Why the base die matters
Nvidia's next high-bandwidth memory, called NVHBM, moves the memory controller off the GPU and onto the base die. That frees space on the processor for more computing circuits. The physical layer that talks to other chips is about 60 percent smaller than in standard HBM. According to the Seoul Economic Daily, the design delivers up to 30 percent more bandwidth and 15 percent less power than HBM4E.
Nvidia has asked Samsung, SK Hynix and Micron to build base dies on TSMC's 3-nanometer process while using sixth-generation DRAM for the memory core dies. That splits a part Samsung has long built in house.
A yield gap forced the hand
Samsung will make the base dies on TSMC's 3nm line rather than its own. Industry estimates put TSMC's 3nm yield above 90 percent, against roughly 60 percent for Samsung's competing 2nm process. The gap made an in-house base die a risk for a part Nvidia wants in volume.
The arrangement would be the first time Samsung and TSMC have worked together on HBM. The two are direct rivals in advanced foundry work. Samsung plans to keep its own production model for standard HBM and for custom products sold to other customers. It is also reviewing whether its foundry could make NVHBM base dies if demand outruns TSMC's capacity.
Memory supply stays tight
TSMC is adding two 3nm lines at Fab 18 in Tainan, with mass production due in the first half of 2027 and of 2028. It also plans 3nm lines at its second Arizona plant and at Kumamoto in Japan.
Memory remains the bottleneck. UBS projects a 10 percent global shortfall in HBM supply next year. China's CXMT expects third-quarter revenue to rise by as much as 619 percent as it expands its DRAM roadmap. Samsung and SK Hynix are also locking in engineers; job-switching among Korean chip workers has fallen to a record low.
For equipment makers, the shift means more 3nm logic capacity dedicated to memory rather than processors, and more vacuum process steps per stack. Each extra deposition, etch or sputtering step adds demand for rotary seals and feedthroughs that hold vacuum while a shaft turns.
---
Reliable Ferrofluid Vacuum Rotary Feedthrough for Semiconductor Manufacturing
In semiconductor manufacturing, maintaining ultra-high vacuum integrity during rotary motion is critical — from single silicon crystal growth and CVD to sputtering and plasma etching. Ferrofluid Vacuum Rotary Shaft Feedthrough from FerrofluidTec delivers hermetically sealed rotary feedthroughs with leakage rates as low as 10⁻¹² Pa·m³·s⁻¹, vacuum degree up to 10⁻⁶ Pa, and customizable shaft diameters from 10mm to 100mm. Built with non-magnetic stainless steel housings and magnetic permeability stainless steel shafts, these feedthroughs provide reliable, maintenance-free sealing for demanding vacuum environments. Visit www.ferrofluidtec.com to learn more.