ADI Frames the Next Phase of AI: Machines That Must Sense, Decide, and Act

Artificial intelligence has mastered the digital domain. The harder frontier, according to Analog Devices Inc., is the physical one — and the Wilmington, Mass.-based company is laying out its case in a new sponsored whitepaper on the race to make AI physical.

From Understanding Language to Understanding Reality

For manufacturing professionals, the distinction matters. Software intelligence operates in controlled digital environments; physical AI does not.

"AI has become remarkably good at understanding language, images, and code. The next challenge is harder — understanding reality," ADI said.

"Physical AI moves intelligence out of digital environments and into machines that must sense, decide, and act in the world."

That shift carries engineering consequences. Physical AI systems face imperfect measurements, timing constraints, and changing conditions — realities familiar to anyone who has specified sensors for a production line. These systems must interact with the world, adapting to changing conditions and responding to situations that cannot be fully predicted in advance.

Where Success Will Be Decided

ADI's position is blunt: the model alone won't carry these systems.

"Success will depend on more than model performance. It will require advances across sensing, simulation, computing, connectivity, and control," the company said.

That framing aligns with ADI's technology focus — sensing, simulation, computing, connectivity, and control for physical AI systems — and with its broader semiconductor portfolio, which the company says drives advancements in automation, robotics, mobility, healthcare, energy, and data centers. Founded in 1965, ADI has staked its claim at the boundary between the analog and digital worlds, which is precisely where physical AI systems live or fail.

The stakes explain why the rise of physical AI has become one of the most closely watched areas in technology for manufacturing.

"The next phase of physical and embodied AI is not simply about producing better answers. It is about predicting and producing useful actions in the presence of imperfect information," ADI said.

ADI at RoboBusiness 2026

The event marks 20 years of RoboBusiness.

The takeaway for engineers: physical AI is not a software problem with hardware attached. It is a systems problem, and the sensing-to-control chain beneath the application layer is where outcomes get decided.