A Buffett of algos for agtech

Introduction

Mental models (algorithms) are rarely formalized, but VCs develop an intuitive sense for them over time. These algos aim to reduce failure. Nonetheless, VCs still make poor investment choices (about 2/3 of all investments return less than 1x).

The main body of this article examines algos for agtech, but given volatile and infrequent agtech exits, I compare with a longer established VC sector, namely biotech. But firstly, is it possible to imagine algos to identify good and bad times to invest in the longest market of them all – public markets?

Note that agtech definition is here mainly confined to its use in green ag (crops and tractors) and does not include red ag (livestock) or blue ag (aquaculture).

PUBLIC MARKET ALGOS

Algo 1. What percent of shareholder equity is Berkshire Hathaway holding in cash?

Warren Buffett’s investment firm, Berkshire Hathaway, is currently keeping vast amounts of cash, something he doesn’t like as an asset. According to Bloomberg, over the years, the proportion of Berkshire’s cash allocation relative to the company’s assets has varied greatly, from as little as 1% in 1994. Records show that as stock valuations rose during prosperous periods, Buffett constantly increased Berkshire’s cash allocation, resulting in a decline in expected return rates, and reduced cash when opportunities arose.

Some of you might remember Barron’s article in December 1999 which asked, “what’s wrong, Warren?”, suggesting he had lost his touch when his returns were poor relative to the stock market. And he had loaded up on cash then too (albeit to less than today’s percentage of total shareholder equity). We all know what happened next, with a stock market peak in March 2000, a mere 3-4 months later. From there, the NASDAQ fell by 78%. Today, as a reminder, cash sits at around a third of shareholder assets...

Algo 2. The Buffet Index

The Buffett Index might indicate if a market is overpriced. It is calculated by dividing the total market capitalization of all publicly traded stocks by the country’s Gross Domestic Product.

Currently it stands at 2 standard deviations away from the long-term average, which explains why he loaded up with cash in 1999 and again today. The correlation between Berkshire Hathway cash and the Buffett Indicator is weaker than expected across multiple decades (0.11). However, in the last 15 years, it appears to be materially stronger (0.76), according to AI. The “signal curve” behaves like a sentiment/valuation proxy: at a total market cap:GDP ratio of <150%, cash is relatively contained; at ~150–200%, there is a gradual cash build; and at >200% there is a convex response in increasing cash assets. The non-linearity suggests Buffett’s behaviour is threshold-driven, not linear.

Algo 3. The forward p/e ratio

Another way to look at expected returns is the forward p/e ratio. The link between returns after one year and forward p/e ratios is weak with the stock market, as Buffett says, acting as a voting machine. But over a 10-year period, it is a weighing machine with returns inversely and linearly linked to p/e ratios. Today’s ratio (>20) indicates 10-year returns are likely to be poor.

From the above three examples which measure the public markets, it is possible to imagine how these might act as a shorthand guide to investors in the stock market, at least if you have patience. I do not vouch for whether they will be predictive in the future or how practically useful they were in the past. But it made me wonder about rules of thumb that might work in agtech investing.

AGTECH ALGOS

Agtech is often cited as a relatively new segment for VC investing. With that in mind, we could learn from another VC target with a more liquid and storied past which is often considered similar, namely biotech. In many ways there are similarities between biotech and agtech: long regulatory pathways, a requirement for efficacy trials and incremental improvement in performance (often crop yield improvements or healthcare outcomes).

But the structural differences are also important. Healthcare represents about 17% of US GDP, while agriculture is about 0.9%, a ratio of 20:1.

Similar ratios between the two sectors exist for other metrics:

• Aside from the SPAC blip, agtech exits have been almost exclusively via M&A; while in biotech, at least a few go public

• There are many biotech buyers – over 1000 – of which the top 20 are 90% of the collective $5t market cap. All of them make over 40 deals, each worth more than $20m, per year. There are far fewer buyers of agtech (mostly tractor manufacturers and agchem) and these have a total market cap of $250-300b. Over the last 20 years they have collectively made per year, about 1-2 deals worth more than $20m each (see my previous essay here). The annual $ spend, averaged over the prior 10 years, is consistently in the range of $270-400m per year (see Figure 4).

• The top 7 biotech buyers each have about the same or greater market cap than the entire agchem/seeds/tractor industry put together

• A 10% yield improvement on a crop is priced completely differently to a 10% improvement in health outcomes for say an oncology drug; the benefit of the first can be calculated according to value of increased crop yield minus any change to input cost, while the second… how much are you willing to spend on the better oncology drug?

