Tokens Built The Digital AI Business. Physical AI Needs A Different Unit
Reported by Forbes1 source ↓
Archetype AI COO Brandon Barbello argues that industrial AI pricing should reflect assets and applications, not token consumption, because sensor data is continuous and systems can run on customer-owned hardware.
Key points
- Barbello proposes licensing physical AI by asset, machine, facility or application instead of tokens
- A municipal customer’s intersections generate about 43 GB of camera data per day each, Archetype AI estimates
- Archetype AI says one wind-turbine deployment found nine previously unknown failure modes across 63 turbines
- Per-asset licensing may be less economical for equipment that needs only occasional analysis
Details
In a Forbes contributed article, Archetype AI co-founder and COO Brandon Barbello argues that token-based pricing developed for digital AI may not suit industrial systems. Sensors generate data continuously, he writes, so billing for each unit of processing can make costs rise with equipment runtime and leave budgets harder to predict.
Barbello says physical AI can run on customer-owned hardware near where sensor data originates, while cloud computing remains useful for training, fine-tuning and fleet-level analysis. He cites an Archetype AI municipal customer whose intersections produce about 43 gigabytes of camera data per day each; the company estimates moving the footage to the cloud would cost $850,000 to $1 million annually before AI processing.
He also cites an Archetype AI deployment that analyzed data from more than 40 sensors across 63 wind turbines and identified nine previously unknown failure modes. Barbello says licensing by asset, machine, facility or application could make costs more predictable, but may be less economical when equipment needs only occasional analysis.
Written from coverage as of
Images
- Forbes