PFN and SEMIFIVE partner on MN-Core L2000 inference chips
Preferred Networks (PFN) and SEMIFIVE are developing the MN-Core L2000 Series, a new generation of AI inference chips intended to deliver more than twice the memory bandwidth of the MN-Core L1000 Series. Development is underway, with tape-out, the handoff of the completed design to the manufacturer, scheduled in the first quarter of 2027. The predecessor is scheduled for release in 2027.
The design vertically stacks logic chips based on PFN’s MN-Core architecture with DRAM in a 3D structure. PFN says this approach provides substantially higher memory bandwidth than GPUs and other AI processors, supporting faster inference and improved power efficiency for applications such as AI agents. The architecture targets data movement limitations that can constrain generative AI performance.
PFN is responsible for processor architecture and logical design, while SEMIFIVE will handle design implementation and supply chain management, including coordination across memory supply, manufacturing, advanced packaging and testing. Samsung Foundry will manufacture the logic chip using its platform technology and provide design infrastructure for die connectivity and power optimization. The partners face technical challenges involving power consumption, heat dissipation and manufacturing yield. Development is supported by Japan’s Acquisition, Technology & Logistics Agency under its Innovative Science and Technology Initiative for Security program.