Intel and SoftBank Unveil ZAM Memory to Challenge HBM Dominance

Intel Returns to the Memory Business

Intel and SoftBank subsidiary SAIMEMORY have unveiled a working prototype of Z-Angle Memory at Intel Connection Japan twenty twenty-six, marking Intel's most significant return to the memory business since it exited DRAM production in nineteen eighty-five.

What Is ZAM?

Z-Angle Memory adopts a vertically stacked chip design along the Z-axis, with interconnects routed diagonally through the die stack rather than using traditional vertical through-silicon vias. This architectural innovation enables single-chip capacities of up to five hundred and twelve gigabytes, two to three times the capacity of current high-bandwidth memory solutions, while reducing power consumption by forty to fifty percent.

The design also addresses one of the biggest challenges in stacked memory: thermal management. Heat generated by each chip is conducted uniformly upward, solving the thermal bottleneck that has constrained traditional planar stacking approaches.

A Powerful Consortium

SAIMEMORY, established as a SoftBank subsidiary in December twenty twenty-four, has assembled a formidable consortium of partners including Intel, Fujitsu, Shinko Electric Industries, PowerChip Semiconductor Manufacturing, RIKEN National Research Institute, and the University of Tokyo. The Japanese government is providing partial subsidies for the effort.

The technology builds on Intel's Next Generation DRAM Bonding initiative, developed under the Advanced Memory Technology programme managed by the United States Department of Energy through Sandia National Laboratory, Lawrence Livermore National Laboratory, and Los Alamos National Laboratory.

Market Context and Challenges

The announcement comes amid a global memory shortage driven by surging AI demand. The HBM market is projected to reach over fifty-four billion dollars in twenty twenty-six, with SK Hynix, Samsung, and Micron locked in fierce competition. SAIMEMORY plans to complete prototypes by fiscal year twenty twenty-seven, with commercialisation targeted for fiscal year twenty twenty-nine.

However, Intel's track record in memory raises questions. The company's previous attempt to disrupt the market with Optane technology ended in twenty twenty-two with a five hundred and fifty-nine million dollar write-off. Whether ZAM can succeed where Optane faltered remains to be seen.