When cotton is picked, roughly 60% to 65% of its weight is the seed, and only about 35% to 40% is the valuable lint. Hauling loose seed cotton long distances to a massive, centralized “cotton elevator” (processing facility) would mean farmers were paying to transport immense bulk and weight, most of which was low-value seed.

A cotton plant doesn’t open all its bolls at once. A single field might be picked over three or four times by hand from late August straight through to December. This created a slow, steady stream of seed cotton that could be taken directly to the local gin in wagonloads as it was picked, eliminating the need for massive, sudden storage capacity.

Because of this, the economic incentive was to gin the cotton as close to the field as possible. Instead of centralized storage silos, as with grain, the South built a highly decentralized network of thousands of local, community-level cotton gins. In the late 1800s and early 1900s, almost every small farming town in the South had its own gin.

The bottleneck in overall production was harvesting speed, not ginning speed.

In the 1940s, a new mechanical cotton picker designed and patented by John Rust became popular. The widespread adoption of mechanical spindle pickers and strippers in the 1940s and 50s meant that a machine could harvest in a few hours what had previously taken a human crew an entire week to harvest.

The bottleneck shifted from harvesting speed to ginning speed after cotton had been harvested. Suddenly, the local gins were overwhelmed.

Mountains of loose cotton piled up in trailers in gin yards, waiting to be processed. Small, older gins could not keep up. To survive, a gin had to invest in larger, faster machinery (bigger saw gins, massive natural gas dryers, and automated presses). Those who couldn’t afford the capital upgrades were forced to close.

The expansion of paved rural roads and the use of motorized trucks meant a farmer was no longer restricted to the gin just two miles down the dirt road. A farmer could easily load a truck and drive 20 miles to a larger, faster, and cheaper gin, destroying the “one gin per town” monopoly.

It led to a sharp decline in demand for manual agricultural labor and resulted in a rapid consolidation of farms (fewer, larger operations) and a shift from labor-intensive to capital-intensive agriculture.

It drastically reduced the number of cotton gins from more than 25,000 at the beginning of the 20th century to about 1,000 at the beginning of the 21st century. Today, according to the USDA, there are only 446 active cotton gins in the United States.

Farmers were picking cotton faster than they could process it

Before 1971, cotton was hauled in wire mesh trailers. If the gin was backed up, the farmer’s trailers were stuck at the gin, meaning the harvester in the field had to stop working because it had nowhere to dump the cotton. Even if a farmer had redundant trailers, the picker was much faster compared to how much cotton could be processed by the nearest gin, when the cotton was coming from multiple farmers

The final blow to the small, local gin arrived in 1971 with a Texas A&M engineering invention.

The cotton module builder compacted the cotton into a massive, dense, freestanding brick right on the edge of the field. It decoupled the storage from the cotton ginning. The cotton could safely sit on the ground under a tarp protected from the elements. This allowed the farmer to keep harvesting indefinitely without buying a single extra axle or tire, while specialized trucks shuttled the modules to the gin at a steady, manageable pace.

Because the cotton could now sit safely at the edge of the field for some time, the frantic rush to the local gin was reduced. Gins no longer needed to be within 20 miles of a farm. This allowed multiple small gins to merge into single “mega-gins” that could pull modules from a 100-mile radius and operate continuously through the winter.

Store and don’t stop harvesting

Another major technological change occurred in 2007, when John Deere introduced an on-board wrapping capability. It eliminated the need to stop the harvester to dump cotton and protected the cotton from rain with plastic wrap.

As the modern harvester drives through the field, it builds, wraps, and drops a 5,000-pound round module out the back without ever stopping. A support tractor equipped with a specialized front-end loader or hydraulic “spear” drives through the field, picks up the dropped modules, and lines them up perfectly at the edge of the field (the turn-row). They are often grouped in sets of four or six, perfectly spaced for a truck to load them quickly.

Before 2007, if a farmer bought another 1,000 acres, the farmer had to buy another harvester, a boll buggy, a module builder, and hire four more people. The round-module harvester allowed one person to do the work of a whole crew. Freed from the labor bottleneck, large farming operations rapidly bought up smaller farms, leading to massive consolidation of farmland.

It drastically reduced farm labor needs (from 5 people to 1) and allowed gins to store cotton outdoors for months, triggering the massive consolidation of small farms and local gins into “mega-operations,” and further reducing the number of cotton gins in the U.S.

The old, rectangular cotton modules required highly specialized, expensive “chain-bed” trucks to winch the 32-foot loaf onto the truck bed. Round modules changed the game by making transport more modular.

Round modules are designed to fit onto standard flatbed trailers. A typical flatbed can hold four to six round modules. Because they are just heavy cylinders, a forklift or tractor can load a truck in minutes. This allows a small fleet of trucks to continuously cycle between the farm and the gin, moving massive tonnage with very little downtime.

When the trucks arrive at the gin, they don’t immediately dump the cotton into the machinery. They unload the modules into a massive, meticulously organized staging area called the “gin yard.” Because the plastic wrap protects the cotton from rain and snow, the gin yard acts as a giant outdoor warehouse.

You will often see tens of thousands of round modules lined up in perfect rows spanning dozens of acres. This “buffer” allows the gin to operate 24/7 at a steady, efficient pace from October through February or March, long after the farmers have finished harvesting.

Conclusion

Throughout the history of cotton, a new technology innovation, whether it was cotton picking, separating the seed from the bolls, or storage on the field with little or a lot of protection from the elements, shifted the bottleneck for the overall throughput and efficiency of the process to a different operation.

A technological change triggers known first-order and often unanticipated second-order changes.

One way to analyze and anticipate second-order effects is to understand how the new technological change is shifting constraints and bottlenecks, and which processes were held back by the old constraints.

This type of thinking will be especially important within food and agriculture as we consider the impact of slowing population growth and eventual decline, an aging population, GLP-1s, automation, and AI.

I want to thank , , and for their review and feedback. Any errors are obviously my responsibility.

The packaged salad is a triumph of food safety innovation, materials science, genetics, supply chain and logistics, vacuum cooling, biology, chemistry, vision systems, robotics, and process improvements.

In 1987, packaged salads were less than 1% of total produce sales.

Producers like Taylor Farms, Dole Food Company, Fresh Express, Earthbound Farm, and Organic Girl dominate the packaged salad category, accounting for a lion’s share of production and sales.

In 2026, the retail value of packaged salads had grown to roughly three times the size of the entire bulk lettuce market. (Bulk lettuce refers to whole-head lettuce or loose salad greens)

Subscribe now

How did we get here?

What’s driving the shift? The most obvious factor is convenience. Packaged salads and greens are easy. The leafy version of ramen. Salad kits with pre-packaged toppings and dressings are actually good.

Salads are healthy. You can buy a packaged salad with all the fixings at your favorite grocery retailer. Salads come in ready to eat, you can customize them with the fixings and dressings, and there is no cutting or cleaning required!

We got here thanks to technology that enables producers to sanitize greens and safely extend their shelf life, so that consumers on the other side of the country can enjoy them.

This is an unbelievably time-constrained endeavor. From the moment fresh lettuces and leafy greens are first harvested to when they are packaged and ready to ship, only 24-72 hours have elapsed.

This is a significant amount of time for produce, which deteriorates each day after harvest due to natural biological processes.

A banana, 7 days after you buy it from the store, looks and tastes very different from when you bought it.

Your raspberries look and taste like crap (which is a “legitimate” food science term!) if you leave them outside for too long, or even if they have stayed in your refrigerator for too long.

The expansion of the packaged salad category into a $15.6 billion industry in 2026 is driven by specific technological and process innovations.

So what are the specific steps to get a packaged salad to you?

Share

First, we use vacuum cooling to evaporate

Field heat is the latent warmth from the sun and the ground absorbed by the produce. It dramatically accelerates respiration and microbial growth. Rapid heat removal is critical because it effectively halts this degradation, preserving quality and extending the product’s commercial life.

Placing harvested lettuce in a vacuum chamber lowers the boiling point of water. A tiny amount of moisture on the leaf “boils off” at room temperature, pulling heat away from the lettuce’s core.

Most packaged salad producers follow the 4-hour rule from “cut to chill”. (It means no more than four hours should elapse between when the lettuce is cut/harvested in the field and when it is cooled.)

This process lowers the temperature from 70°F to 34°F in 20 minutes. It does so by uniformly dissipating heat without causing damage.

Vacuum Cooling adds 2 to 3 days to shelf life.

Then we cut it, wash it (three times), and dry it

As shown in the diagram above, leafy greens are cut before being washed, dried, and packaged.

Bacteriophage Cocktails

Instead of relying solely on chemicals to clean and process lettuce, processors are now using bacteriophages-viruses that specifically target and kill ‘bad’ bacteria like Listeria and E. coli.

This approach directly addresses consumer concerns about pathogen risks by providing a natural, targeted method that achieves a 99% reduction in harmful bacteria, ensuring safer salads without compromising quality. Phage treatments are applied as a fine mist post-wash and remain active during storage, offering ongoing protection during transport.

Triple-wash

Modern wash systems are designed to prevent “cross-contamination” (where one bad leaf ruins the whole batch. Rather than relying on high levels of chlorine, processors use wash enhancers that stabilize the water’s pH and ensure the sanitizer remains effective even when the water becomes “dirty” with organic matter (dirt and juice).

In 2026, more facilities are integrating phages, which are “good” viruses that specifically hunt and kill Listeria and Salmonella. The phages are misted onto the leaves as a final safety layer that continues to “work” inside the bag during transport. It acts as a buffer, maintaining a precise pH (typically 6.0–7.0).

