GlobalFoundries, the publicly traded semiconductor foundry, has announced customer availability of its UX platform technologies. The platform is designed to power a new generation of edge computing, connectivity, and what the company calls “Physical AI” applications.
Physical AI, in plain terms, is the idea of embedding artificial intelligence directly into the devices and objects people interact with in the real world, rather than relying on cloud-based processing. Think AI-enabled glasses, smart hearing devices, connected appliances, and software-defined vehicles.
What the UX platform actually does
The UX platform is built to enhance intelligence, sensing, and detection capabilities in physical devices. It integrates with GlobalFoundries’ existing technology stack, which includes RF, imaging, ultra-wideband, and audio functionalities, with a particular emphasis on multimode sensing applications.
The platform fits into what GlobalFoundries describes as its STAC architecture: Sense, Think, Act, Communicate. It’s a framework for building end-to-end systems where a device can perceive its environment, process that information locally, take action, and communicate results, all without round-tripping data to a distant server.
Under the hood, the UX platform draws on GlobalFoundries’ ultra-low-power CMOS/FDX process technology, RF-SOI and SiGe connectivity solutions, and advanced packaging capabilities.
The $18 billion bet on Physical AI
GlobalFoundries has identified a serviceable addressable market exceeding $18 billion for Physical AI by 2030. The design wins the company has cited span several categories: AI-enabled glasses, hearables, smart home appliances, and software-defined vehicles.
For context, GlobalFoundries trades on Nasdaq under the ticker GFS.
Building the IP moat
GlobalFoundries completed the acquisition of Synopsys’ ARC processor business, a deal that brought more than 150 patents and roughly 300 customers into GlobalFoundries’ orbit.
On the technology roadmap, the company is also working on embedded RRAM (resistive random-access memory) on its 22FDX+ platform, with volume production expected in 2026. Embedded non-volatile memory is a key ingredient for edge AI chips because it allows models and data to be stored directly on the processor, eliminating the latency and power cost of accessing external memory.