Published 6 days ago • loading... • Updated 5 days ago

DIGITIMES ASIA: AI workload surge creates opening for alternative memory and chiplet technologies, expert says

Chuck Sobey, founder of ChannelScience, warned that AI workload growth is forcing a data-center redesign and causing historic memory shortages, per DIGITIMES, ahead of December 5, 2025.
Modern AI workloads now demand memory bandwidth as deep learning training, large language model inference, and retrieval-augmented generation access gigabyte-sized tensors with GPU memory access up to 10 terabytes per second, far exceeding standard DDR5 bandwidth.
Suppliers face strained inventories as high-performance server memory hits a 70% fulfillment rate with price surges of up to 50%, while rack redesigns support up to one megawatt.
Chiplet advocates note the chiplet architecture enables transition from monolithic ASICs and opens entry for magnetoresistive RAM , resistive RAM , and phase change memory , with Sobey urging suppliers to act before pricing and supply chains settle.
The industry is already treating data-tiering shifts as the long-standing five-minute rule accelerates to a five-second rule, and Chuck Sobey will discuss chiplets at the DIGITIMES webinar on December 5, 2025 and the Chiplet Summit in Santa Clara, February 17–19, 2026.

Insights by Ground AI

25 Articles

ABC FOX Montana

+24 Reposted by 24 other sources

Center

DIGITIMES ASIA: AI workload surge creates opening for alternative memory and chiplet technologies, expert says

TAIPEI, Dec. 1, 2025 /PRNewswire/ -- The rapid expansion of artificial intelligence (AI) workloads is forcing a fundamental redesign of data center infrastructure, driving historic memory shortages and opening a rare market window for alternative memory technologies, according to Chuck…

6 days ago·Missoula, United States

Read Full Article

Think freely.Subscribe and get full access to Ground NewsSubscriptions start at $9.99/year