Improvements in washing technology have increased shelf life by 2-3 days.

Modern production lines use hyperspectral imaging and AI to ‘see’ things humans cannot. Cameras scan leaves in real time, detecting spectral signatures of spoilage, fecal contamination, and foreign materials such as plastic or insects.

This technology ensures only high-quality, safe products reach consumers by reducing contamination risks and increasing the percentage of good products shipped, thereby enhancing food safety and quality assurance.

High-speed air ejectors that fire a burst of air to “flick” the specific bad leaf off the belt handle rejections.

These vision and robotics systems increase the percentage of good products shipped and reduce the risk of product contamination.

Genetics for processability

In the 1990s and 2000s, breeders realized that standard lettuce was too fragile for factories. They began selecting and breeding for “processable” traits. Genetics helped change the cellular architecture so that leaves with smaller, tightly packed cells and stiff walls survived the mechanical spin-dryers in packaged lettuce plants without bruising.

Genetics also played a major role in eliminating “pink rib”, a genetic predisposition in some romaine varieties to turn pink/red when stressed or cut.

Lately, the industry has shifted from traditional cross-breeding to genomic selection to identify and select plants that keep lettuce crisp even in suboptimal fridge temperatures.

These genetic improvements have improved lettuce processability in salad factories and increased the likelihood that it stays fresh and appealing to consumers.

Then we put the lettuce to sleep in a special bag

One of the earliest innovations that extended shelf life and transportability for packaged salads was slowing the biological process by “suffocating” the lettuce and putting it to “sleep”.

In 1989, Fresh Express introduced Modified Atmosphere Packaging (MAP) technology. It uses breathable bags that balance oxygen and carbon dioxide levels, preventing leaves from wilting. MAP uses semi-permeable membranes to create a custom atmosphere.

Normal air is roughly 21% oxygen, 78% nitrogen, and 0.04% carbon dioxide. In a salad bag, the mix is flipped over to put the leaves into “hibernation”. The target oxygen level is set to 1-5%, the target carbon dioxide level is set to 5-15%, and the remaining is nitrogen, used as an inert filter.

This process puts the lettuce to “sleep.” It prevents the leaves from consuming their own sugar reserves too quickly. It slows respiration and prevents spoilage bacteria from growing.

The specific environment delays lettuce oxidation, adding 4-6 days of shelf life to the packaged salad bag.

Modified Atmosphere Packaging extends shelf life by 4-6 days.

In the early 2000s, packaging moved from simple plastic to laser-perforated films. The bag is a semi-permeable membrane with laser-created holes. Lasers are used to poke holes in the plastic. The specific respiration rate of the salad mix determines the size and frequency of these holes.

For example, the respiration rate of baby spinach is very high. So the perforation strategy requires a high density of larger holes to allow oxygen to be consumed and constant venting to prevent fermentation. The holes in the bag can be as small as 50 to 200 microns.

Romaine or Iceberg Lettuce has a lower respiration rate than spinach, so the perforation strategy calls for a low-to-moderate hole density. The lettuce leaves are structurally tougher and so breathe slowly, whereas too much “air” will cause the cut edges to turn pink.

The perforation strategy adds another 2-3 days to shelf life.

Then we eat

The packaged salad travels through the cold chain to the grocery retailer or wholesaler’s distribution center. It ultimately ends up on a grocery store shelf, in a restaurant kitchen, or in a food service cooler, before it is consumed. At the same time, its freshness is still intact by consumers like you and me.

So the next time, when you are at your desk eating a healthy packaged salad, as one of the 251 million Americans who consume them, and getting ready for your next video call in 12 minutes, take a moment to thank all the technology that made it possible to get the product to you a few thousand miles from where the greens were grown and harvested.

The video shows how a packaged mixed salad is processed and packaged. It shows vacuum cleaning, vision and robotics systems, triple washing, drying, and packaging using specialized bags with pores.

Note: I do want to acknowledge that even though packaged salad is mostly healthy and safe, sometimes we do see recalls due to various concerns. You can look up all FDA recalls and safety issues here.

I want to thank Mike Riggs, Venkatesh Ranjan, and Abby Shalek-Briski for their feedback.

About 150 years later (today), less than 200 miles north of the Chincha islands, ships line up to get into the brand-new deep-sea, highly automated Chancay port, which opened in November 2024.

The stevedores of Baltimore from “The Wire” would have been totally against this port.

The ships are there to carry minerals, industrial goods, and, most importantly, agricultural outputs such as blueberries, avocados, grapes, and mangos. Most of the cargo is headed to China and other parts of Asia.

Not only is the Chancay port new, but some of the products in those containers, such as blueberries, are relatively new exports for Peru. In fact, Peru didn’t produce and export many blueberries till about 10-15 years ago.

The growth of blueberry exports is a good story of what can happen when public and private enterprises come together, build critical infrastructure, enable the right policies, and provide support for the right technology.

Counter seasonality

About 15 years ago, if you went to your local grocery store to buy fresh blueberries in October or November, you would have a very tough time finding them. This is due to the fact that the main blueberry growing regions of the world, the United States (mainly California), Europe (Poland and Portugal), and Chile, did not harvest many blueberries during the September through November time period.

So, if you could grow blueberries such that they could be harvested during September through November, and you could have the logistics infrastructure to get the blueberries in different parts of the world in a timely manner, you could command high prices during that time period.

You can be the only game in town for a quarter of the year!

This is exactly what Peru has achieved. But to begin with, blueberries are not native to Peru.

The conventional wisdom was that blueberries could not grow in Peru because the plant typically requires “chill hours” (prolonged periods of cold weather) to flower, which the Peruvian coast lacks.

But industry pioneers began experimentation in the middle of the ‘oughts. They brought different varieties to Peru to test if any could adapt to the coastal desert climate.

The industry discovered that certain varieties (most notably Biloxi) could thrive in the mild climate of the Peruvian coast without requiring deep cold. They treated the crop almost like a tropical evergreen rather than a temperate bush.

By managing water and nutrition precisely, growers realized they could “force” the plants to flower and fruit at specific times, allowing them to control the harvest window.

Because the weather is stable, Peruvian growers can prune their plants to ensure they produce fruit exactly when prices are highest globally (September–November), the window when the US and European harvests are finishing, and the Chilean harvest hasn’t fully started yet.

This gives Peru’s blueberry business a unique advantage, assuming they have the supporting infrastructure of reliable demand, logistics, and irrigation infrastructure, the ability to experiment to be able to realize and capture the most customer value, and a policy framework that allows the business to thrive and grow.

As we will see soon, all of these factors came together to make Peru a blueberry powerhouse in less than 20 years.

The Usain Bolt of Blueberries

One of the challenges in establishing a new blueberry crop is that a blueberry bush may take 3-4 years to reach full production, especially in traditional regions such as the United States and Canada.

But in Peru, due to the climate and genetics, blueberry plants can begin producing commercially in their first year.

Peru blueberry bushes grow fast and reach harvest maturity in just over half a year. As in the US, in Chile and Argentina, they grow in three years. The huge difference in growing time accelerates both productivity and learning. Peruvian firms can conduct 3-4 times as many experiments with different varieties and crop management techniques.

In Peru, because there is no winter dormancy, a plant enters commercial production in its first year. This allows investors to get their money back much faster. The short time-to-cash flow with blueberries (assuming you have demand and a logistics network) allows for lower investment and less deep pockets to experiment with blueberries in Peru, compared to other parts of the world.

This accelerated learning process cannot be underestimated. The Peruvian coast is an arid desert with stable temperatures and acts as a natural open-air greenhouse. This allows Peru to harvest blueberries year-round, though it typically targets specific windows to command the best prices.

Stable temperatures in Peru provide greater control over “programming” and demand locking-in, thanks to their reliable supply. For example, if Peruvian growers want fruit in October to fetch high prices, they prune blueberry plants 5-6 months before harvest. This reliability (not available in California because weather dictates harvest timing) allows Peruvian growers to sign large, guaranteed contracts with major grocery retailers.

Share

Blueberries are labor-intensive (hand-picked). Peru has access to a large agricultural workforce in the north at much lower prices than in California (La Libertad/Lambayeque). In California, finding enough workers during harvest is a chronic crisis, often forcing farmers to leave fruit on the bush or invest in expensive mechanical harvesters.

Genetics to the rescue

Because the Peruvian industry is young (circa 2012), most of their fields are planted with modern, high-yield varieties. California has many older fields with legacy varieties that are less efficient and have softer fruit.

In terms of fruit profile, Peruvian growers prioritize blueberries that are vigorous, firm, extra-large, sweet, less acidic, and crunchy. These qualities make the berries ideal for international markets. Maintaining the in-field characteristics and quality of berries from production to consumption requires specialized genetics and logistics.

Companies like Camposol, Inka’s Berries, and Agrovision (now called Fruitist) did multiple trials to determine the best Chilean blueberry varieties suited for Peru’s climatic conditions. With its counter-seasonality advantage and fast-growing berries, Peru has captured a large share of the international export market. Now, Peruvian growers are working to increase value by developing more innovative products through genetic research and identifying new varieties.

For example, newer varieties like Ventura are larger and sweeter than Biloxi, yield fruit earlier, and allow growers to hit the high-price window in August and September. Super varieties like Sekoya Pop (and Sekoya Beauty) prioritize a “crunch” texture, a long shelf life (45+ days, which is vital for shipping to Asia), and a jumbo size.

Infrastructure and Policy Support

Despite the chronic political instability in Peru (8 presidents in the last decade), the Peruvian state has provided the necessary infrastructure and policy support for the blueberry business to flourish and for Peru to become one of the world’s largest exporters of blueberries.

