Japan's AI Unicorn Is Building What the Rest of the World Is Still Arguing About
I came across a CNBC piece on Preferred Networks this week and I want to talk about it because it is the kind of story that puts the AI conversation in a different frame than we usually get. While most of the debate in US tech circles revolves around generative AI, chatbots, and whether AI will steal everyone's job, Preferred Networks has been quietly building AI that solves things that are physically real. Autonomous trucks. Healthcare applications. Industrial robots. They are not building tools that write your emails. They are building tools that drive the truck at 3am when there are no drivers left.
Preferred Networks is Toyota-backed, Japan-based, and valued at over a billion dollars, which makes it Japan's largest AI unicorn. Founded in 2014, it has been working on applied deep learning for over a decade with a focus on what its CEO calls real-world problems. That framing matters to me because there is a version of AI investment happening right now that is almost entirely focused on abstractions, and then there is this version, which is focused on what happens when a physical system meets a computational one. The latter is considerably harder and considerably more interesting.
The trucking problem they are tackling is a good example of why this work is so important. Japan has an aging population and a declining number of truck drivers. The hours are brutal, the routes are long, and the supply of people willing to do it is shrinking. AI-powered autonomous trucking does not replace human truckers out of laziness or cost-cutting. It steps into a gap that is genuinely threatening the supply chain. That is a fundamentally different kind of problem than the ones most AI companies are solving for.
From an IT perspective, what Preferred Networks is building requires an entirely different infrastructure conversation than enterprise SaaS automation. We are talking about edge computing, real-time decision systems, physical sensor integration, and AI that has to work correctly in conditions that are not controlled environments. My lab is a nice climate-controlled server room. Their lab is a highway at midnight in the rain. Respect.
The CEO's goal of making the latest technology available in the shortest possible time is a philosophy I can get behind. There is a tendency in technology circles to develop things for the sake of developing them, then figure out the application later. Preferred Networks appears to be running that process in reverse. Find the problem, build the solution, deploy it into the world. That is the approach I have always preferred, and it is apparently also what Japan's most valuable AI startup is betting on. I will be watching this one closely.
https://www.cnbc.com/video/2025/03/24/how-japans-largest-ai-unicorn-is-shaping-he-future-of-deep-learning.html
