The Story Everyone’s Laughing At
A New Zealand startup called Halter just became one of the most talked-about companies in AI — and it has nothing to do with chatbots, image generators, or large language models.
Halter makes smart collars for cattle. Solar-powered. GPS-enabled. Connected to a smartphone app. A farmer draws a boundary on a map, and the collar enforces it — using vibration and sound cues to guide the animal away from the edge. No physical fence required. No farm hands. No dogs.
Peter Thiel’s Founders Fund just backed the company at a $2 billion valuation. That’s double where it was less than a year ago. The funding round is reportedly oversubscribed.
They trademarked the algorithm running the whole system. They called it the Cowgorithm.
Most people’s reaction to that word is a smirk. Which is probably the point.
What Halter Actually Does
Strip away the novelty and Halter is a genuinely sophisticated piece of engineering. The collar tracks location in real time, monitors health indicators, maps grazing patterns, and — crucially — learns the behaviour of individual animals over time.
Within days of wearing the collar, cattle stop testing the virtual boundary. The vibration becomes a conditioned cue. The fence exists entirely in software, enforced through learned response.
Farmers manage their entire herd from a phone. 700,000 animals across New Zealand, Australia, and the United States are already wearing these collars. The company charges between £4 and £6 per animal per month on a subscription model — which means the more cows wearing collars, the more predictable and locked-in the revenue becomes.
Halter’s founder, Craig Piggott, came from Rocket Lab — where he built spacecraft. He left to solve what he described as “the bigger unsolved problem.” Farming.
Why Thiel Backed It — And Why It Matters
Founders Fund has a pattern. Defence technology. Aerospace. Robotics. Companies that apply cutting-edge technology to industries that have barely changed in decades.
Halter fits that pattern precisely. Agriculture has lagged behind almost every other sector in digital transformation. Farmers have relied on manual processes and local knowledge for generations. Precision agriculture — using technology to manage farms more efficiently — is growing fast, but adoption has been slow and the landscape is littered with agtech companies that declared bankruptcy trying to convince farmers to change.
Halter broke through that resistance. And the reason it broke through tells you something important about where AI is actually heading.
The Real Moat Isn’t the Collar
It’s easy to look at Halter and see a hardware company. A clever device. A good app.
That’s not the investment thesis.
The moat is the data. Hundreds of thousands of animals. Years of behavioural patterns. Health indicators. Grazing data. Fertility cycles. A proprietary machine learning model trained on a dataset no competitor can replicate without starting from scratch and waiting years.
This is the template for what investors are now calling Physical AI — artificial intelligence embedded into the real world, in industries nobody thought were technology industries. Not a chatbot. Not a dashboard. An AI system that changes what physically happens in the world, collects data in the process, and gets more accurate and more entrenched with every passing month.
If you’ve been following the AI After 40 newsletter, this is the pattern worth watching. The most defensible AI businesses aren’t going to look like AI businesses. They’re going to look like cattle management software. Or luxury property logistics. Or agricultural supply chains. The AI is the layer underneath — invisible, proprietary, and increasingly impossible to dislodge.
The Part Nobody Wants to Say Out Loud
Here’s where it gets interesting — and where the smirk fades.
Consider what the Halter system actually does, described without the agricultural context:
A wearable device tracks location 24 hours a day. Real-time health and behavioural monitoring. An invisible boundary is drawn on a map. The subject receives a vibration cue when it approaches the boundary. Within ten days, the subject stops testing the boundary. It simply stays inside.
The collar is on a cow. The cowgorithm is trained on cattle.
But the underlying architecture — geolocation, predictive behavioural modelling, invisible compliance boundaries, conditioned response — is not inherently agricultural.
Peter Thiel, it’s worth noting, also co-founded Palantir. A company built on large-scale data aggregation, pattern recognition, and population-level behavioural analysis. The through-line between these investments is not farming. It’s the infrastructure of behavioural influence.
This isn’t a conspiracy. It’s just pattern recognition. The same kind the cowgorithm is doing, in a field in New Zealand, right now.
What to Take From This
Halter will almost certainly succeed as a business. The product works, the data moat is real, the market is enormous, and the subscription model creates compounding revenue. For investors and entrepreneurs, it’s a masterclass in where defensible AI value actually gets built — not in another chatbot wrapper, but in physical industries with proprietary data and high switching costs.
For everyone else, it’s worth sitting with the broader observation for a moment.
We are in an era where the most sophisticated AI systems are being embedded into the physical world — into farms, into cities, into logistics networks, into the devices people carry in their pockets. They learn behaviour. They shape it. They get better at shaping it.
The cowgorithm is genuinely impressive.
For cows.
Ads Meah writes about AI, luxury travel, and the intersection of emerging technology and real life. Follow on AI After 40 for practical AI insight for professionals over 40