One of the architects of the policy framework is Pierro Ghezzi, the former minister of production for Peru. You should see his January 2026 interview with Vox Dev. He mentioned four key policy pillars that underpinned the agro-export boom along with logistics.

Pierro Ghezzi highlighted the role of “Mesa Ejecutivas” (Executive Roundtables or ERs) in the Global Policy publication and the Inter-American Development Bank article, “Executive tables in Peru: A technology for productive development.”

Latin American countries face the challenge of sustainably increasing their productivity. Changes in production methods toward modern methods across various sectors, including natural resources, and in private-sector regulations and standards, offer opportunities.

These increased coordination needs cannot be met with standard public policies or traditional industrial policies. A modern industrial policy is required, one that must be a process of strategic public-private collaboration. It must include learning, experimentation, coordination, monitoring, evaluation, and correction.

The main objective of the ERs (Executive Roundtables) is to identify the constraints that limit the productivity of a sector or a factor and to implement solutions to eliminate them. Their focus is on execution, not on generic dialogues about competitiveness. But experience shows that execution provides a clear direction.

The Executive Roundtables (Mesa Ejecutivas) provided a key unifying and continuous structure to push through many policy changes and provided continuity despite a volatite political situation.

Free-trade agreements

Peru established a free-trade agreement with the United States in 2009. Peru also has free trade agreements with the EU, UK, ANZ, Mercosur countries, Japan, Mexico, China, and many other countries. In return, Peru permits certain types of foreign direct investment.

The trade policy has ensured that if Peru can provide high-quality products at competitive prices, there will be a market for them in different parts of the world.

Water in the desert

The coastal desert of Peru receives almost zero rainfall (less than 10mm per year in some places). The coastal desert is not suitable for agriculture without human intervention to bring water there for irrigation. On the east side of the coast, the mighty Andes are snow-capped, and rivers flow out of them, but no major river comes to the desert. The idea of drawing water from the Santa River and creating north-south canals was conceived in 1912. Construction began in the 1960s, and Phase 3 began in 2012.

It takes water from the Santa River (which flows down from the Andes) and diverts it through a massive 150km+ network of open-air canals and tunnels that run north along the desert coast.

In another ambitious project (Olmos Irrigation Project), which was initially conceptualized in 1924, but real work started on it in 2006, brings the waters of the River Huancabamba through a 20 km long Trans-Andean tunnel through the Andes. Once the first phase of the tunnel project was operational, the scheme supplied more than 2 billion cubic meters (500 billion gallons) of water annually to irrigate 130,000 acres of farmland.

The project is an engineering marvel. It takes water from the Huancabamba River (which naturally flows east toward the Amazon/Atlantic), captures it with the Limón Dam, and forces it through the tunnel to the dry Pacific side. (as shown in the schematic below).

These two massive public irrigation projects created large, efficient, industrial-scale plots that are perfect for mechanical harvesting.

Phytosanitary

When it comes to fresh produce, you need to prove to your buyer (especially in the case of exports) that your product is not going to bring pests to their country. There is a reason why, as individual passengers, we are not allowed to carry fruits and vegetables across borders. One time, a friend of mine was fined $600 for bringing a single mango into the United States!!

Peru’s government agency, SENASA (National Agriculture Health Service), played a key role in certifying and enforcing standards to assure buyers that Peruvian fruit is free of pests. Initially, Peruvian growers used methyl bromide asa chemical fumigant for post-harvest treatment. Fumigation affects fruit quality, including softness and reduced shelf life.

When Peru finally struck a deal with China, it followed the temperature-controlled cold chain method of storing and transporting blueberries at a temperature below a certain threshold for 3 weeks. This kills pests without fumigation but requires a highly efficient and reliable cold chain.

Lately, Peru has been using a systems approach to certify fields instead of the final product. If the fields are shown to be pest-free, the fruit can be exported as natural, without any treatment. This allows them to compete for higher-value products, such as the organic, premium, crunchy berries category.

Peru’s special efforts on phytosanitary have given it the freedom to operate in various regulatory environments and have given the authorities and consumers confidence in Peruvian blueberries (and other products)

Sector-specific agrarian law

The Peruvian government enacted the Agrarian Promotion Law in 200, which reduced the income tax rate for agricultural enterprises to 15%, accelerated depreciation for agricultural investors, and made the sector attractive to investors. This helped fuel additional investment, provided more time for experimentation, and catalyzed the development of Peru’s blueberry sector.

The law was modified in 2021, raising the income tax to 30%. These modifications could lead to a decline in future investment.

Subscribe now

Logistics infrastructure

As I said at the beginning of the piece, Peru opened a brand-new deep-sea port in Chancay by utilizing financing and support from China’s Belt and Road Initiative. The port has reduced the shipping time from Peru to China from 35-40 days to just 3 weeks. This significant reduction allows Peru to ship fresh berries (with limited shelf life) to China by water rather than by air freight.

Peru’s ability to move perishable products quickly gives it a competitive advantage. The Pan-American Highway, with its cold-chain supporting infrastructure and farm-level vertical integration through large growers, can move fruit from farms to ports quickly and efficiently without sacrificing quality.

Peru’s continuous harvesting capability and flexibility give it a competitive edge over countries reliant on seasonal farming. Here, the Biloxi variety stood out. Optimal for production from late August to early December, Biloxi blueberries had a key market advantage: they avoided direct competition with northern hemisphere supplies (harvested May-August).

A perfect storm for rapid growth

One way to think about the growth in blueberry exports from Peru is to think that it happened in less than 20 years. Another way to think about it is that it took almost 100 years for some of the irrigation projects to come to fruition.

One thing is clear that for the Peru’s blueberry industry to take off, it needed many factors to come together at the same time to take advantage of the counter-seasonality possible for blueberries, whether it was to find the right markets, public-private partnerships, logistics capabilities, policies for product safety and reliability, favorable tax and investment environment, and an irrigation infrastructure built over decades.

Other growing regions can take lessons from Peru’s blueberry adventure. A large-scale positive change requires multiple factors, such as technology, policy, trade, and infrastructure, to come together.

If you want to bring about large-scale change, you need to identify your edge (like Peru did with counter-seasonality for blueberries) and work with the ecosystem to catalyze it and drive positive change. These changes don’t happen overnight and can take 15-20 years (or a few decades).

Share Software is Feeding the World

“Flooding” in the Yuma desert is quite common.

The waters of the Colorado River, the irrigation infrastructure, Yuma’s senior rights for the water of the Colorado River, abundant sunshine 330 days a year, good soil, and advances in technology created a perfect storm for Yuma to become the winter salad bowl. The Colorado River’s water availability may become challenging in the future, potentially triggering changes in Yuma’s agricultural landscape. This week’s newsletter goes back in time to the history of Colorado's waters, how things are changing, and ultimately what it means for AgTech innovation in the future. Last week’s newsletter focused on the automation of lettuce harvesting in Yuma, and it is a good companion read with this week’s edition.

I recently spent 2.5 days in Yuma. I landed in Phoenix early in the morning, and we drove to Yuma, about a 3-hour drive.

You leave the urban landscape of Phoenix behind you and keep going down from an elevation of 1100 feet to 150 feet as you go to Yuma Valley.

You encounter the massive Saguaro cactus, which looks like Poseidon’s giant tridents towering over the landscape. It is ironic, given Poseidon’s trident symbolizes power over oceans and waves, but the landscape is arid and moonlike.

Driving over the Gila Mountains and Telegraph Pass (about 20 miles east of Yuma), as you drop into Yuma Valley, the stark desert is replaced by a massive, lush green grid of industrial agriculture (lettuce, leafy greens, etc.).

The waters of the Colorado River and the amazing irrigation infrastructure built over a few decades are one of the main reasons to turn this “desert” into the winter salad bowl of the United States.

According to the Arizona Water Factsheet, Yuma County, (November 2025)

Nearly all surface water in Yuma County comes from the Colorado River, which is the only source of water for communities located adjacent to the river.

Colorado water is used by 7 US states and Mexico. How the water of the Colorado River is divided up was decided by the Colorado Compact in the 1920s.

It is important to understand the history and numbers behind the Colorado Compact, the assumptions it makes, how it has affected Yuma agriculture, how it will impact the future of Yuma agriculture, and what it means for AgTech innovation.

The Colorado Compact

In the early 20th century, the seven southwestern and western states of Arizona, California, Colorado, Nevada, New Mexico, Utah, and Wyoming created an agreement to manage the river’s water by dividing into the Upper Basin (CO, NM, UT, WY) and Lower Basin (AZ, CA, NV). The agreement was established in 1922 and is called the Colorado River Compact.

This agreement and subsequent negotiations dictate how water is shared among these states and Mexico (because the Colorado River flows into Mexico and into the Gulf of California).

One of the biggest issues in setting up the Colorado Compact and in how water is shared between states was an overestimation of how much water the Colorado River carries over the long term.

The agreement used data on water flow from 15-20 years prior to the contract to determine how much water would flow to the Upper and Lower Basins.

The fly in the ointment is that the 15 years preceding the agreement were among the wettest on record, so it overallocated 7.5 million acre-feet to the Upper Basin and 7.5 million acre-feet to the Lower Basin states, for a total allocation of 15 million acre-feet.

The data for the 15 years preceding the 15-year period used for the agreement showed a total water flow of only 13.5 million acre-feet, and the long-term average for the Colorado River is closer to 13.5 million acre-feet and is declining.

