Simulation is becoming the new standard for ag robotics

Good morning. Eternal.ag exited stealth this past week with €8 million and a fully autonomous tomato harvester built entirely in simulation before it ever touched a real vine. It's the latest sign that the way ag robots get built is changing.

In today's newsletter:

Greenhouse robotics has a graveyard problem.

Startups raise money, build a robot that works great in demos, deploy it commercially, and then watch it fall apart. The gap between "works in the lab" and "works 22 hours a day in a real greenhouse" has humbled a lot of smart people.

The industry is starting to figure out why. And the answer is changing how the best teams build.

Eternal.ag, a Cologne-based startup, exited stealth this past week with €8 million and a fully autonomous tomato harvesting robot called Harvester. The funding and the launch are interesting. But what's more interesting is how they built it.

Instead of learning in the field, Eternal.ag trained and tested their robots entirely inside virtual greenhouse environments before deployment — what they call simulation-first development. Failures that cost nothing. Iterations that used to take months now take days.

Once deployed, every action feeds data back into the system. The robot gets smarter with every harvest.

And the timing matters. European greenhouse labor availability has fallen roughly 30% since 2010 with no signs of reversing. This is a pattern playing out across ag sectors globally. Growers can't wait years for robots to learn on the job anymore.

The broader shift happening across ag robotics is that the best companies are treating the virtual environment as seriously as the physical one. Train in simulation. Deploy with confidence. Improve with real data.

Getting farmers to adopt new technology is hard. Sales cycles are long, trust is low, and the upfront costs are big.

The 2026 Farm Bill is trying to change that.

Two weeks ago the House Agriculture Committee advanced the Farm, Food, and National Security Act of 2026. Buried inside it are a few provisions that should have every autonomous ag startup paying attention.

First, the bill reimburses farmers up to 90% of the cost of adopting precision agriculture and AI technologies through the existing EQIP program. The normal cap is 75%. That extra 15 points is the government essentially saying — we'll absorb most of the risk so farmers don't have to.

Second, AGARDA gets reauthorized. Think of it as DARPA but for agriculture. A high-risk, high-reward research fund specifically designed to back the kind of moonshot technology that private VCs won't touch. It's been chronically underfunded since it was created in 2018. The new bill pushes to change that, and mandates the USDA enter agreements with the Department of Defense and the Department of Energy to share autonomy research.

DARPA and the DoD, sharing data with agriculture. That's not a small sentence.

The bill still needs a full House vote before Easter and Senate approval after that. Nothing is done yet.

But the direction is clear. The government is moving from passive observer to active accelerator in autonomous agriculture. For startups that can show their technology reduces chemical inputs or improves farm resilience, there's now a federal program designed to put it in front of farmers.

The best sales rep just got a much bigger budget.

Meet Emma. She doesn't take breaks, doesn't mind the heat, and works through the night without complaints.

Emma is an AI scout for vineyards built by Budbreak Innovations, a startup founded in 2025 by Jonathan Moon.

Emma moves row by row, vine by vine, capturing high-resolution images of every single plant. Her computer vision detects early signs of disease, maps plant health, and even forecasts yields. All of this in real time.

The problem she's solving:

Viruses are one of the biggest threats to winegrape farmers. All it takes is one missed vine and an outbreak can spread across an entire vineyard, causing millions in losses. By the time a human scout spots it, a lot of the times it's too late.

Emma is already deployed in 14 vineyards and orchards across California and New York, including pilots with some of the largest vineyards in Northern California and NASA. She's been covered by NPR and USA Today.

Budbreak recently joined Reservoir Farms in Sonoma (the world's first on-farm robotics incubator) where they'll develop and validate Emma alongside other agtech leaders.

The bigger picture:

Early disease detection could reduce pesticide use by up to 90%. Less waste, better yields, healthier land.

One cool thing: Jonathan is currently traveling to any farm with Emma and scouting 5 acres for free. If you know a vineyard that could use an AI scout, this is the time. This is the link to get your free scan.

RaaS stands for Robotics as a Service. And it's changing how farmers access technology.

The old model was simple. You buy a machine, you own it, you use it. A tractor, a harvester, a sprayer. Big upfront cost, but it's yours.

The problem is that agricultural robots are expensive. Most farming operations, especially small and medium ones, can't justify that kind of capital investment for technology that is still evolving fast.

RaaS flips that model. Instead of buying the robot, you pay for what it does. Per acre treated, per hour of operation, per season. The company owns the machine, handles the maintenance, and sends updates. You just pay for the results.

Think of it like the difference between buying a car and using Uber. You get where you need to go without owning the vehicle.

The company deploying the robot stays responsible for keeping it running and improving it over time. For the farmer, it shows up, does the job, and leaves.

It's worth noting that some operations eventually buy the equipment outright once they see the returns. RaaS is often the entry point, not the permanent arrangement.

That's all for this week.

See you next Sunday.

—Enzo, Founder of Field Autonomy