Following Harvey, the next $10B success for vertical AI could potentially be in the agriculture sector.

      TL;DR Harvey has surpassed an $11 billion valuation by integrating legal data with a foundational model. Agriculture's fragmented $500 billion data landscape has emerged as the next frontier for vertical AI, bolstered by USDA's $300 million agreement with Palantir, which supports the notion that industry-specific intelligence, as seen with GrowersTech, is essential for advancement.

      In the legal sector, Harvey reached an $11 billion valuation, while Legora is quickly pursuing opportunities in Europe. Abridge has developed a multi-billion-dollar enterprise that transforms clinical discussions into structured medical records in healthcare. In customer service, Sierra is now valued at over $15 billion, making it one of the fastest-growing AI companies ever. The structure in each example follows the same pattern: a foundational model enhanced by proprietary data, a domain ontology, and specific workflow logic for a complex industry. Vertical AI has emerged as a consistently promising investment in the current AI landscape.

      Investors are now curious about which sector will be next and the potential size of that market. One recurring response is agriculture.

      The underlying statistics highlight substantial potential. McKinsey suggests that connecting the fragmented data of agriculture could contribute $500 billion to the global GDP. U.S. crop farmers reportedly spend about $72 billion annually on seeds, fertilizers, and crop protection. The AI market in agriculture is projected to grow from $2.43 billion in 2025 to over $8 billion by 2031, according to Mordor Intelligence. A critical constraint lies in the data layer that informs every decision within the global food system, which is currently flawed.

      What Harvey demonstrated

      Harvey’s successful $200 million funding round and $11 billion valuation stemmed not from a superior foundational model, but from its surrounding layer of reliable legal data, an ontology reflecting the flow of legal work, evaluation frameworks aligned with legal standards, and context that ChatGPT cannot generate on its own. Its collaborations with A&O Shearman enabled AI tools for antitrust filings, fund formation, and loan reviews, while its partnership with LexisNexis integrated trusted content and workflow tools. This was not merely a story about models; it was fundamentally about context.

      According to the MIT Project NANDA's 2025 GenAI Divide report, 95% of organizations see no return from their GenAI efforts, not due to weak models, but because horizontal AI seldom transforms complex industries independently. Harvey addressed this challenge in law, Abridge in clinical documentation, and Sierra in customer service. Observers are now closely monitoring companies positioned to resolve similar issues in agriculture.

      The U.S. government has also committed $300 million to this cause.

      A $300 million endorsement from USDA

      In 2026, the U.S. Department of Agriculture initiated its “One Farmer, One File” program aimed at unifying systems within the Farm Service Agency, Natural Resources Conservation Service, and Risk Management Agency into a single record for farmers. Shortly after, USDA partnered with Palantir in a $300 million Blanket Purchase Agreement supporting the National Farm Security Action Plan.

      This partnership confirms long-held views among agri-tech professionals: agriculture faces infrastructure-level data challenges. Palantir offers a horizontal solution, but the vertical model that incorporates agronomic logic as its core foundation represents the future of this sector.

      This concept underpins GrowersTech, an Israeli-American consortium that unified the Agmatix data platform with GROWERS, a U.S. retailer-loyalty organization that captures transactional data within the American agricultural supply chain. Its main engine, Axiom, is founded on a neuro-symbolic AI architecture combining a knowledge graph of pre-trained agronomic ontologies with field-level data and transactional signals between manufacturers, retailers, and farmers.

      “Agriculture doesn’t have a data problem. It has an intelligence problem,” stated Ron Baruchi, the company’s CEO. “The data exists; what’s lacking is infrastructure that interprets its meaning.”

      Why generic AI falters in agriculture

      While a generic model may understand nitrogen, it cannot determine the ideal quantity, which varies based on growth stages, soil types, past crops, and upcoming weather conditions. Agronomic decisions require contextual insight, which generic AI solutions in agriculture have failed to recognize, rendering these tools ineffective.

      GrowersTech approaches the problem from a different angle. Its ontology—an intricate framework detailing the workings of agriculture—is pre-trained by agronomists before any customer-specific data is introduced. Key relationships among crops, soils, products, and outcomes form its foundational structure. New implementations adapt this ontology without needing to reconstruct it.

      Currently, the system processes over 1.5 billion field-trial data points sourced through collaborations with leading agricultural universities. The company’s research findings have been published in Nature, a prestigious recognition rarely attained by private agri-tech firms. The platform is already operational within global crop input manufacturers, major food and beverage supply chains, U.S. agricultural retail cooperatives, and governmental agricultural departments, addressing product performance forecasting,