According to the EOS report,

“The allocation of 7.5 million acre-feet per year of consumptive use for each basin was grounded neither in the best available hydrologic calculations nor in climate variability projections. Rather, it was a compromise Hoover proposed between two endmember figures [Kuhn and Fleck, 2019]. One end was 8.2 million acre-feet per year, half the assumed annual average discharge at Lee Ferry of 16.4 million acre-feet per year, which itself was derived from a report by the U.S. Reclamation Service (now the Bureau of Reclamation) [Fall and Davis, 1922]. The other end was 6.5 million acre-feet per year, a figure proposed and advocated by the Upper Basin commissioners that reflected a roughly 50-50 split of the river discharge at Yuma [Kuhn and Fleck, 2019].

In the decades following the 1922 compact, a plethora of acts, orders, and agreements were written and signed to fine-tune the compact’s provisions, to authorize construction of dams for water storage and power generation, to build water transfer infrastructures, and to resolve interstate and intrastate disputes. Particularly significant was the 1944 treaty between the United States and Mexico—still in effect today—that guaranteed 1.5 million acre-feet of Colorado River water annually for Mexico, bringing the total allocation to 16.5 million acre-feet per year.”

Hydrologist Eugene Clyde La Rue, who had studied the Colorado River for quite some time, argued that decision-makers should use longer- term averages in estimating river discharge, which were not used.

The amount of water flowing down the Colorado River has decreased over the years (or fluctuated significantly), creating interesting issues for water allocation under a priority-based access system. There is also an issue with the allocation of water to tribal homelands.

Subscribe now

According to Wikipedia,

Since 2000, dwindling Colorado River flows and consumption rates in excess of natural replenishment have provoked a need to plan for shortages. Because of its lower priority, shortages should fall primarily upon the users of the Central Arizona Project, the consequences for some Arizona farmers and urban sectors could be dire. If the federal government acts unilaterally to restrict other water uses, it risks litigation. For these and other reasons, river users are exploring voluntary measures to address shortages.

The Colorado Compact does not have an end date, but some of the interim agreements negotiated in 2007 are set to expire in 2026 (this year!). The Bureau of Reclamation is leading a process to develop long-term management plans, with a draft Environmental Impact Statement released in January 2026, aiming for new guidelines by October 2026 before the next water year begins.

It is safe to say that access to and allocation of water for agricultural irrigation (which accounts for 97% of Yuma's freshwater use) could be a challenge in the future.

Gravity is free

Based on the reading so far, one might not be wrong to think that, given the water challenges in the present and the future, Yuma might be moving to the most efficient irrigation system in the world: drip irrigation.

But changing a particular practice can have a cascading effect.

The whole lettuce growing business is a finely choreographed dance. A five-degree difference in temperature is used to start planting in one part of Yuma Valley, and as temperatures go down further, planting spreads to other parts of the valley.

The irrigation method used during the early part of crop growth depends on whether the crop is direct-seeded (seeds sown directly into the ground) or transplanted from another location. These practices have been refined over many years and are closely intertwined.

So if you come in with a technology solution that changes one practice, you have to really understand the impact of that change for this finely tuned system, you will end up with solutions that do not work or cause adoption challenges.

Coming back to irrigation, leafy green production (lettuce, spinach, kale) in Yuma uses a specific hybrid irrigation strategy that surprises many people because it relies heavily on “old school” flood irrigation methods that have been optimized with high-tech precision.

They typically do not use drip irrigation throughout the crop's lifecycle. Instead, they switch methods halfway through the season.

Most lettuce fields in Yuma follow a strict two-phase irrigation schedule, especially for direct-seeded crops. During the germination phase, aluminum pipes with sprinkler heads are laid out by hand across the field immediately after planting seeds. The sprinklers cool the soil and protect the seed. It ensures that every inch of the seedbed gets wet, to guarantee a uniform “stand” of lettuce

Once the plants are a few inches tall (after “thinning”), the sprinkler pipes are removed. Water is then sent down small trenches (furrows) between the rows of crops. The water comes from a network of open canals that start at the Colorado River, and water allocation is controlled by the local water districts. Growers siphon out water from the canal through movable pipes and send it down furrows.

The Colorado River is salty. Sprinklers can leave salt crusts on the leaves as water evaporates. Furrow irrigation pushes those salts down below the root zone to prevent salt burn.

Ideally, the water seeps sideways into the roots without ever touching the leaves. Yuma has heavy alluvial soil that holds water well, allowing it to flow down a furrow without sinking immediately. This is important for food safety, as the water stays on the ground rather than on the leaves. It minimizes the risk of pathogens touching the edible part of the plant.

Even if Yuma could be looking at water issues, drip irrigation might not always be the answer, due to another “green” commodity - dollars!

Lettuce is a fast crop (60–90 days). Installing and removing drip tape for every rotation is incredibly expensive and labor-intensive. Drip irrigation can push salts to the edge of the wet zone (right where the plant roots are), whereas furrow irrigation flushes the salt deep down.

A big expense would be the energy required to make drip irrigation work. The energy requirements can make drip irrigation quite expensive, whereas furrow irrigation (flood irrigation) does not require energy.

As one grower said,

Even though furrow irrigation looks primitive, Yuma farmers have made it about 80–90% efficient using Laser Leveling. Before planting, GPS-equipped tractors and laser systems grade the fields to be perfectly flat or have a microscopic slope.

This level of precision ensures water flows at “Goldilocks” speed. The flow is fast enough to reach the end of the row, but slow enough to soak in evenly without pooling or running off.

So even though the cost of water is lower with a drip system, the setup costs (PVC mainlines, plastic drip tape, etc.), energy costs, and maintenance costs increase the total cost of ownership and operations for a crop like leafy greens.

Having said that, due to water challenges, growers have been shifting the crops they grow and adopting efficient methods such as laser leveling, newer genetics, and fertilizers to become more water-efficient. Modern lettuce varieties grow faster and are denser, requiring fewer irrigation runs due to a shorter season.

For example, the water fact sheet for Yuma County from the University of Arizona says,

Water used for irrigation in Yuma County has decreased 15% since 1990 based on a reduction in irrigable acres, expanded use of multi-crop production systems that require less water, and significant improvements in crop and irrigation management and infrastructure. The number of acres planted to vegetables has increased nearly 6x over the past 40 years, while acreage committed to full-season crops (citrus, cotton, sorghum, and alfalfa) has declined 43%.

Given the context of the water situation, climate change, and labor challenges (which we covered last week), one might be tempted to focus heavily on reduced water use, whereas the challenge is how to allocate water (which itself might be changing) most efficiently.

Given this context, what types of solutions should the AgTech ecosystem consider for Yuma?

Subscribe now

What does it mean for AgTech?

AgTech innovation in Yuma must pivot from a pure optimization mindset to one of resilience, embracing diverse, evolving farming practices that adapt to new technologies and a harsher environmental reality.

Genetics

As I said earlier, climate change and heat might shrink the growing window for Yuma. If geneticists can develop heat-resistant crops, they can help protect or extend the shoulder season. Expansion of the shoulder season helps protect or restore profitability lost to extreme weather events or climate change.

Any technology that can reduce field temperatures during the shoulder season could also protect or expand the growing window in Yuma, potentially increasing profitability.

Also, there is a salinity time bomb over a reasonably long horizon, as the Colorado River is getting saltier. It might require genetically modified crops designed to tolerate higher salinity water.

Photovoltaics

As and when water becomes expensive, the sun might become one of the cheapest assets of Yuma. The integration of solar panels over crop canals and fields can reduce evaporation, and transpiring plants can cool the solar panels. The development of lightweight, translucent solar arrays that let through the exact light spectrum plants need for photosynthesis while blocking the heat-generating infrared rays.

Tracking salt and heat

Current soil moisture sensors can measure soil moisture, but it is difficult to map and monitor salinity in a field. Cheap sensors that can track EC and canopy temperature can help growers make better decisions.

Water accounting

As the government starts paying farmers not to farm, there is a massive need for technology that verifies that water was actually conserved. AgTech companies that can become the QuickBooks of Water could be quite valuable.

Automation to address labor issues

As I said last week, there are challenges with labor access, and labor costs are rising. The labor force is also aging and shrinking. Crop robotics for weeding, thinning, and, most importantly, harvesting will be key areas for the AgTech ecosystem. We might also have the need to breed “robotics-ready crops”.

For example, lettuce varieties that grow on longer stalks or have uniform shapes designed to be grabbed by a machine, even if they look slightly different than what consumers want.

Connectivity infrastructure

It will be difficult to implement automated, autonomous systems on farms without connectivity, which is a challenge in Yuma. Yuma is tackling that problem head-on by building a “middle mile” fiber-optic network to provide high-speed internet access at the farm and field levels. With the ease of creating “edge computing” devices, connectivity becomes even more important to help control and manage remote equipment.

Share Software is Feeding the World

Conclusion

The agricultural landscape of Yuma is a complicated web of historical, social, political, cultural, economic, environmental, and technological issues. The Colorado River, the amazing irrigation infrastructure, and the resilience and ingenuity of growers have created the “winter salad bowl.”

The complexity is similar to that of other agricultural regions, but it has its own set of considerations. As the underlying factors change, the agricultural business built on them will have to evolve over the coming years.

Technology will play a key role in the process.

I had the privilege to visit multiple harvesting operations during my visit to Yuma, Arizona, last week. I also had the chance to connect with other players in the ecosystem who have worked on various aspects of the problem.

These included forward-thinking Farm Labor Contracting (FLC) companies, local manufacturers, AgTech adoption services companies, research institutions, and, of course, many growers, shippers, and packers. Just as with any population, I encountered people with extremely strong opinions about how things should be done!

