GlobalFoundries unveils 7nm-class FDX Fusion chip platform with 2028 Dresden manufacturing target

GlobalFoundries is making a significant push into the physical AI market with a new 7nm-class FDX Fusion chip platform, targeting real-time sensing, processing, and edge connectivity for devices that need to think fast and run cool. The company plans to begin manufacturing at its Dresden, Germany facility in 2028, timed to coincide with a major capacity expansion that has been years in the making.

The announcement puts GlobalFoundries squarely in competition for a market it believes could exceed $18 billion by 2030, centered on what the industry is calling physical AI: the layer of intelligent processing that sits between raw sensor data and real-world action.

What FDX actually is, and why it matters for edge AI

FDX stands for Fully Depleted Silicon-on-Insulator, a planar chip architecture that differs meaningfully from the FinFET processes used by leading-edge fabs like TSMC and Samsung. Think of FinFET as a three-dimensional transistor that stands upright on the chip, and FD-SOI as a flatter design that sits on an insulating layer, which dramatically reduces power leakage.

Physical AI applications, whether in a car’s sensor array, an industrial robot, or a connected medical device, often run on constrained power budgets in environments where heat is a problem. GlobalFoundries says its FDX platform supports operation up to 150 degrees Celsius, which covers most automotive and industrial use cases.

The company’s FDX platform has been in volume production since 2019, and more than 3 billion FDX chips have shipped across that period. The new 7nm-class designation is a marketing claim about performance parity with more advanced node architectures rather than a literal transistor size.

The next step in that roadmap is the UX platform family, which includes 40UX and 22UX variants. GlobalFoundries plans to make production design kits available to customers in September 2026, giving chip designers roughly two years to develop products before manufacturing is expected to come online at scale.

The Dresden expansion is where the money is

The platform announcement is strategically linked to a major infrastructure investment. GlobalFoundries is expanding its Dresden fab under a program called the SPRINT project, with approximately €1.1 billion earmarked for the effort. The goal is to push production capacity beyond 1 million 300mm wafers per year by the end of 2028.

Automotive is a natural fit. Modern vehicles are packed with sensors and compute nodes that need to operate reliably across extreme temperature ranges and long product lifespans. IoT devices and industrial edge hardware face similar demands.

What this means for the physical AI market

Physical AI describes AI inference that happens at the point of action rather than in a data center: the chip in a robotic arm that decides when to grip, the processor in an autonomous vehicle that decides when to brake, the sensor node in a factory that flags an anomaly before it becomes a failure. Latency, power, and reliability are the constraints that matter, not raw throughput.

GlobalFoundries’ $18 billion serviceable addressable market estimate by 2030 reflects the convergence of edge computing growth, automotive electrification, and industrial automation.