Intel Eyes Memory Market Return

According to Intel, the company is reassessing its approach to the memory business as artificial intelligence (AI) drives rapid growth in computing and memory demand. Intel CEO Lip-Bu Tan recently said that memory is no longer simply a traditional “commodity” business, with the company exploring new memory architectures and 3D stacking technologies that could more closely integrate CPUs and memory.

In a recent podcast interview, Tan said that AI inference and agentic AI are increasing demand for CPUs. In addition to developing new CPU architectures, Intel is exploring co-design approaches that integrate processors and memory, including potential solutions that stack DRAM directly with CPUs.

Tan noted that he had previously been reluctant to invest in the memory market because it was widely regarded as a highly standardized and cyclical commodity business. However, the rapid expansion of AI workloads and the emergence of new technologies are changing the industry, creating more opportunities for technological innovation.

New memory architectures are reportedly among Tan’s areas of interest. Intel recently appointed former SK hynix CEO Seok-Hee Lee to oversee its foundry and advanced packaging operations, a move that could support the company’s exploration of advanced processor and memory integration. Tan said Intel was not yet ready to disclose specific plans but indicated that the company was evaluating new approaches to memory and computing.

Intel has a long history in the memory industry. The company originally developed products such as the 3101 SRAM and 1103 DRAM before gradually exiting the conventional DRAM market. Intel later worked with Micron to develop Optane memory based on 3D XPoint technology, but the business was discontinued in 2022.

Today, Intel is also working with partners on emerging memory technologies such as XBM and ZAM. These solutions are aimed primarily at addressing the high-bandwidth memory requirements of AI computing and could use new interconnects and memory architectures to help alleviate the “memory wall” that limits performance in increasingly data-intensive AI systems.

As AI servers continue to drive demand for high-bandwidth memory, advanced packaging and high-performance CPUs, Intel is exploring ways to bring processors, memory and advanced packaging closer together. However, Tan emphasized that these memory initiatives remain in the research and planning stages. Intel has not officially announced specific products or mass-production plans for a return to the conventional DRAM market.