GlobalFoundries targets 2028 Dresden manufacturing for FDX Fusion
GlobalFoundries has outlined FDX Fusion, a chipmaking platform aimed at AI systems that sense and respond to their surroundings. Its October 9 roadmap targets Dresden manufacturing in 2028, with demonstration chips in early 2027 and an initial design kit in the middle of that year.
The performance goals include seven nanometer class digital capability and more than double the density of the original FDX generation.
These remain development targets. The release provides no independent workload benchmarks or production yield data, and its performance label does not establish physical transistor dimensions.
An existing expansion reaches a new stage
GF also marked the topping out of SPRINT, its existing €1.1 billion Dresden expansion. The company says this construction phase finished on schedule and budget, allowing the remaining work to continue.
The project aims to lift annual site capacity above one million 300mm wafers by the end of 2028. GF currently lists capacity of up to 950,000 wafers, depending on product mix. Those are totals for Dresden across technologies, rather than a dedicated Fusion allocation.
A separate condition applies to the technology program. GF says the actual start of FDX Fusion development depends on German federal approval for its Dresden project under the planned IPCEI Advanced Semiconductor Technologies program. The construction milestone does not clear that approval requirement.
Why integration matters at the edge
FDX uses fully depleted silicon on insulator technology. GF’s current platform combines low leakage operation with adaptive body biasing, which lets designers adjust the balance between speed and energy use. Its feature set also spans radio frequency circuitry and embedded memory.
For a machine that must interpret sensor input and act within a power budget, the practical question is how those functions work together. A useful evaluation would measure the complete workload, including communication and control, rather than judge the design solely by a headline digital performance class.
The next useful evidence
A process design kit supplies the models, rules and libraries that connect chip design software to a manufacturing process. That makes its delivery an important engineering checkpoint. Teams still need to check whether the available components, simulation models and verification tools cover the product they intend to build.
For product planners, the sensible next step is to map those design requirements before committing a launch calendar. A factory schedule and an initial design kit leave practical questions about validation, cost and production readiness to resolve.
For the broader manufacturing picture, ByteForward’s TSMC September sales report examines what monthly revenue can reveal about demand for AI chips.
Photograph from August 2022 by Albrecht Voß / Stadtarchiv Dresden, via Wikimedia Commons. CC BY-SA 4.0. Original unchanged. Display crops remain under the same license.
