Micron Unveils New Memory Tech For AI Data Centers, Nvidia Team Up

Author: Anusuya Lahiri | October 22, 2025 11:04am

Micron Technology, Inc (NASDAQ:MU) has begun customer sampling of its 192GB SOCAMM2 (small outline compression attached memory modules), designed to accelerate adoption of low-power memory in AI data centers.

The move comes as rapid artificial intelligence innovation prompts data centers to opt for energy-efficient infrastructure.

In partnership with Nvidia Corporation (NASDAQ:NVDA), Micron pioneered low-power server memory for data centers over the past five years.

Also Read: Micron Rides AI Wave With Upgraded Outlook, Analysts See More Gains Ahead

SOCAMM2 brings LPDDR5X advantages—low power consumption and high bandwidth—to AI system main memory, meeting the demands of massive-context AI platforms. It delivers the data throughput and energy efficiency required for both AI training and inference, positioning SOCAMM2 as a key memory solution for next-generation AI platforms.

Micron transforms low-power dynamic random access memory or DRAM originally designed for mobile devices into data center-class modules through specialized design and rigorous testing.

Customer samples of SOCAMM2 are now shipping in capacities up to 192GB per module with speeds up to 9.6 Gbps, and high-volume production is aligned with customer launch schedules.

Micron stock gained over 140% year-to-date topping Nasdaq 100 index’s over 19% returns.

Wall Street analysts identified Micron’s key growth drivers as demand for AI-related memory in data centers, particularly DRAM and high-bandwidth memory (HBM), as well as favorable supply conditions that support strong pricing. RBC Capital's Matt Bryson noted DRAM and HBM momentum fueled by AI infrastructure spending.

Price Action: MU stock is trading lower by 1.15% to $199.95 at last check on Wednesday.

Read Next:

• Who Benefits From AI Spending? JP Morgan’s Top Picks

Image: Shutterstock

Posted In: MU NVDA