One tap and 600,000 cows start walking to milking stations: Inside the AI farm startup Peter Thiel backed at a $2 billion valuation

Somewhere on a farm in New Zealand, a dairy farmer is still in bed at 4:30am. In the old world, he'd already be pulling on boots, whistling for his dog, and heading out to round up 600 cows from scattered paddocks. That would take forty-five minutes. In this world, he reaches for his phone, taps a single button, and puts it back on the nightstand.

Outside, 600 animals begin to move on their own. No dog. No farmhand. No fence. Just a small device around each cow's neck that makes a sound, and animals that have learned, just as a dog learns to sit, that the sound means walk this way. The milking shed fills itself. This is Halter, and Peter Thiel just valued it at $2 billion.

Building a physical fence costs $15,000 to $40,000 per mile. It is permanent, needs constant maintenance, and does nothing for other problems such as the daily labour of herding, health issues that go unnoticed, and fertility windows that pass unseen. For decades, the fundamental mechanics of running a farm had not changed. Farmers relied on fences, dogs, and their own observation. Serious technology to fix these inefficiencies did not exist until now.

How one tap gets thousands of cows walking to milking stations

A farmer opens the Halter app, sees a satellite view of their land, and draws a boundary with their finger, similar to circling an area on Google Maps. That circle becomes the virtual fence. There are no posts or wires. The boundary is transmitted as GPS coordinates to every collar in the herd within seconds.

Each collar tracks the cow's location four times per minute, continuously. When a cow approaches the boundary, the collar emits a sound on one side of her neck. The cow has been trained to associate that sound with turning away. She turns, the sound stops, and she continues grazing. This process happens automatically, dozens of times a day, without human involvement.

When the farmer wants to move the herd to the milking shed, a single tap activates the system. The AI calculates the position of every animal, determines a logical path for each one, and sends directional cues to every collar in a coordinated sequence. Cows at the back begin moving first, while slower animals receive stronger prompts. The entire herd reaches the shed within minutes, guided by a system that knows the precise location of every animal.

The Pavlovian trick that makes it all possible

The core of the system is not GPS or AI, but behavioural science. When a farm first adopts Halter, the onboarding process takes about a week. The farmer walks with the herd while the collars emit sound cues, guiding cows physically in the correct direction. Each time the cow responds correctly, the sound stops. Over time, the physical guidance is no longer needed. Within days, the association becomes automatic. Within a week, the farmer no longer needs to be present in the paddock.

Cows are naturally social animals. Once a few respond to the signals, the rest follow. The herd learns collectively, making the process faster and more efficient. Hundreds of cows can be trained within days, responding reliably to an algorithm without direct human supervision.

The Cowgorithm and continuous monitoring

The collar does more than guide movement. It continuously monitors each cow. It tracks chewing behaviour, body temperature, walking speed, rest patterns, and fertility signals. This data creates a behavioural profile for every individual animal.

Halter’s AI, known as the Cowgorithm, is trained on data from hundreds of thousands of cows over several years. It understands what is normal for each animal. If a cow begins chewing less, moves differently, or shows subtle signs of illness, the system flags it. In many cases, this happens 24 to 48 hours before visible symptoms appear. Early detection can significantly reduce treatment costs and improve outcomes.

Fertility tracking is another major advantage. A cow’s breeding window is short, often lasting only 12 to 18 hours. Missing it delays the cycle by weeks. The system detects behavioural changes associated with this phase and alerts the farmer at the right moment, improving efficiency across large herds.

The numbers behind the growth

Across the United States, Halter users have created more than 11,000 miles of virtual fencing, comparable to the perimeter of the continental United States. This has resulted in estimated savings of $220 million in physical fencing costs. The system is now used on more than 700,000 cattle across New Zealand, Australia, and 22 US states.

The company operates on a subscription model, charging between $5 and $8 per cow per month. The collars are leased rather than sold. A farm with 500 cows may pay around $3,000 per month, while larger operations pay significantly more. Once a farm restructures its operations around the system, switching away becomes difficult due to the reliance on digital infrastructure.

Peter Thiel and the $2 billion bet

Peter Thiel’s Founders Fund has been involved since the early stages, participating in Halter’s 2018 Series A round when the company was still in its early development phase. This is not a recent discovery but a long-term investment.

After observing the company’s growth to hundreds of thousands of managed animals, Founders Fund is now leading a funding round that values Halter at approximately $2 billion. The decision reflects confidence in both the technology and its long-term potential.

Why this is bigger than farming

At its core, Halter has built a system that coordinates thousands of living animals in real time using data, sensors, and behavioural cues. While the current application is in agriculture, the underlying approach has broader implications.



The same principles could be applied to areas such as wildlife management, logistics, and robotics. More importantly, it represents a shift in how physical environments are managed. Tasks that once required constant human effort can now be guided by software systems operating continuously in the background.

There is also a significant labour dimension. Agriculture has long depended on physically demanding work and increasingly scarce labour. Systems like Halter reduce the need for manual intervention while improving efficiency.

A former rocket engineer returned to farming and applied engineering principles to a traditional industry. Years later, that decision has resulted in a company valued in the billions. It suggests that even the oldest systems can be transformed when viewed through a different lens.