Meet the Bay area startup zapping weeds with superheated veg oil

Tensorfield Agriculture is pioneering the commercial adoption of thermal micro-jetting for weed control and thinning in vegetable crops. But how does zapping weeds with superheated vegetable oil compare to other approaches from laser weeding and herbicides to hand-weeding?

AgFunderNews (AFN) caught up with cofounder and CEO Xiong Chang (XC- pictured above second from left) at the World Agri-Tech summit in San Francisco to find out more.

AFN: Give us the 60-second elevator pitch

XC: We’re a team of computer scientists and engineers with backgrounds in robots and computer vision based in the Bay Area. We kill weeds using vegetable oil with a focus on specialty row crops such as carrots using tech developed at UC Davis and the University of Bonn-Landtechnik.

Our machines operate semi-autonomously with each nozzle array covering one acre per hour whether there’s 50,000 weeds in that acre or 500,000 at 1.2 miles per hour, with 1/4 inch precision.

AFN: How does this approach compare to laser weeding?

XC: Laser weeders work wonders but they’re very expensive and very heavy, so you can get issues with soil compaction and bed defamation. Our machines are about a quarter of the weight, use about a fifth of the energy, and have a travel speed that’s two to four times faster.

AFN: How hot does the oil have to be to get the job done?

XC: We’ve found that 160 degrees Celsius works, but in really cold weather that can be a bit less effective, so you sometimes have to go up to 180 degrees at the point of delivery. It varies a little depending on the weather.

AFN: How precise can you be?

XC: Across the standard 80-inch bed there are 232 nozzles, each controlling an area a few millimeters wide.

AFN: What if the ground is uneven?

XC: We’ve developed a self-leveling mechanism that keeps the sprayers within three inches of the target, while depth cameras map out the bed to account for any irregularities in height that need to be compensated for in our spray targeting algorithms. As a result, we can hit smaller targets with greater accuracy.

AFN: What’s the ROI for the grower/farmer?

XC: For organic carrot growers, the saving comes from labor, because they’re hand-weeding, whereas for something like conventionally grown lettuce, the ROI would be from savings on herbicides. Even if growers are still hand-weeding weeds that are too close to the crops and they want a safety buffer, they are still saving money [by zapping most of the weeds with hot oil].

AFN: How do you heat up the oil safely in a moving vehicle?

XC: We have a commercial deep fat fryer attached to the machine which heats recirculating food grade industrial thermal fluid [using propane], which is completely sealed from the rest of the system. This gets sent through to heat exchangers and heats the canola oil at the point of delivery.

One of the issues with heating canola oil is that if you keep recirculating it, it will oxidize and polymerize, so we need to make sure that it’s heated for the shortest period of time possible, which means we heat on demand at the point of delivery.

AFN: What’s the capacity of one of the vehicles before you have to re-fill the veg oil?

XC: That depends on weed pressure—the more weeds, the more oil you’ll use—but the design goal was for 10 acres of use before having to refill.

AFN: Your vehicles are semi-autonomous?

XC: They are self-propelled and the see-and-spray decisions for the micro jetting are completely autonomous. But we have a human pilot on board that controls the motion of the vehicle, although that step could be automated away.

AFN: How easy are the machines to operate?

XC: Very easy. The interface is actually a PlayStation 4 controller, and because both of us [Chang and cofounder and CTO Sandeep Mirchandani] grew up playing Mario Kart, we’ve tried to map the controls as similarly as we can to that, so it’s actually quite intuitive!

AFN: What’s the sweet spot for your technology?

XC: It’s particularly well-suited for any crop where hand weeding costs are high and you can’t go in with a mechanical solution, such as organic, high-density crop such as carrots, but also conventionally grown spring mix or baby leaf spinach. Any case where hand weeding bills are in excess of $250 an acre makes sense for us.

AFN: Is this just for the first weeding pass?  

XC: We can only really access that first flush of weeds; once there’s significant crop canopy occlusion, this solution will end up killing carrots. But what it does is severely reduce the amount of hand labor required for that first weeding pass. So in a carrot, which is 75 to 90 days from seeding through to harvest, you might do two passes. And how many weeds you get on that first pass [using hot oil] has an impact on how much you have to spend [using human labor, for example] on the second one.

AFN: What’s your business model?

XC: Our initial go to market model is a $50 fixed fee per acre, plus we charge half a penny for every weed that we kill. So if we go into a relatively clean field, we’ll charge less, but either way, the grower consistently gets savings.

AFN: How has the tech evolved since you started out?

XC: Our first prototype that actually hit a customer field was in 2020. It had an oil dosing and heating mechanism on it, but it wasn’t anywhere near the level that was needed to be able to do useful work. However, it de-risked the tech stack in that we showed we could tell the difference between crops and weeds.

Our next prototype, v2, was in 2022, where we proved we could heat oil on the fly and dose that onto specific targets. However, we had issues with uneven ground with having a static distance from the nozzles to the bed, so we had to go back to the drawing board and build an articulating payload in v3 so the nozzle array could move to accommodate the variability we see out in fields. We also increased the resolution [for more precise spraying].

Now, we have got paid trials with growers of organic carrots.

For the next machines that we build that are covering three 80-inch beds at once, we’re projecting that 33 days of operation will pay off the capex that’s required to build a machine. So it is a high margin business in the specialty high density space.

AFN: Can you get your machines to go faster?

XC: At the moment, we’re traveling at 1.2 miles per hour, independent of weed pressure. But the ability to go 5-10 times faster is purely just a function of can we get nicer solenoid valves on board. So we certainly have the potential to go faster, although the weed detection starts to get a bit dicey once you go over 5x our current speed.

AFN: How have you funded the company so far?

XC: We have great backers including SOSV, but we’re now at that inflection point where having provided value to a customer we’re looking to build more machines to prove out the unit economics, so we are looking to raise more money.

AFN: What are the biggest barriers you are facing?

XC: There have been various amounts of healthy skepticism as to whether thermal micro-jetting works because many growers have been let down by [ag robotics] before. One of the first growers we spoke to was an organic carrot grower who said that he had tested other [precision spraying] systems but had gone back to hand weeding.

AFN: What are your plans for the year?

XC: In the immediate term, we want to build five of these machines and get to $5 million of ARR (annual recurring revenue] to show that the unit economics work.