Harvesting lettuce (and many other specialty crops) is still very much a majority human-driven endeavor.

A brief history of automation and mechanization attempts

According to a 1973 research paper, serious attempts to mechanize lettuce harvesting at public research universities began in 1962. We have been trying to fully automate lettuce harvesting for almost 64 hours years, but we have not succeeded yet.

And it is not for lack of trying.

There are multiple startups and other custom shops working to automate this process.

According to the latest crop robotics report from Better Food Ventures, there are more than 450 robotics companies, with many focused on harvesting operations.

The Bracero Program - Farm Labor in the 1950s-60s

US Growers, shippers, and packers prior to the 1960s benefited from the Bracero program.

This was a government-to-government agreement. The US government acted as the labor contractor in the early years, transporting workers to distribution centers. It was massive in scale and focused almost exclusively on Mexican nationals.

Growers paid low wages, and the program itself was accused of suppressing domestic wages. There was little financial incentive to replace Braceros with machines, since labor was so affordable.

The Bracero program was terminated in 1964. This created a huge challenge for growers, unless they had access to undocumented workers.

Right after the program was terminated, UC Davis began commercializing a tomato harvester for processed tomatoes, which was quickly adopted. Some acreage for labor-intensive crops, such as asparagus and strawberries, moved south to Mexico.

The Bracero Program ended in 1964, and many farmers expected to continue to hire Mexican workers under the H-2 program. However, the DOL issued regulations on December 19, 1964, that required employers of H-2 workers to offer and pay the AEWR to any U.S. workers they employed (Martin, 2007, pp. 15–16). Unlike bracero employers, H-2 employers were also required to offer and provide U.S. workers with the same housing and transportation guarantees that were included in bracero contracts. At a time when U.S. farm workers were not covered by minimum wage laws, most U.S. farmers adjusted to the end of the Bracero Program by mechanizing or changing crops rather than continuing to rely on migrant workers (Clemens, Lewis, and Postel, 2018).

The post-Bracero period saw many attempts to mechanize lettuce harvesting. But these early machines were heavy, slow, and often crushed the lettuce. The technology of the time (analog sensors and hydraulics) lacked the finesse of a human hand.

With the rise in undocumented workers, the pressure to automate harvesting decreased a bit.

From the 1980s to the 2010s, the industry paused on full automation and switched to doing “harvest aid” machines.

The harvest aid machine typically consists of a slow-moving tractor that pulls a long conveyor belt wing. A crew of about 20 people walks behind the conveyor belt. They continue to cut, trim, and place lettuce heads on a belt.

On the right-hand side of the picture, you can see about 20-25 people selecting, cutting, and cleaning lettuce heads, then placing them on a conveyor belt that moves from right to left.

There are about 3-6 people at the end of the conveyor belt who are responsible for quality-checking the lettuce. The QC people are called “ojeros” as they keep an eye on the quality of produce coming through.

If the field has a high-quality product, the harvesting crew will have more cutters and fewer cleaners. If the product quality in the field is not top-notch, the harvesting crew will have more QC personnel and fewer cutters.

For shed pack products (as seen in the picture above, which will be used for processing, the lettuce heads are kept clean through a spray of water and then end up in large bins (left of the picture), which are then immediately moved to a cooling facility for further processing for salads, etc.

The two or three people standing by the bins make sure the bins are filled uniformly. This part of the harvesting crew is called “reinas” (queens, as they stand and work higher up on the machine).

For field packs, the lettuce heads are immediately wrapped in plastic and placed in boxes for shipping to a cooler.

Harvest aid machines reduce physical strain on workers by reducing walking and bending (depending on the product) and increasing harvesting speed.

The most critical process in harvesting lettuce requires the ability to spot a mature head, cut it at the right angle and at the right spot, and trim the bad leaves. The robotics capabilities of the 1990s and 2010s didn’t have the capabilities to replicate it cost-effectively, whereas a trained human can do it quite effectively.

(This should quell the notion that farming is an unskilled job)

As a result, human harvesting still accounts for more than 50% of the total production cost for specialty crops like lettuce and berries.

The University of California, Davis, publishes detailed cost breakdowns by different crop types in different parts of California. The table below shows a screenshot of the total operating costs per acre for field-packed iceberg lettuce.

I have highlighted the harvest/field pack costs of $7,200 per acre, out of the total operating costs/acre of $14,584. This data is for 2023 for iceberg lettuce in the Central Coast region of California. Labor costs will be slightly lower in Yuma, Arizona, but the percentages will remain the same.

H2-A program - a new far labor program from the 1990s

The H-2A program began in 1986, though the first visas were not issued until 1992. The H2-A program's goals and structure were in sharp contrast to those of the Bracero program.

After the Bracero program was terminated for suppressing wages, the Department of Labor (DOL) implemented strict rules for H-2 visas. They created the Adverse Effect Wage Rate (AEWR).

AEWR is a minimum wage floor designed to prevent foreign workers from being paid less than US workers.

For example, today in Arizona, the hourly legal rate for an H-2A worker is around $18, and they are required to have housing and other amenities, which bumps the hourly rate to $25 to $30 per hour.

For the first 20-25 years, the H2-A program was hardly used, as it was cheaper and easier for farmers to hire undocumented workers. As a result, in the first 20-25 years of the program, H2-A issued only 30,000 to 40,000 visas nationwide each year.

Around 2010, due to stricter border enforcement and demographic changes in Mexico, the cheap, undocumented labor supply started to dry up. It forced farmers into the H-2A program as the “labor of last resort.”

As H2-A program costs have risen, it has once again created a financial incentive for growers to automate as much of the harvesting process as possible to reduce costs and stay competitive with other parts of the world, such as Mexico and South America. My friend and fellow AgTech Alchemist, Walt Duflock, has written extensively about these challenges, and you should follow him on LinkedIn.

Subscribe now

As you can see from the images below, harvesting lettuce requires speed, keen observation, precision, and stamina. Harvesting lettuce across Salinas, California, and Yuma, Arizona, is a 6-day-a-week, 52-weeks-a-year, around-the-clock operation.

Humans are very dexterous and versatile in their skill sets. Picking romaine or iceberg lettuce is not straightforward.

For example, lettuce picked for shed packing for value-added products like salads has to be cut at the right angle, a few leaves have to be removed, and the core is stabbed with a coring knife.

This lettuce is then put into large bins, which are sent to the cooler for further processing.

For a product like romaine heart lettuce, the human has to discern which lettuce to pick and which to leave behind, cut it at the right spot, and then remove the right amount of leaves to leave a beautiful-looking heart, which is then immediately packaged.

This is a bit of a consumer expectation problem. Consumers expect beautiful-looking produce that stays fresh for at least a reasonable amount of time in their refrigerator before it is used.

As a result, a large amount of potentially usable product is left behind (by some estimates, 40-50%). This is plowed back into the ground before the next crop goes in, providing some nutrients to the soil.

As you can see from some of the examples here, it is challenging for a machine to completely replace a human, as it must match the speed, quality, and cost of the human crew. The harvesting crew has to perform the following three main operations for lettuce.

Cutting, cleaning, and packing.

Share

Harvesting process capabilities and requirements

There are 5 different buckets of capabilities that an automated machine should be able to meet to reduce the amount of human labor used or almost eliminate it.

These required capabilities will illustrate why, even though the ecosystem has been working for more than 64 years, we still have not fully automated lettuce harvesting.

Though with powerful vision systems, sophisticated AI algorithms, edge processing, advances in materials engineering, and rising H2-A costs, it might be feasible to address this problem.

Performance and Throughput

The system must match or exceed the efficiency of manual labor crews to justify the capital expenditure. Current crews work fast and move through fields quickly, easily processing thousands of heads per hour.

The crew has to maintain the quality of the cut, the cleaning, and the packing while maintaining a high throughput.

As I said earlier, lettuce harvesting happens 6 days a week, all year round. The system must be capable of near-continuous operation during the harvest window of April–November in regions like Salinas, CA, and November through March/April in Yuma, Arizona. The changeover time between fields has to be kept to a minimum.

Economic and Commercial Viability

The machine must reduce the human workforce requirement over a longer time period, and it should have a reasonable and acceptable ROI timeline for adoption.

Obviously, any new machine will not meet these requirements on Day 1, and so there should be a plan to go down the cost curve by learning with enough reps in the field.

Food Safety and Hygienic Design

The machine must adhere to the Produce Safety Rule (PSR) under the Food Safety Modernization Act (FSMA). The Produce Safety Rule has specific requirements around agricultural water, biological soil amendments, sprouts, domesticated and wild animals, worker training and health and hygiene, and equipment, tools, and buildings.

Food safety is absolutely critical. All the operations I saw took food safety very seriously. We always had to wear gloves, hair nets, and beard nets (whether you had a beard or not)!

Machines are washed, and surfaces are cleaned regularly. Everyone has to constantly wash their hands to maintain a hygienic environment.

One of the growers had a pet peeve about birds! Birds can be a nuisance as they can damage the crop. If birds do their “business” while flying over the field, it makes harvesting the product challenging from a food-safety standpoint.

I heard stories about how 100s of acres of crops had to be abandoned due to excessive bird excrement! This particular grower had hired an 8-person crew with the sole responsibility to scare the birds away. Innovative attempts to scare the birds away include drones chasing birds, specific frequency sounds, etc.

Product Quality and Grading

The cut must be made exactly at the “collar” (where the leaves meet the stem) to prevent a reduction in shelf life. The automated system must identify and reject heads with visible defects. Mechanical damage should be kept to a minimum and meet USDA standards for different grades of produce.

