Physical AI
AI is moving from the cloud into the physical world.
The first AI era was built in the cloud. The next will operate in the real world — where sensing, data movement, memory, latency, power and thermals must be optimized as one system.

Always-aware systems need a new architecture
Physical AI has a data problem
Most compute architectures weren’t designed for systems that continuously sense, interpret and respond. More sensors, higher resolutions and always-on operation now create growing pressure on bandwidth, memory, power and thermals.
Centralized-only AI architecture does not scale efficiently
FotoNation shortens the pipeline itself: bringing memory closer to compute, prioritizing bandwidth over frequency, and enabling intelligence to operate locally with extreme efficiency.
Cognitive Compute™ puts computation where it matters
Process high-rate perception close to the sensor, keep transient data and computation close together, and use centralized AI compute where global context, larger models or multimodal reasoning add value.
Our Target
Humanoid Robotics

Human-like machines require human-like perception.
Humanoid robots concentrate the hardest Physical AI constraints into one mobile system. They must continuously see, hear, sense, understand and act while operating within strict power, memory and thermal limits.
FotoNation products turn raw multimodal sensing into compact, useful information for higher level intelligence — combining distributed perception, near-memory cognition and centralized AI compute where each is most effective.
Humanoids are the proving ground, not the limit.
TriSilica® is built to address a challenge shared across Physical AI: turning continuous sensor data into real-time intelligence without unnecessary movement, power or thermal load.
By processing perception close to where data is generated, keeping transient data and compute together, and using centralized AI where broader context adds value, the same Cognitive Compute™ architecture can scale across drones, industrial robotics, automotive, automation and AI in an Edge Box.












