Samsung Electronics unveiled a new generation of high-density memory Tuesday designed to ease one of artificial intelligence’s fastest-growing bottlenecks: moving and storing the enormous volumes of data required by increasingly complex AI systems.

The company’s V10 Bonding V-NAND uses more than 400 layers and a wafer-bonding architecture that increases storage density by approximately 58% from the previous generation. Samsung said the design also improves reading, writing and data-transfer performance while using power more efficiently.

The announcement matters because the AI infrastructure race is no longer centered only on graphics processors. Advanced models require large pools of memory and storage that can feed data to accelerators quickly enough to prevent expensive computing capacity from sitting idle.

Samsung manufactures the memory cells and supporting circuitry on separate wafers before bonding them together. That approach allows the company to add capacity without relying entirely on taller and more difficult conventional chip structures, which become harder to manufacture and cool as additional layers are added.

The technology is aimed primarily at high-capacity solid-state drives and storage systems used in AI data centers. Higher density can reduce the physical space and electricity required to store the same amount of data, two increasingly important considerations for operators facing power constraints and rising construction costs.

Samsung also outlined new concepts for placing high-bandwidth memory closer to AI processors. Its proposed zHBM architecture would stack memory vertically above accelerators, shortening the distance data must travel and potentially improving bandwidth, energy efficiency and heat management.

Those designs remain under development, while V10 Bonding V-NAND is a more immediate part of Samsung’s effort to regain momentum in advanced memory. The company has faced intense competition from SK Hynix, Micron and other suppliers that benefited earlier from surging demand for high-bandwidth memory used with Nvidia’s AI chips.

For data-center developers, the wider shift could broaden the AI spending cycle beyond chip designers. Memory manufacturers, storage suppliers, cooling companies and electrical-equipment producers are becoming just as important to capacity growth as the processors receiving most investor attention.

Samsung’s announcement also points to the next constraint confronting AI companies. Building larger models will require not only more computing power, but memory systems capable of delivering data quickly without adding unsustainable energy use, heat and infrastructure costs.

JBizNews Desk | Wall Street

© JBizNews.com All Rights Reserved.
Reproduction or distribution without written permission is prohibited.