Operation and Environment

Meeting operational and environmental requirements is one of the most challenging aspects for these machines. The harvesting environment and the time period are characterized by mud, dust, and high heat.

Furrows might be wet and muddy. Machines have to turn around at edges, which may have different characteristics. It should be able to enter a field and be easy to transport from one field to the next.

Growers are very passionate about how they grow their crop. Some growers are extremely passionate about their bed width, and it would be very hard to get them to change it if it meant an easier operation for their harvesting machine.

Due to this, a harvest machine should be flexible and able to adjust to different bed widths (ranging from the low 30s to mid 80s inches).

Ecosystem Support

For any product, especially a startup, it is impossible to get all of these right in the first version. The product development process has to start with a particular capability and then build experience in the field through a real, growing operation.

Given the number of startups working on robotics, it is often challenging for startups to reach the right type of growers to get feedback on their products and then have the infrastructure to test and iterate on their product.

This is where the ecosystem comes into play. I had the privilege to connect with a few public and private players in the space in Yuma.

For example, YCEDA (Yuma Center of Excellence for Desert Agriculture) under the University of Arizona umbrella is

dedicated to bringing scientific research and industry together to find solutions that bring value to stakeholders by addressing “on-the-ground” needs of Desert Ag production.

YCEDA members can provide technical and research assistance, field acres for testing, and ecosystem support to startups and other players working on bringing innovation to market.

An organization like The Reservoir provides support, office space, field testing infrastructure, and grower connections.

Reservoir helps founders build, test, and scale technologies that farmers can put to work—faster, fitter, and with lasting impact.

The Western Growers Association serves as a voice for specialty crop growers and provides important connections, expertise, and perspectives on technology adoption, policy, and industry needs.

A private organization like Axis Ag provides “Technology services for the agricultural community.”

Axis Ag empowers farmers with innovative agricultural solutions that drive sustainable growth, optimize productivity, and foster environmental stewardship. We are dedicated to advancing technology and partnerships that enhance the future of farming, building resilient food systems, and supporting the communities we serve.

Axis Ag has an interesting model. They have years of experience in specialty crop agriculture, strong grower connections, and the expertise to field-test different AgTech solutions, provide feedback to startups, and find the right set of early adopters to test and scale them.

They have partnered with multiple AgTech companies to help them bring their products to market and to scale.

The services and support provided by YCEDA, The Western Growers Association, The Reservoir, Axis Ag, and many others are critical to bringing these very much needed AgTech innovations to market and scaling them.

The normal conversations within AgTech can feel depressing. The farm economy is definitely in trouble, but my trip to Yuma, Arizona, showed me that the AgTech ecosystem is working hard, is collaborating, and is focusing on solving the right type of problems with rigor.

It might take longer than sending a human to the moon, but many of the existing problems, will get solved in the near future!

Next week’s edition

I will write about the amazing irrigation infrastructure in Yuma, Arizona, how Yuma became the winter salad bowl of the US, the challenges the region faces, and the role AgTech can play.

Happenings

I will be leading a two-and-a-half-hour session during the “AI in Agriculture Forum” at World Agritech San Francisco on March 16, 2026. The focus of the session will be on AI benchmarking. I hope to see many of you in San Francisco in March.

I will be hosting a session titled Robotics & Autonomy: The New Agricultural Workforce on March 17th, 2026, at World Agritech, featuring speakers from CNHi, Kubota, Tavant, and AgZen.

AgTech Alchemy will host an event in Monterey, California, on February 17th, in collaboration with AgSafe. The AgTech Alchemy team will share more details shortly.

There have been a few changes at SFTW over the last few weeks. The paid tier is disabled, and all past and future content is available for free to everyone.

As part of the move, I have also migrated from Ghost to Substack (though the old Ghost site is still up), as Substack is free if you are not charging your users. No change is required on your side. You can access all previous content on SFTW.

2025 has been a tough year for Agriculture and AgTech, to say the least.

The issues continue to be complicated, multi-faceted, and often (or always) there are no silver bullet solutions.

My last agriculture-related conversation in 2025 brought this home very strongly.

It happened at the Annual AgTech Alchemy event at Shapeshifter’s Cinema in Oakland, CA. I promise I was not in an inebriated state, even though the proprietor of Shapeshifter’s Cinema serves some fine home-brewed beer, and there was a generous supply of wine and bourbon!

The discussion was civil but very spirited when we began discussing water issues affecting California agriculture. We discussed automation technology, smart irrigation, water policy (e.g., SGMA), labor issues, supply chain infrastructure, and how each affects farming and food outcomes in the short term and the long term.

The discussion highlighted the challenges facing agriculture, particularly in resource utilization, policy changes, financing mechanisms, social and cultural issues, and scaling technology.

The discussion was a great reminder for me personally.

Technology is a great enabler for solving small and large problems.

But technology has to be adopted at scale for it to have an impact.

Adoption at scale requires the right business models, the right alignment of incentives, the right distribution networks, economic value creation for the ecosystem, stage and problem/solution-appropriate financing mechanisms, and the right policies in place (especially true in agriculture).

So what can you expect from SFTW in 2026?

Founder and Startup stories

Passionate founders and startups play a critical role in any ecosystem, and agriculture is no exception. SFTW will continue to bring the perspectives of founders and their startups to the SFTW audience.

Widening the lens

SFTW will broaden its lens to examine how technology is adopted, from startups through scaling. SFTW will always take a technology-centered lens but will also consider other factors relevant to scaling technology, such as distribution, incentives, financing mechanisms, supply chain, and policy issues.

I recently signed on as a Senior AI Advisor to the Gates Foundation for its AI advisory work on smallholder farming, and you may hear more about it as well.

Funding models

In agriculture, it is fashionable to blame VCs for wasting money and not having a long-term perspective. Yes, many VCs can definitely be blamed for making bone-headed moves, but if every VC investment were successful, then it would not be called venture capital.

We need different funding models and approaches that align with the problem and solution type, technology maturity, and company growth stage. I had explored a few questions about this in last year’s seminar.

You will see more programming from SFTW in 2026 on this very important topic.

Trend Analysis and Future Casting

One of the most common types of questions I get is

Where do you think <a particular thing in ag> is heading?

Or

Where do you think <this concept/topic/idea/sector> will be in 5 years?

Or

What do you think will be the impact of <AI or some technology or capability> over the long term?

It is not easy to answer these questions, but I will do my best to provide my rationale for having a particular point of view. Many of these “future castings” will prove inaccurate, and I would appreciate your perspective on this topic.

Here’s to another interesting year ahead as we explore these topics together!

I came to know Brad Zamft when I was working at Mineral at X, the moonshot factory. Brad was working on a sibling project around using AI for genetic improvement in crops. The project is now called Heritable Agriculture, an independent company outside of X. Brad is a physicist by training, but transitioned to plant biology over a few years.

Brad has a bold vision to make genetic improvements in crops much more affordable. These genetic improvements can result in higher-yielding crops, spare agricultural land and deforestation, and result in nutritious and drought-resistant crops that use fewer resources.

I hope you enjoy this conversation as much as I did having it, and hopefully Brad’s passion comes through it.

Summary of the Conversation

Brad Zamft discusses the evolution of his career from physics to plant biology, the challenges and opportunities in agricultural innovation, and Heritable’s mission to revolutionize crop genetics using AI.

We discuss how the convergence of genetic data, combined with powerful AI tools, can lead to a golden age of low-cost genetic improvements, especially for a wide variety of crops that are typically ignored by the world’s agricultural behemoths, which focus only on corn and soy. Brad and I talk about how it could be transformative for what’s possible in the Global South with better genetics for underserved crops and the potential impact on global security and health.

Rhishi: You trained as a physicist, not a plant biologist. How did that transition happen? How has your physics education shaped or helped the work you’re doing now?

Brad Zamft: I chose my undergraduate degree when I was around 16. Back then, I thought I wanted to be the next Albert Einstein. I really loved physics in high school. I’m a conceptual thinker; I think in pictures. And the way biology was taught in high school, at least when I was there, didn’t click with me. It mainly involved memorizing organelles and rote facts, which I didn’t enjoy and didn’t do well in. Physics, on the other hand, was all about drawing arrows on free-body diagrams. That resonated with me.

But then I got to college, and it gradually became clear, though it took me four years to admit it, that I wasn’t going to be the next Einstein. I was in over my head. That realization led to a bit of an existential crisis.

After I graduated, I did a round-the-world backpacking trip. I went to Cambodia, rural villages in China, and Jordan, for example. That experience completely changed my life. I became fascinated by the diversity of people, perspectives, and ways of living around the world.

That’s when my mindset shifted. I decided I wanted to devote my life to making an impact and serving people who are less fortunate than me. I had already been accepted into the UC Berkeley Physics PhD program during that trip. So I came up with a grand plan: I’d join the program, get a master’s in physics, take all the pre-med classes for free, then become a doctor. My goal was to join Doctors Without Borders and work in war zones and underserved communities.

But once I started taking those biology classes, I got really fascinated with engineering biology. That shifted my path again. I dropped the med school plan and finished the PhD.

After that, I joined George Church’s lab as a postdoc, which got me into bona fide synthetic biology. And honestly, I think physics prepares you surprisingly well for many fields. There are physicists doing biology, linguistics; you see them everywhere. That kind of quantitative, first-principles training translates well across disciplines.

Everything is becoming more quantitative now. And having that strong foundation in quantitative analysis has been one of the biggest assets in my career as a biologist.

Rhishi: You and I both worked at X, the moonshot factory. X has undergone significant change over the past few years. X follows a particular model of innovation. How well did the X model work for what you were trying to build? What were the pros and cons of working within that framework?

