Inside The Silicon Mind · Oct 7, 2026 · 31m · 23m saved
With Ivan Poupyrev, Co-Founder and CEO of Archetype AI
About 85% of world GDP comes from physical industries such as manufacturing, construction and logistics, Ivan Poupyrev said, while most of today's AI was built to understand words.
Language models read text. Poupyrev said a turbine, a car or a power line produces measurements instead, and his company is building one foundation model, called Newton, to read that sensor data across every kind of equipment.
"Physical world does not speak in language." — Ivan Poupyrev
Ivan Poupyrev, Co-Founder and CEO of Archetype AI, on Inside The Silicon Mind, spent eight years at Google building radars and other complex sensors and shipping them in products, after earlier work at Disney and Sony.
The full interview is covered here so you can skip it. 31 minutes of audio, 8 minutes of reading.
Here are the 9 arguments that matter.
Key Takeaways
Physical AI works on measurements, not language, so a model has to learn from sensor data
The fastest gains will come from equipment that is already online, not from new robots
Poupyrev expects 10–20% productivity gains on existing infrastructure
One general model can replace the patchwork of single-purpose industrial software, as large language models did for text
Fewer workers are joining physical industries while demand for energy, chips and data centers grows
Large companies arrive late to new technology because management does not care until a new area earns about $100 million a day
AI raises the importance of every hire, because smaller teams leave no room for a weak link
1. 85% of GDP Is Physical
The host asked whether the next trillion-dollar company will be built on AI that understands the physical world. Poupyrev said he believes so, and gave a figure he called rough.
"If you think about it, 85% of the world economy and this is kind of rough now but being thrown around 85% of the world economy by GDP coming from physical world" — Ivan Poupyrev
He named manufacturing, construction and logistics, and said applying AI there to make operations more efficient, safer and more predictable will matter more than the online world.