• Ag trials don’t have the clear outcomes that Phase I-III trials have in healthcare, nor the same impact on valuation

• The few agchem buyers are currently distracted, reducing their appetite for M&A, for example:

• BASF’s agchem division is going public
• Corteva is splitting up
• Bayer is struggling with its Monsanto acquisition

I wondered whether another way to look at agtech is via Axtell curves[1]. These are double-log plots of frequency of deals versus deal value. Aside from showing the current structure of the M&A landscape between agtech and biotech (and software), I wanted to explore the hypothesis that the ratio between biotech and agtech deals was still of the order of 20:1.

Each industry’s core economic drivers uniquely shape its M&A transaction landscape:

• Agtech (Steep Slope / Compressed Market): Dominated by a few giant conglomerates and a cluster of tiny startups. Mid-tier transactions are rare. Once outside the $30m–$100m bracket, deals drop off sharply (indicated by the steepness of the curve), limiting massive standalone exits.
• Biotech (Polarised Power-Law / Binary Model): Driven by clinical trial outcomes, resulting in many small preclinical deals and a few multi-billion-dollar blockbusters. This leaves the $300m–$1b mid-market notably empty.
• Tech Software (Flat Slope / Continuous Market): Driven by predictable revenue multiples that allow companies to scale smoothly. This creates a highly active mid-market with continuous exits from $50m up to $5b.

Ultimately, software offers a balanced, healthy ecosystem. Biotech operates as a high-stakes blockbuster lottery and AgTech remains highly restricted by legacy strategics (reinforcing the point that M&A is the dominant exit route).

In conclusion, as the agtech-biotech Axtell lines aren’t quite parallel, it isn’t possible to give a single ratio, but at the upper end it is somewhat less than 20x, at just under 5x.

All but the 5 top agtech exits (all overpriced SPACs[3]) took place below $1b, while the top 20 agtech exits in the last 20 years, excluding SPACs, have averaged $400m. If this $400m average continues to hold true (optimistic based on Figure 5 which uses more recent data), and each new investor at entry anticipates a 3x step up between rounds[4], then this investor might follow this power law logic:

• The last round (Series B) prior to an exit at $400m would limit the post-money of the last round to $133m, or a pre-money of about $100m
• Prior round (Series A), at a post-money of $33m, or a pre-money of about $25m
• Seed round, at a post-money of $11m, or a pre-money of about $8m

Necessarily, all exits do not all take place at the highest valuation, which would dictate even more conservative round valuations are required to provide returns for the full waterfall of Preferred and Common shareholders. This has been SGV’s experience in practice as well, where companies with valuations falling in this range at each round have been more successful, while those with substantially higher intermediate valuations have not. One might conclude that high intermediate valuations are delusional and toxic to early-stage companies. Many from the agtech ecosystem believe we are at the bottom of the cycle. As the above valuations are out of kilter with rounds presently taking place, it appears the current valuations have further to fall.

The corollary to valuations is equity raises. The total equity raised in the sequence above is $44m, but many agchem/OEM smallcos have raised larger amounts of money. Do they succeed? According to a Pitchbook report[5],[6], no:

“…for both investors and founders, capital discipline is critical. Companies raising under $1M failed at least 78% of the time, but those raising over $100M before failing destroyed the most capital. The sweet spot is $15M–$60M total raised, exiting at $75M–$200M to a strategic acquirer.”

This range captures the $44m equity amount in the previous paragraph. Yet multiple privately held agtech companies still exist that have raised such high levels of equity, meaning new investors in the last round might be disappointed; big rounds enable a high burn which is not sustainable. How these remaining well-funded companies fare will be an important determinant of future money flows into agtech.

Some agtech sectors are very difficult to fit within the power law hopes of VC, such as capex intensive vertical farms (where reaching price parity with traditionally grown crops is tough) or genetics for row crops which suffer from high fixed costs, long timelines to revenue, regulatory burdens and the need to partner with a tiny handful of relatively poor seeds companies in order to scale. This echoes Buffett’s scepticism towards the airline industry which also has

• High Burn Rates: Like airlines, row crop genetics requires up-front capital for research, field trials and regulatory approval before seeing a profit
• Unpredictable Adoption: When times are tough, airlines lower prices just to fill seats, making it a “disaster for capital”. Farmers are slow to adopt new technologies where the benefit is uncertain, making sales cycles long and revenue volatile
• Weather: weather can hamper smooth airline operations and for farmers it disguises season-to-season yield benefits, making them sceptical of uncertain future outcomes but certain of up-front expenditure
• Thin profit margins: Airlines require massive amounts of money to buy and maintain hardware, yet they suffer from razor-thin profit margins, frequent bankruptcies, and intense price wars. Not so different to farming.
• Market Vulnerability: Just as airlines are at the mercy of fuel prices and pandemics, farmers are at the mercy of commodity prices and crop pests and disease.

