Danny Bernstein left Big Tech to fund farm-bots | E2310

Danny Bernstein spent close to 20 years in Silicon Valley, including 10 years at Google, before leaving to study complex outdoor industries. He observed technology abundance at Google and decided to apply Silicon Valley principles to specialty crop agriculture, a sector he viewed as underserved relative to broad acre crops like corn and soy, which benefit from subsidies and programmatic farming approaches.

The Reservoir operates as both an incubator and a fund built around a 40-acre agricultural innovation campus in Salinas, California, one hour from Silicon Valley. Reservoir Farms is two years old, operates over 100 acres across multiple farms in the American West, employs 10 full-time staff, hosts 20 startups paying between three and six thousand dollars a month in rent, and is expected to generate between five and seven million dollars in ARR. Reservoir VC completed a proof-of-concept fund of three million dollars and wrote five checks, with one portfolio company acquired and rebranded as Bonsai Robotics, formerly called FarmNG. Bernstein said fund and incubator investment decisions are kept separate, and Reservoir does not put money into companies and then reclaim it as rent.

Labor costs are the central economic pressure driving Reservoir's focus. Labor accounts for 40 percent of production costs across specialty crop agriculture broadly, 60 to 70 percent in strawberries, and 80 percent in table grapes. The American Farm Bureau found that over 400,000 farm jobs were posted last year and fewer than one percent received a domestic applicant. Bernstein said domestic workers not wanting farm jobs is categorical and clear, and that H2A visa labor pools do not scale as production expands into new regions. He views technology as the only mechanism that scales in place of unavailable domestic labor.

Reservoir focuses on three problem sets: chemical reduction, chemical replacement using technologies such as electric and laser weeders, and automation of the hardest-to-fill jobs. Bonsai Robotics runs autonomously through vineyards doing precision application of biologicals as bio-based alternatives to pesticides. Bernstein described strawberry harvest as the white whale of AgTech, with two or three startups attempting it every year. A prior portfolio investment, Root AI, pursued robotic strawberry picking using computer vision from MIT but got caught up in SPACs and did not work out.

A structural problem Bernstein identified for AgTech startups historically was lack of access to farmland, since most teams were based in San Francisco. A survey of startups found the average time to access an actual farm after raising venture capital was nine months. Reservoir addresses this directly by placing entrepreneurs onto working farms in the Central Valley. Bernstein said Reservoir Farms expects to be profitable next year and chose not to pursue cash-for-equity swaps in order to build a sustainable operating business first, given hard costs including farm and tractor leases.

Bernstein also raised a competitiveness concern for domestic growers, noting that USAID subsidized agricultural infrastructure in Peru as a deterrent to drug trade activity, that the South American blueberry industry was built substantially on that infrastructure, and that the US is now importing more fresh produce than ever. He said it is difficult for California and Michigan growers to compete because US policy effectively subsidizes agricultural production outside the United States.

Separate from the AgTech discussion, the hosts covered several other topics. On AI output quality, one speaker predicts design output from tools like Claude or Codex will improve from roughly 8.85 out of 10 to a 9 or 9.5 within a year, attributing the current gap partly to users accepting first-pass AI results without iterating. One speaker disclosed purchasing Figma stock, citing a valuation of approximately seven times sales and roughly 70 times earnings, a loyal designer customer base, and a roughly 10 billion dollar market cap giving it acquisition currency to move into adjacencies previously dominated by Adobe.

On education accountability, a US Department of Education policy would cut federal student loan access to programs where graduates do not out-earn comparable non-college workers. An estimated 800,000 students currently attend programs that would fail this test. One speaker argued colleges should be required to pay half of any defaulted student loan when a graduate cannot find work, and characterized the current system as colleges charging up to 250,000 dollars with no financial accountability tied to student outcomes. A dataset from an economics course illustrated AI-assisted cheating at scale, with roughly 90 percent of students appearing to have used ChatGPT on a take-home midterm. Student 52 scored 100 on the AI-assisted midterm but only 16.5 on the in-person final, the largest score gap in the dataset.