Micron Technology (NASDAQ:MU) is expanding capital spending and deepening long-term customer collaborations as artificial-intelligence workloads increase demand for higher-capacity, higher-bandwidth memory, according to Sumit Sadana, the company’s executive vice president and chief business officer.
Speaking at The Six Five Summit: AI Unleashed 2026, Sadana said memory is in “significant shortage” across market segments and that the company does not yet have visibility into when industry supply will catch up with demand. He said customer demand projections have continued to rise each year.
Memory Moves Beyond Off-the-Shelf Components
Sadana said customers increasingly view memory as a means to differentiate AI systems rather than as a standardized, plug-compatible component. That shift has led to closer collaboration between Micron and customers on multiyear product roadmaps, including custom capabilities that may not be available through conventional JEDEC-compliant products.
He pointed to Micron’s work with NVIDIA on low-power DRAM for data-center use as an example. Micron was the first company to introduce the capability in the data center and was the sole source for a period, Sadana said. He added that more customers are now considering lower-power DRAM because of potential density, size, performance and power-consumption benefits.
According to Sadana, closer memory-processor co-design is becoming necessary because processor and memory development cycles are long. Customers are seeking partnerships that can span five to seven years of research and development, he said, with arrangements resembling ASIC-style engagements more than traditional memory supply relationships.
- Micron is pursuing multiyear strategic customer agreements involving committed supply and committed demand.
- Sadana said the agreements provide more robust returns on investment and can support capital-expenditure decisions over a multiyear period.
- He said customers may work deeply with only one or two suppliers on specialized designs, potentially resulting in periods of sole- or limited-source supply.
AI Demand Extends From Data Centers to Devices and Robotics
While data centers have been the initial focus of AI infrastructure investment, Sadana said intelligence will increasingly move to edge devices, including PCs, smartphones, automobiles, industrial equipment and emerging device categories.
For battery-powered products, memory must enable higher performance within lower power envelopes, he said. Smaller AI models that can run locally on PCs or smartphones could also support privacy and confidentiality by limiting the need to send data to the cloud, according to Sadana.
He described autonomous vehicles, factory automation and robotics as future areas of AI-driven memory demand. In particular, Sadana said humanoid robots could eventually become a major market, initially serving more structured factory tasks before advancing into more complex environments such as homes.
Each humanoid robot could require hundreds of gigabytes of DRAM and terabytes of SSD NAND storage, Sadana said. He characterized robotics as a potential growth driver in the latter part of the decade and into the 2030s, while noting that broader adoption would take time.
Micron Raises Investment Plans as New Capacity Takes Time
Sadana said the growth in AI-driven demand cannot be met solely through technology transitions that increase memory-bit output. The industry needs additional wafers and new cleanroom capacity, he said, requiring construction of greenfield manufacturing sites and supporting infrastructure.
Micron’s capital expenditures were just over $13 billion in fiscal 2025, and the company expects to spend roughly double that amount in fiscal 2026, Sadana said. The company has also indicated fiscal 2027 capital expenditures above $45 billion.
He said Micron raised its planned U.S. investment to $250 billion from $200 billion and accelerated the anticipated timing of that investment. The company is developing projects in the United States, Taiwan, Japan and Singapore, with about 20 investment projects underway globally across front-end and back-end manufacturing capacity.
Among the projects Sadana cited, Micron’s Idaho One fab is expected to have first wafer output in the middle of next year, while Idaho Two is expected to have first wafers near the end of calendar 2028. A Tongluo fab acquired in Taiwan is expected to begin wafer starts in 2027, he said. Micron’s first fab in a planned four-fab cluster in New York is expected to begin output in 2030.
Despite those efforts, Sadana said meaningful new supply is expected to begin ramping only in 2028, with momentum building in subsequent years. Construction, permitting, infrastructure, technician availability and the need to develop skilled local workforces all constrain how quickly capacity can be added, he said.
Manufacturing Complexity Remains High
Sadana emphasized that advanced memory production involves substantial technical complexity. He said a memory wafer can require roughly 2,000 process steps, and the cycle from starting a wafer in a fab to shipping a finished product can take about five months, including assembly, packaging and testing.
He also highlighted the complexity of high-bandwidth memory products, which can stack 12 DRAM dies above a base die connected to a graphics processor, creating thermal, power and packaging challenges. On the NAND side, he said the industry is building devices with more than 200, 300 and 400 layers, while SSDs can reach capacities of as much as 245 terabytes.
“It will take a while for the industry to find a new equilibrium point,” Sadana said of memory supply and demand. “But when exactly that happens, we don’t have line of sight to yet.”
About Micron Technology (NASDAQ:MU)
Micron Technology, Inc is a global semiconductor company that designs and manufactures memory and storage solutions. Its product portfolio includes dynamic random-access memory (DRAM), NAND flash memory, solid-state drives (SSDs), memory modules and embedded memory solutions for a wide range of computing and electronic devices. Micron supplies components used in data centers, enterprise and cloud infrastructure, client computing, mobile devices, automotive systems and industrial applications, and also markets consumer-facing products under the Crucial brand.
Founded in 1978 and headquartered in Boise, Idaho, Micron has grown into an international manufacturer with research, development and production facilities across multiple regions.