And if you were in Astro Teller’s shoes, what tweaks would you have made to give a company like yours a better shot at success?

Brad Zamft: I can really only speak to the version of X I experienced, which was also the version you experienced. There was an earlier era of X, the one that produced Waymo and Verily, when budgets were enormous. Then came the X we were part of, where budgets were more modest. And now there’s today’s X, which I don’t feel qualified to comment on.

But what’s remained consistent throughout is Astro Teller’s vision of a moonshot. I think that vision is essential. The world needs it.

Alphabet needs it. Companies like Alphabet have to disrupt themselves because if they don’t, someone else will.

Even a company with a technological moat as deep as Alphabet’s, your phone, your computer, your search history, your GPS location, and access to almost every kind of data, can get caught off guard. Five years ago, if you’d asked me whether anyone could ever touch Google in the search business, I would’ve said no way. They had everything: data, infrastructure, dominance.

And then along came ChatGPT.

That’s what these disruptive moments look like. And even the most prominent players, Alphabet and Microsoft, have to sit up and pay attention.

In my view, that’s the core purpose of X: to build things that have the potential to disrupt the Alphabet from within. And beyond that, from a societal standpoint, we need people working on ideas that seem a little bit crazy, not impossible, not violating the laws of physics, but bold and far enough that you couldn’t fund them in a traditional industrial R&D setting.

We also need that work to stay focused on impact, which is something academia sometimes loses sight of. That’s why the world needs X, and it needs X-like organizations around the globe.

Where Heritable is today is a direct result of X. We wouldn’t be where we are now without the patience, the ambition, and the audacity that X made possible. I have no regrets. I have nothing but appreciation for X. I hope it continues doing what it does for a long time to come.

So spinning out Heritable, and letting it raise capital from investors who understand ag, who have the right connections and context, that’s the right move.

And honestly, I wouldn’t change anything about Heritable’s path. It unfolded the way it needed to. And I’m proud of that.

Rhishi: If someone were walking down the street and asked you, “What’s Heritable?” how would you explain it in plain language?

Brad Zamft: When I explain it to folks on the street, I say this: we use artificial intelligence, along with the revolutions happening in DNA and RNA sequencing, remote sensing, and satellite imagery, to help everyone grow better crops.

Our goal is to improve all crops across all locations and land types using the combined tools of AI and biology. That’s the grand vision. But we have to focus, especially as an early-stage startup. So we’ve gone where the customer traction is strong, where we’ve found the right partners, and where the opportunities are clearest.

Right now, that’s in vegetables, and, believe it or not, in forestry.

Forestry represents a massive opportunity. It’s been largely underserved, mainly because of its long improvement timelines. But that’s precisely the kind of challenge we’re built to take on. At Heritable, our mission is to accelerate those timelines and reduce the cost of crop improvement. Forestry is a perfect example of a high-impact space that desperately needs those efficiencies.

Your audience probably knows this already, but there’s a massive drop-off in resources when you move away from corn and soy. Everything else, whether it’s leafy greens, brassicas, or root vegetables, gets significantly less attention and funding.

That’s why we’re focused on the spaces where modern computational techniques can make the most significant difference: in crops and geographies that matter profoundly for nutrition, food security, and climate adaptation, but haven’t yet seen the full benefit of today’s genomic and AI-driven tools.

Rhishi: When I worked at Bayer, corn and soy were the king and queen crops, however you want to label them. Everything else tends to be treated like a stepchild. Corn and soy is just a bigger business, plain and simple. Also, the cost of bringing a new genetic variety to market isn’t all that different whether you’re working on corn or a smaller crop. From a company’s perspective, it makes perfect economic sense to focus where the returns are highest.

So that brings me to you: How are you solving that problem? Is your secret sauce that you’ve figured out a way to develop new genetic traits much cheaply? Because if you can change the economics, then maybe you can change the entire strategy.

Brad Zamft: It comes down to making crop development cheaper and faster. That’s our core mission. At the highest level, we want to improve all crops in all geographies using biology and AI. But just one level down, our focus is on making crop development more affordable and more efficient.

But when you look deeper into agriculture, you realize there are thousands of other players, smaller companies, research institutions, breeding programs, who would love to access corn-level genetic improvement, but simply can’t afford it.

We have more biological information now than at any point in history. Twenty years ago, if you sequenced 50 bases of DNA, you could publish in a top-tier journal. Today, sequencing an entire metagenome might not even be enough for publication. We’re swimming in genomic data.

At the same time, we’ve seen similar leaps in how we measure and observe the world: satellites, drones, weather stations, soil sensors. We can capture environmental context at incredible resolution.

The key is integration: bringing together biological and environmental data and making it actionable.

For the first time, we have the computational tools to analyze these massive, multimodal datasets and extract insights that just weren’t possible before. The tools have changed. And that’s precisely where Heritable comes in. We’re challenging that assumption. We’re saying: now is the time to go beyond corn and soy, because we finally have the means to do it.

Rhishi: When you talk about crop improvement, what are the specific dimensions you’re targeting for improvement? And what are some of the most in-demand or highest priority traits you’re focused on?

Brad Zamft: We currently offer three products.

The first is genomic prediction. We take the entire genome and predict how that plant will perform. What makes our approach unique is that we incorporate environmental data: soil, weather, and climate. This lets us create a digital twin of a plant, a simulated version of that plant, at a 10-meter resolution, anywhere in the world.

The second product is designed for situations where you don’t fully understand the genetic architecture of a trait. If you don’t know what genes control a trait, we use multiomics, a complex mix of molecular data types inside the plant, to identify the key genes responsible.

But even knowing the genes often isn’t enough. If you already know the genes, the question becomes: How do you control them? How do you turn them on or off? How do you activate them in specific tissues or under specific stress conditions?

That’s where we leverage large language models. We use them to pinpoint the specific DNA bases you can edit to control gene expression, timing, location, and conditions.

Now, to your question about traction: where are we seeing the most demand?

The first product, genomic prediction and breeding, has generated by far the most interest. And it’s not just because we can improve genomes; we also model the environment. That opens up a whole new dimension: not just how to change a plant, but where to place it.

Let’s say you want to expand your geographical footprint. Maybe you’re facing climate risk, geopolitical risk, or you just want to optimize where you grow different varieties. Our system helps identify which environments are best suited for which genomes, and vice versa.

We’re seeing a lot of demand from companies and growers who want to make those strategic decisions with more confidence.

Rhishi: One of the fundamental challenges in agriculture is context, the agroecological conditions, the soil type, the rainfall patterns, and everything else. All of those factors vary, even for the same crop.

Seed genetics are developed by optimizing for a broad set of conditions, which work well in major production zones like northern Iowa. For example, you design a corn variety with a specific relative maturity that performs well across a large region.

You could imagine a world where every field has its own optimized seed. Maybe every plant has its own genotype. That might sound crazy, but conceptually, you could go that deep if the economics and tech made sense.

Brad Zamft: I do think it’s worth dreaming for a moment. Let’s go back to our X days, where thinking exponentially was encouraged.

There’s a future, if we think in exponential terms, where the produce you buy at your local farmers market comes from genetics that were custom-made for that exact farm. The crop ripens the day before the market. It’s fresh, it’s nutritious, it’s high-yielding, it’s cost-competitive, and it’s sustainable.

And here’s the thing, we already have genetic segregation. I’m not even sure we need to make custom seeds for every single person. That would be death by a thousand paper cuts, the operational cost of that would crush you. But in a true science fiction future, and honestly, we are living in the science fiction I read as a kid, we’ll be able to use natural segregation, predict it, and match the right genetics to each farmer, even down to individual preferences.

More importantly, and this goes back to your roots, and to my own revelation when I was backpacking, we’ll be able to deliver competitive, high-performing genetics to smallholder farmers. That’s what could really lift the quality of life in those communities.

You’re right, though operational complexity is the killer. That’s why we’re not starting there.

But we’re also not starting from the dominant paradigm either. We’re not designing genetics that average across all of Iowa, because while that makes perfect sense for large seed companies trying to manage complexity, it’s not what we’re here to do.

It’s all about finding the right partner.

I was just in Canada two days ago. We’ve announced a great partnership with Paul J. Mastronardi, one of the largest growers of indoor vegetables. He grows year-round, primarily in controlled environments.

It’s not a smallholder farmer scenario, and it’s not an operational behemoth like corn or soy. It’s a focused use case. And even though this market is big, it’s still not big enough for the major seed companies. They’re primarily focused on field crops, such as strawberries, grown outdoors.

That’s where we come in.

We’re not chasing hyper-fragmented logistics, but we’re also not playing in the high-scale, high-commodity world. We’re working with a partner who sells massive volumes of fresh produce each year and needs better genetics to keep going.

Rhishi: I visited some greenhouse tomato operations in Canada about 12 years ago. At the time, it felt like something out of science fiction. They had all this automated vine-grown production on trellises. The tomatoes would rotate, and a lot of hydroponic work was also happening.

That might be a good segue into your business model and partnership strategy.

What’s your go-to-market plan? When you’re evaluating partnerships, what’s your rubric?

Brad Zamft: We’re looking for partners who want to take the journey with us. That means real cost-sharing and real value-sharing, if I’m being honest and getting down to brass tacks.

We want to build new entities. We want to help expand your business and take you into places you can’t reach right now. That could mean developing an entirely new trait that you never thought was economically feasible. It could mean entering a new region. Or it might mean taking on an entirely new crop.

But it comes at a cost. We’re building that business together.

