Harvesting deep tech: How startups are putting Google Cloud AI on the frontiers of agriculture

We are in an agricultural revolution that’s never stopped growing. Farmers today face complex issues like population change, evolving supply chains, growing costs, and climate change. Yet just as in generations past, rapid technological advancements always seem to keep pace, giving us new opportunities to operate more efficiently and sustainably.

Already AI is taking its place alongside the tablets and tractors that are as plentiful as cornstalks and soy beans in the field, and a group of fast-moving startups are the newest farm hands. Some 17% of the 5,500 farmers McKinsey surveyed said they are currently using AI to help with farm-related tasks, while 72% expect it to have some impact within the next three to five years.

At Google Cloud, we’re working with a number of exciting founders in this area, helping to keep crops healthy, boost the harvest, maximize sustainable practices, monitor the weather, and make up for worker shortages. It’s one of the most exciting and promising areas of AI for good we’re seeing at a time when the world needs all the help it can get in the fields..

To see first hand how a number of these startups are uniquely approaching and reinventing aspects of agriculture with Google Cloud’s unique AI stack, our team joined them at Reservoir Farm’s inaugural Ruggedize, an event dedicated to showcasing advancements by the industry and startups to apply frontier AI to agtech.

We often like to talk about deep technology in the startup space, the stuff that’s really pushing the boundaries of what’s possible. Here we got to see just how big an impact there can be for AI from farm all the way to our table. And on the flip side, there may be no better proving ground for physical AI and advanced models than the fields and orchards of the food sector.

Let’s have a look at just how rugged this technology can be, and how agtech might show the way for other industries.

• Bonsai Robotics builds autonomous hybrid-electric robots and retrofit kits powered by advanced vision AI to tackle tough outdoor tasks in dusty, GPS-denied environments. They leverage services like the Google Maps API and our Gemini models to help orchestrate connected robotic fleets at scale, as well as AI infrastructure to train their world model and foundation model

• Hippo Harvest develops autonomous greenhouse farming systems powered by robotics and AI to cultivate fresh produce using a fraction of traditional resources. They leverage autonomous mobile robots, or AMRs, and precision computer vision to micro-dose water and nutrients at the plant level, scaling climate-resilient indoor agriculture while drastically reducing water use, fertilizer run-off, synthetic pesticides and carbon emissions. Google Cloud provides the data and AI backbone: Cloud Storage holds the plant imagery and sensor data models are trained on, Cloud GPU instances run computer-vision inference on every harvest, and a cloud Postgres replica powers analytics independently of the farm's on-site systems.

• Orchard AI gives farmers the ability to effectively understand every plant they grow through an AI-powered camera system mounted onto existing tractors and farm vehicles. Able to scan millions of trees, vines, and plants, Orchard captures precise data on crop yield, size, health, disease, and more. Leveraging the Gemini API and leading models such as Gemini 3.8 Flash, Orchard transforms this plant-level intelligence into practical recommendations that help farmers make better decisions across tens of thousands of commercial acres. This saves on chemical sprays, labor inputs, and operational efficiency while making the farm more profitable and sustainable.

Katherine Hecht, a forward deployed engineer at Google Cloud working with agtech and robotics startups, believes the future of physical AI spans a continuous spectrum of intelligence. As she shared at Ruggedized, this new chapter in the agricultural revolution means connecting planetary foundation models, simulated physics environments, and real-time robotic action:

“Agriculture is one of the best proving grounds for physical AI because of its challenging and unforgiving nature. You're operating off the grid, in extreme and unpredictable weather, under strict compute and power constraints, and with spotty connectivity. Rather than isolating field machines to onboard vision alone, teams can draw on an entire toolkit to provide boots-on-the-ground intelligence.

“There’s planetary and atmospheric physics models like WeatherNext, NeuralGCM, and AlphaEarth; interactive world simulations and physics engines like Genie and MuJoCo to explore cause-and-effect safely; and lightweight edge models like Gemma alongside Gemini Robotics ER 2.0 and VLAs for real-time action in the field.”

Agtech, and robotics more broadly, represent some of the fastest growing use cases for AI from startups. This is particularly notable, considering the myriad challenges that come with deploying AI and hardware in rugged, often rural environments.

These are also areas where Google Cloud’s AI stack plays a key role. Today, Google Cloud is providing robotics startups with a unique choice of AI accelerators, models capable of multimodal understanding and generation, data tooling, agentic platforms, and infrastructure, which are powering many of the use cases on display at Ruggedize.