Applications

Physical AI

The same infrastructure cycle that reshaped the data center is now echoing into tens of billions of devices, machines, vehicles and robots. We expect this next wave, Physical AI, to scale faster than any technology wave before it.

Devices and machines that sense, think, act & communicate in real time

Physical AI runs on a closed loop known as STAC—Sense, Think, Act, Communicate: capturing signals from the real world, deciding locally in real time, acting with precision and exchanging data securely across distributed systems. STAC is the core architecture behind every scalable Physical AI system and maps directly to the technologies GF has been investing in for more than a decade.

Built on ultra-low-power CMOS, RF, silicon photonics, advanced packaging, embedded non-volatile memory and now MIPS and ARC real-time RISC-V processors, GF delivers an end-to-end platform for intelligent autonomous systems backed by a resilient global manufacturing footprint across the U.S., Europe and Asia.

Why Physical AI

The past decade of AI infrastructure build-out has set the stage for what’s next: pushing intelligence onto the device, where milliseconds and milliwatts matter. Physical AI enables systems to sense the physical environment, think deterministically, act with precision and communicate securely across distributed networks. This shift is driven by the need for lower latency, higher reliability and greater energy efficiency than cloud-only approaches can deliver.

As intelligence moves into devices, vehicles and robots, GF delivers the semiconductor technologies that power this new era of real-world, real-time computing.

Proven momentum where Physical AI is already taking shape

Physical AI is already transforming industries where real-time decision-making, safety and energy efficiency are non-negotiable, and where GF technology is embedded.

Automotive: From ADAS sensor fusion and radar to efficient domain controllers, combining safety‑capable, deterministic compute with robust RF connectivity for split‑second performance on the road

Robotics: Multimodal sensing, distributed intelligence and reliable actuation for humanoids and autonomous systems, delivered through ultra‑low‑power compute, fast embedded memory and advanced packaging

Healthcare: On‑device imaging and monitoring systems to process signals locally for instant response, privacy and power efficiency in constrained form factors

Next‑generation IoT: Wearables, AR/VR and smart home devices that balance responsiveness, battery life and connectivity for always‑on experiences

GF + MIPS + ARC: A holistic technology partner for Physical AI

GF and MIPS combine real‑time, event‑driven RISC‑V compute with GF’s differentiated technologies to accelerate physical AI innovation. Together, they deliver a deterministic, safety‑capable compute platform engineered for devices that must sense, think, act and communicate in real time.

MIPS brings a software‑first microarchitecture and proven multithreaded heritage ideal for physical‑world workloads, while GF provides FDX™ FD-SOI, inFET, RF‑SOI, SiGe, BCD/GaN and advanced packaging to optimize power, performance and connectivity. This pairing enables a closed‑loop intelligence stack that customers can tailor to workload, power and cost.

Read a Q&A with MIPS CEO Sameer Wasson (opens in a new tab)

GF business webinar series: Enabling the evolution of edge to Physical AI

This webinar showcases how GF MIPS and CMOS technologies are driving the evolution of smart, autonomous, and connected devices from edge to physical AI applications.

Watch webinar

Latest news & insights

Physical AI

Frequently asked questions

Physical AI brings intelligence into the physical world—machines and devices that sense their environment, process information locally, make decisions and act in real time. It reduces dependence on the cloud for latency‑sensitive tasks and enables autonomy at the edge.

As applications like humanoid robotics and advanced autonomous systems move from research into real‑world deployment, demand for real‑time embedded control and energy‑efficient, on‑device intelligence is accelerating. A massive data‑center build‑out has primed the ecosystem; the next cycle pushes intelligence to billions of devices.

We provide the semiconductor foundation—ultra‑low‑power compute (FDX FD-SOI & FinFET), RF and connectivity, advanced packaging and photonics, embedded memory and deterministic real‑time compute via MIPS to build reliable, secure, low‑latency systems at scale.

Edge AI describes where computation happens, running AI models close to the data source to reduce latency and reliance on the cloud. It delivers faster processing, but its role is primarily analysis.

Physical AI goes further. It brings intelligence directly into devices and machines so they can sense their environment, make deterministic decisions, act through motors or actuators and communicate in real time. It’s a full closed-loop system built for real-world operation, not just local inference.

Edge AI is a building block of Physical AI. It provides the localized compute foundation, while Physical AI adds the capabilities needed for autonomous, real-time action in the physical world, like precision sensing, ultra‑low‑power compute, embedded memory, secure RF connectivity and reliable actuation. Physical AI doesn’t replace edge AI; it extends it into real-world impact.