Samsung’s 2nm Chip Supply Timeline Uncertain for Preferred Networks

Samsung has announced a collaboration with Preferred Networks to develop AI accelerator components that support generative AI. The partnership will see Samsung providing components made up of 2nm chiplets and advanced packaging. However, specific details such as the timeline for production and the exact 2nm process variant to be used have not been disclosed.

Earlier, Samsung had introduced its SF2 2nm process, slated for mass production in 2025 and 2026 for mobile products, focusing on optimizing power consumption. In contrast, the SF2X process, tailored for AI applications, is expected to enter mass production in 2026, with back-side power distribution enhancements scheduled for 2027.

Meanwhile, industry rival TSMC is reportedly commencing risk production for Apple, indicating a competitive landscape in the advanced semiconductor manufacturing sector. Samsung's introduction of gate-all-around (GAA) nanosheet technology in its 3nm node has positioned the company ahead of TSMC, reinforcing its claim of technological leadership in the 2nm process domain.

The collaboration with Preferred Networks includes the integration of the 2.5D I-Cube S technology, a heterogeneous package technology enabling multiple chips within a single package. The silicon interposer for this technology was developed by Gaonchip, a design services company.

Junichiro Makino, Chief Technology Officer at Preferred Networks, expressed enthusiasm about the collaboration, stating, "We are excited to lead in AI accelerator technology with Samsung Electronics' 2nm GAA process. This solution will significantly support Preferred Networks' ongoing efforts to build highly energy-efficient, high-performance computing hardware to meet the demands of generative AI technologies." Taejoong Song, head of Samsung's foundry business development team, emphasized the significance of the collaboration, highlighting Samsung's 2nm GAA process and advanced packaging technology as ideal solutions for next-generation AI accelerators.