Intel partners with Saimemory to commercialize Z‑Angle DRAM for AI infrastructure

Intel-Saimemory pact aims to unlock new DRAM for AI infrastructure

Z-Angle Memory collaboration targets AI-driven DRAM bottlenecks

Intel is partnering with Saimemory, a SoftBank subsidiary, to push a next-generation DRAM architecture known as Z-Angle Memory (ZAM) toward commercial use for AI and high-performance computing. The companies plan prototypes by the fiscal year ending March 31, 2028 and aim to scale production in fiscal 2029, leveraging Intel technical IP and manufacturing know‑how developed under the U.S. Department of Energy’s Advanced Memory Technology program. Intel says the new assembly approach improves DRAM performance while reducing power consumption and costs, responding to memory designs that Intel Fellows and others describe as inadequate for modern AI workloads.

The collaboration emphasizes energy efficiency to address mounting concerns about the large power draw of AI training and inference clusters. Saimemory, created in December 2024, is expected to coordinate joint testing facilities and iterate prototypes rapidly with Intel, while partners such as Fujitsu perform large‑scale validation across data‑centre workloads, AI models and server platforms. If validation meets expectations, ZAM could alter DRAM roadmaps by delivering higher effective bandwidth and lower total cost of ownership for operators of AI accelerators and servers, and by easing supply constraints that have emerged as AI-driven memory demand outpaces capacity.

Industry observers say success would change the competitive dynamics of the memory market as hyperscalers and chipmakers seek both performance and energy savings. Intel’s participation brings manufacturing experience that could accelerate yield ramp and integration with server ecosystems, while the multi‑party nature of the project aims to ensure compatibility with existing infrastructure. The companies present ZAM as a potential medium‑term remedy to shortages and a strategic complement to chipmakers’ efforts to co‑design memory and compute for AI applications.

Intel names GPU chief architect amid broader AI push

Separately, Intel signals a deeper push into the AI compute stack by appointing a new chief architect to lead GPU development, CEO Lip-Bu Tan says at a Cisco AI Summit. The hire — unnamed — reflects the company’s effort to attract scarce GPU engineering talent as it competes with specialized rivals in a market driven by large language models and data‑centre demand.

Market and supply context

The ZAM announcement comes as the industry copes with memory shortages and as Intel works through production snags that have affected recent quarters. Government and private investments in chips and infrastructure are intensifying, and Intel’s moves on memory and GPUs position it to compete across both hardware and integrated AI platforms.