Skip to main content
Back to timeline
Preferred NetworksSource publication:

PFN and SEMIFIVE team up on MN-Core L2000 inference chip: double the memory bandwidth, design handoff set for Q1 2027

Synopsis

PFN and SEMIFIVE will collaborate on the MN-Core L2000 series, an AI-inference semiconductor that is the successor to PFN's MN-Core L1000 series, planned for sale in 2027, with more than twice the memory bandwidth of L1000. To handle the structural complexity of stacking logic and DRAM vertically, SEMIFIVE provides design implementation and supply-chain management while manufacturing uses Samsung Foundry's platform, and the two plan to complete design and hand off final data in the first quarter of 2027. The approach bets on memory bandwidth as the bottleneck for generative-AI inference, but power, heat and yield issues in 3D stacking still have to be resolved before mass production.

Source-provided article image: PFNとSEMIFIVE、AI推論向け次世代半導体 MN-Core L2000シリーズ の開発で協業
Preferred Networks

Interpretation

PFN and SEMIFIVE will collaborate on developing the MN-Core L2000 series, an AI-inference semiconductor that is the next generation of the MN-Core L1000 series PFN plans to start selling in 2027. The MN-Core L1000 series is not yet on sale, and L2000 is a follow-on model that expands memory bandwidth to more than twice that of L1000. The two companies are already working together and plan to complete the design and hand final design data to the manufacturer in the first quarter of 2027.

The bandwidth gain comes from stacking logic semiconductors and DRAM vertically rather than only raising compute speed. The MN-Core L series focuses on memory bandwidth, which governs generative-AI inference performance, and uses 3D stacking to widen it, a different route from the GPUs commonly used for AI processing. PFN handles processor architecture and logic design, SEMIFIVE contributes design implementation and supply-chain management with its 3D-IC design solution, and manufacturing uses Samsung Foundry's platform.

Perspective

The report covers the two companies' collaboration announcement and development plan: the L2000 series is scheduled to complete design and hand off manufacturing data in the first quarter of 2027, the L1000 series is planned for sale in 2027, and the two will exhibit jointly at PFN's booth at SC26.

3D-stacked AI semiconductors must address more complex technical issues in power, heat dissipation and yield than before, and the announcement gives no specific values or validation results for these metrics.