Micron builds a long-horizon memory R&D hub for the AI era

November 19, 2026

Micron builds a long-horizon memory R&D hub for the AI era

Image: ChatGPT

Micron is establishing Micron Research Labs as a long-term research center in Boise, Idaho, backed by a planned US$10 billion (approximately €9.1 billion) investment over the next decade. The facility is intended to go beyond the company’s usual five- to seven-year technology horizon and to concentrate on critical memory technologies, advanced memory and compute architectures, packaging, and future semiconductor manufacturing.

The investment also includes university collaborations, satellite labs, and broader ecosystem partnerships. Micron expects to begin construction in 2027 on a site designed to host hundreds of researchers. Deirdre Hanford, formerly of Synopsys, has been named corporate vice president and president of the new lab, with responsibility for the research infrastructure, operating model, and partnerships.

The lab’s first publicized outreach effort was a planned investment of up to US$50 million (approximately €45.5 million) over five years for the U.S. Department of Energy’s Genesis Mission Graduate Fellowship Pilot Program and related national laboratory research. Micron says the broader research push adds to previously announced or forecast investments that have risen sharply over the past four years.

The company’s leadership sees the shift as a response to how memory development has changed. Near-term work on DRAM1. dynamic random access memoryMemory technology or system for temporarily or permanently storing digital information.2. dynamic random-access memoryDRAM is a volatile memory technology scaling and yield improvement remains important, but the AIartificial intelligenceAI stands for artificial intelligence era increases the need for longer-horizon research. Memory performance, memory bandwidth, and power have become central constraints in data centers as well as at the edge, in vehicles, and in robotics.

That diversity of applications also raises the burden on research. For data-center use, LPDDR capacity and performance targets have changed significantly, while robotics and automotive systems add advanced packaging and 3D implementations. The resulting projects increasingly require direct work with key customers rather than a simple transition to the next scaled DRAM1. dynamic random access memoryMemory technology or system for temporarily or permanently storing digital information.2. dynamic random-access memoryDRAM is a volatile memory technology node.

Industry observers describe the cost structure as a major reason for deeper partnerships. Research spending tends to rise by about 30 % with each node transition, while a competitive fab adds a second major cost layer. A typical DRAM1. dynamic random access memoryMemory technology or system for temporarily or permanently storing digital information.2. dynamic random-access memoryDRAM is a volatile memory technology manufacturer allocates roughly 20 % of revenue to capital spending and about 10 % to R&D. Shared research through organizations such as imec, CEA-Leti, and university or national-lab programs helps spread both cost and risk.

Emerging memories face the same pressure in a different form. Weebit Nano has used partnerships with CEA-Leti, universities, and research institutes to advance ReRAM, including work on neuromorphic computing and in-memory compute. Its Weebit Matrix In-Memory Compute Research Program gives industry and academic researchers a ReRAM-based hardware development kit for algorithms, system prototyping, programming methods, and architecture exploration.

Micron’s own history also shows how uncertain memory R&D can be. The company spent billions on 3D XPoint before ending the program after deciding it had learned what it could from the effort. For Micron, a new memory technology must clear three hurdles at once: technical viability, manufacturability, and customer acceptance. Micron Research Labs is meant to strengthen that long-term position as memory and logic become more tightly integrated in AIartificial intelligenceAI stands for artificial intelligence systems.

Source: EE Times

More background: Fundamentals

← All posts