Physical World Data Machine
Amitt@amitt
−2.9%
since Sep 30
The idea
I believe AI is running out of fresh digital data, so the next edge comes from whoever can capture new data from the physical world. This basket spans all three layers: the sensors and vision chips that let machines perceive, the robot and vehicle fleets that collect the data, and the companies already sitting on hard-to-copy physical-world datasets. It stays broad rather than concentrated, tilted toward businesses already selling into this now, with cash held back to add on dips over the next three to five years.
What it owns
- Every data-collecting machine needs sensors and vision systems, so makers of these components profit across all robot and vehicle platforms. · Teledyne TechnologiesIt makes the cameras and sensors that provide vision for drones, satellites, factories and defense systems., NVIDIA CorporationIt supplies the chips and robotics platforms that serve as the compute layer for most machines collecting physical-world data., Sony Group CorporationIt dominates image sensors, supplying the chips inside most of the world's cameras and increasingly cars and robots., Keysight TechnologiesIt sells the instruments that measure and validate signals, chips and networks in every sensor and radio system., Cognex CorporationIt provides machine vision cameras and software that let factory and logistics machines see and understand what they handle., ON SemiconductorIt makes the image sensors inside cars and industrial machines where photons first become digital data., Ambarella, Inc.It makes chips that process vision at the edge, turning physical-world data into actionable information right on the device., Ouster, Inc.It sells lidar sensors that provide depth sensing for machines moving through the physical world safely.31%
- Machines that operate in the physical world generate fresh data that AI models need to train on. · Tesla, Inc.Its fleet of millions of cars logs the physical world daily, creating one of the largest sources of real-world driving data anywhere., Intuitive Surgical, Inc.Its surgical robots record data from millions of procedures inside patients, creating a dataset that cannot be scraped from the internet., Deere & CompanyIts autonomous tractors and sprayers collect field-level data across millions of acres, a physical-world dataset digital scraping cannot reach., Rockwell Automation, Inc.Its controllers and software sit on factory floors worldwide, generating operational data where machines meet materials., Symbotic Inc.Its warehouse robots generate fresh physical-world data with every pick and move inside customer facilities.25%
- Scientific instruments produce the highest-quality physical data, the kind AI for chemistry and biology requires for training. · Thermo Fisher ScientificIt makes instruments that generate lab-grade data about molecules, cells and materials at the highest precision., Agilent TechnologiesIts instruments turn chemistry and biology into data, from drug labs to food testing, creating physical-world information., Danaher CorporationIt owns companies that measure and process biology across drug manufacturing and diagnostics, generating lab-grade data., Bruker CorporationIt builds instruments that measure matter at the molecular and atomic level, data only obtainable by touching the physical world.17%
- Companies holding large proprietary datasets of real-world operations become more valuable as high-quality training data grows scarce. · Samsara Inc.It holds one of the largest datasets of physical operations data from trucks, sites and equipment across hundreds of thousands of customers., Trimble Inc.It owns precision positioning and mapping data across construction, agriculture and transport, knowing exactly where things are., Garmin Ltd.Its watches and devices collect body and motion data from millions of people, capturing physical-world data about humans., Planet Labs PBCIt photographs the entire Earth daily, creating a physical-world dataset that cannot be reproduced by web scraping.16%
- Cash11%