We’re not interested in the dominant model, where you license a trait, take a 1% royalty, and the seed company keeps the rest, splitting it three ways between themselves, the distributor, and maybe the farmer. That’s not what we’re after.

We’re looking for true partners, people who want to help build new subsidiaries, new business units, maybe even brand-new entities down the line.

Right now, we’re working with Paul J. Mastronardi on strawberries. That’s a great example of the kind of partnership we want to keep building.

Rhishi: Strawberries have a pretty complex genome. One significant challenge in agriculture is that it’s really hard to transfer knowledge and tools from one crop to another.

It’s not like being great at corn breeding means you can just turn around and start breeding strawberries. The skill sets, tools, and data pipelines they’re all different. It’s a whole new playbook.

So, going back a bit to the technology side, and maybe this ties into the business model too, how easy or difficult is it for you to switch from one crop to another?

Brad Zamft: I mean, we haven’t reached that point yet. Sure, there’s a science fiction future where we’ve built a simulator for all of biology, or maybe a little less ambitiously, for all of plant biology, or all dicots, or all vegetables. We’re not there today. But I do believe that kind of future will become reality. It’s just a matter of timeline.

What is here today, and what we’ve developed and productized at Heritable, is our approach. And that approach has several key components.

First, we have one of the best teams in the world at running multi-omic field trials. There are plenty of people out there who know how to collect DNA and follow genetic markers in breeding programs. But we’ve worked with other groups who say, “Don’t worry, we’ll handle everything, we’ll extract the RNA, we’ll get you all the molecular data.” And I can tell you, when we do it ourselves, it’s better.

We’ve done this over and over, across many crops. We know how to design the trial. We know how to do tissue sampling, extraction, and sequencing. We’re world-class at that.

Then there’s data cleanup, a very underappreciated step in any machine learning or AI pipeline. After that, there’s data integration, and that’s where we bring in our major differentiator: environmental modeling.

We can predict how a plant will grow anywhere in the world at 10-meter resolution, and do so reproducibly. Our ability to integrate the environment into the model is foundational and transcends crops.

That’s really the point: this approach transcends crop. Yes, there are minor differences, the tissue you sample will depend on the crop. The trait you’re optimizing will vary depending on the project. But over the last five years, we’ve fine-tuned every part of this workflow.

We’ve also dialed in the model’s architecture. We’re not constantly rerunning architecture searches or asking which algorithm will perform best. We’ve learned those lessons. We know how to do this. We don’t need to relearn it every time.

Now, is the holy grail full transferability, where a model trained on corn instantly knows how to predict outcomes in strawberries? Sure, that would be amazing. But we’re not there yet. And honestly, I’m not even sure that level of transferability is necessary right now.

There’s already so much demand in the industry that we can run the field trials we need, and the cost isn’t a significant barrier. So while we’re not in the science fiction future yet, we’re delivering fundamental, scalable, high-impact tools that are working right now.

Rhishi: When I was at Mineral, getting access to data to train our models was one of the biggest challenges. How are you managing the process of getting the data you need to train your models?

Brad Zamft: Yes, everyone is still cautious about sharing their data with us, and rightfully so. But that really comes down to relationship management. It’s about finding the right partner and building trust. That doesn’t happen quickly. I haven’t had a sales cycle for under six months yet.

Reducing the sales cycle requires building trust one-on-one, but it also happens more naturally over time, once we’ve proven ourselves through successful projects. As people see that we’re a trustworthy partner, that we’re not out to misuse anyone’s data, that we work securely and transparently, those barriers start to drop.

And again, it depends on the partner. Some organizations understand that the future of agriculture will rely on specialized collaboration. You need people focused on artificial intelligence and predictive tools, and others focused on operations such as growing, packing, shipping, and distribution.

Rhishi Pethe: Let’s say you’re working on a particular crop, you’ve done the genome characterization, identified the key genetic markers, and now you want to understand how that plant will perform under different environmental conditions. You need data to validate your models.

What are the most significant barriers to driving down the cost and increasing the velocity of what you’re doing? For example, we saw the cost of sequencing a human genome drop by over a million-fold in about 20 years, while the speed of sequencing skyrocketed by several orders of magnitude.

Brad Zamft: Most of agriculture still has a long way to go, and we can get very far with better breeding.

Just look at corn. For 80 years, it’s shown genetic gains of about 1.7 bushels per acre per year. And that mostly comes from agronomic breeding, dihybrid crosses, and computational biology, not from GM.

After 70+ years, we’re still not done. The genetic potential of corn hasn’t been exhausted.

Now shift your focus to crops like sorghum, or even to trees. In the time we’ve been breeding corn, trees have maybe gone through five breeding cycles. There’s so much low-hanging fruit if we just improve the way we breed, especially if we incorporate environmental data better.

That’s why the vast majority of our business model is focused on breeding. There’s an enormous unmet need and massive opportunity in just doing that better.

Of course, we’re happy to support gene-editing projects today when partners ask for them. But let’s look at the addressable market.

You start with everyone who uses genetics to deliver a plant product. Then you narrow it down to those who are in an industry or geography where biotech-modified products are actually accepted, either by regulators or the public. That cuts out a huge portion of the market.

Then you narrow further to those who are not only willing to take that risk but also have the resources to do it. Gene editing often requires tissue culture and lab infrastructure, which are significant barriers.

Image source: BBC Research

And it’s not always about regulation or public perception. Sometimes, the logistics and cost of implementation are the bigger reasons to hold off.

That’s why, for now, we focus primarily on breeding. It’s where we see the most potential, technologically, operationally, and economically.

Rhishi: I recently read a book titled “Globalization of Wheat.” The author argues that, while the Green Revolution was undeniably transformative, breeders focused heavily on wide adaptation. She means they intentionally bred wheat varieties to perform consistently across a broad range of agroecological conditions.

The tools available at the time pushed them in that direction. And they achieved that consistency not just through genetics, but by leveling out environmental differences with intensive irrigation and heavy fertilization. That was part of the strategy: make the environment more uniform so the seed could be more uniform too.

What I find interesting about your work is that you’re not aiming for widespread adoption. Your approach seems more about fine-tuning. And in a more speculative or science fiction version of your work, it could go all the way down to the field, or even plant-level personalization.

How do you think your tools or technology could shape agriculture in the Global South?

Brad Zamft: We have to do this. We have to figure it out. No one’s solved it yet. People have tried, but the conditions have changed, and now we have the tools to give it another shot.

I’m not an expert in international development, and I don’t pretend to be. No one in the audience should take me for one. But from what I understand, there’s a critical inflection point when a subsistence farmer starts producing just a little more than what they need to eat.

That shift, from subsistence to a surplus, can be a quantum leap in someone’s life. It allows them to sell what they grow. And once they can sell, they can start focusing on other things: education, health, and improving their quality of life. So we have to figure out how to enhance the genetics of those farmers. And when we do, it will have a massive, immediate impact.

But there’s another side to the story.

Management and business structure matter. Many subsistence farmers grow open-pollinated varieties. They don’t buy seed. So how do we monetize improved genetics when there’s no established genetics industry? No supply chain? No infrastructure?

This is where we need collective thinking. This is where the whole community, especially people more qualified than I am, needs to step in and ask: How do we build an industry model in which a farmer can pay for improved seed and make a return on that investment?

And maybe the key isn’t starting with the lowest-resourced farmers. Perhaps the answer lies in starting with the middle or lower-middle segment, those with a bit more capacity to adopt, adapt, and build momentum.

It’s not a solved problem. But we’re doing our part. And for me, this is one of my life’s missions: to help those less fortunate than us improve their quality of life through better crops, better tools, and better opportunities.

Rhishi: If Heritable is as successful as you hope it will be, what tangible impact would you hope to see? What would make you say, “Yes, we did something that mattered.” Feel free to touch on any areas that matter most to you, whether it’s forestry, nutrition, climate resilience, or something else entirely.

Brad Zamft: As we continue to de-risk our technology, we’ll move closer to that science fiction future we talked about earlier. And when we do, we’ll unlock a world of applications, because plants truly are miraculous. They’re solar-powered, carbon-negative, self-assembling machines. If you get the genome right, you can literally put a seed in the ground and return later to find that you’ve grown the function you wanted.

If we succeed, we’ll see sustainable, diversified forestry, not vast monocultures of genetically identical trees planted in rows, but ecosystems that reflect a deep understanding of biodiversity and environmental dynamics, supported by genetics that back it up.

In row crops and vegetables, we’ll have nutritious food that competes on price with less healthy options. You’ll be able to buy strawberries that taste amazing, and your kids will actually prefer them over Cheetos. And they’ll be priced competitively. You’ll have supply chain certainty, especially in indoor agriculture, where one of the main goals is year-round, local production. So yes, you’ll have strawberries all year, and they won’t have to travel 3,000 miles to reach your plate.

And then we can really begin to embrace science fiction.

Ten years from now, we’ll have dropped the technical and operational barriers enough that we can start sequestering carbon at scale, not through inputs or amendments, but through the very plants we’ve designed. These plants, these “machines”, will pull CO₂ from the air and push it into the soil, rebuilding our soils automatically.

We’ll also have food as medicine. In fact, we already do. Many of our pharmaceuticals are plant-derived. But we’ll go further. We’ll optimize plants to express those medicinal compounds more consistently and at higher levels. You’ll eat a salad not just for nutrients, but for specific health benefits, because we’ll have designed that into the plant itself.

Everything becomes possible once we remove the barriers and accelerate the timeline for improving crops across species, across systems, and across the world.

Rhishi: That’s a very inspiring vision, Brad. Thank you for your time today, though I do love my Cheetos!