Channelling Buffett: “I have an 800 number now that I call if I get the urge to buy a vertical farming company which has a high burn rate. I call at two in the morning and I say: ‘My name is Michael, I have a capex addiction, and I’m looking for a fix.’ And then they talk me down.”

The structural constraint in agtech vs. biotech (agtech is 5-20x smaller) means inherently more conservative assumptions for agtech VC across valuation, capital, and strategy than biotech. Combining the above and a past essay[7], I attempt to provide a high-level codification – a list of five shortcut algos – to improve agtech exit outcomes.

For investors:

1. Laser focus on entry point valuations: empirically, an agtech exit of $400m is truly awesome. Back-solving with 3x step-ups implies tight round valuations (Seed ~$8m pre, A ~$25m pre, B ~$100m pre). Once companies exceed these values, exit upside becomes structurally constrained. Inflated rounds are toxic to returns and company survival. A simple (still very generous) filter might be “Can this still deliver >3x if exit is <<$400m?”
2. Avoid over-capitalized companies with >$50m already raised or have high future equity requirements: high burn and dilution lead to weak waterfall outcomes and presage future down rounds. Prioritize capital-efficient startups with controlled burn and a clear path to breakeven.
3. Look for acquisition signals: exits appear more likely when collaboration or equity investment exist pre-acquisition.
4. Avoid some sectors with a difficult fit to VC power law needs: Areas like vertical farming and row-crop genetics face airline-like economics: high upfront capex/burn, long timelines, slow/uncertain adoption, and weather/commodity volatility, making venture outcomes challenging.
5. Actively manage the buyer universe: Recognize the structural limitation of a small, cyclical buyer pool (OEMs/agchem). And consider expanding into adjacent buyer segments (food, climate, fintech, industrials, AI-enabled players).

Boards might consider:

• Tight round valuations: Returns in agtech are driven by valuation discipline and capital efficiency. Counterintuitively, it seems taking the highest paper valuation prior to exit might be your enemy, leading to down rounds or onerous terms, or both.

• The siren call of IPOs: Accept the structural reality that agtech IPOs are infrequent and, even then, always underperform. M&A is the dominant exit path:

• Build strategic relevance early: Align product development with clear buyer use cases in OEM/agchem/seeds, then
• … partner with likely acquirers, while maintaining exit flexibility: Adopt a “try-before-you-buy” model [8]: commercial pilots → strategic JV → minority stake → acquisition

• Don’t overshoot the runway: Buyer appetite is cyclical and linked to external drivers (e.g. commodity prices) or internal distraction within the largecos. Long runways maintain flexibility of exit timing when buyer economics or attention improves. Achieve them by low cash burn, not by large raises.

In summary, investment cases and board level focus should be built around realistic exit ranges that avoid overfunded/overvalued companies, credible and diversified exit routes and acquirer engagement.

As a prospective agtech investor into early-stage companies, my perspective may be conflicted, but it is conservative and reflects the structural constraints at exit for agtech. To the extent that the above is predictive, I am mindful of Buffett’s quote: “We’ve long felt that the value of stock forecasters is to make fortune-tellers look good.”

So don’t quote me when all the above turns out to be bunkum.

Footnotes:

1. Log-log curves exist in many diverse phenomena, including across multiple aspects of businesses. See Axtell, Science, 293, 1818-1820; see Ball’s book titled “Critical Mass” for an accessible summary
2. I created my own curves for deals from the last 20 years using Traxcn data for both biotech and agtech and got broadly the same result as the graph shown here, AI generated. AI source data for Biotech were cited as: ⁠EY’s 2026 Biotech Beyond Borders Report; McKinsey & Company; ⁠J.P. Morgan Biopharma & Medtech Deal Reports. For software: ⁠Software Equity Group & The SaaS M&A 2026 Report; ⁠Kroll Global Software Sector Update; ⁠Bain & Company’s Global M&A Report. For Agtech: ⁠Capstone Partners AgTech Market Update; ⁠PitchBook’s Agtech VC & Exit Trends
3. frothy as SPACs have dropped collective valuations by 98% since IPO
4. In practice, the step up between rounds is about 1.6. https://pitchbook.com/news/reports/q2-2026-pitchbook-analyst-note-what-does-a-good-investment-look-like-in-agtech. An excellent sector specific report
5. https://pitchbook.com/news/reports/q2-2026-pitchbook-analyst-note-what-does-a-good-investment-look-like-in-agtech.
6. https://seanmulvany.com/2026/05/03/there-are-returns-to-be-had-from-agtech/
7. www.agnavigator.com/Article/2025/06/25/the-whys-of-agtech-exits-the-empirical-evidence/
8. Ibid