Micron has officially started mass manufacturing of its HBM4 memory stack designed for Nvidia’s Vera Rubin GPU. Revealed at the 2026 GTC event, this 36GB 12-high (12-Hi) HBM4 module achieves pin speeds exceeding 11 Gb/s and pushes memory bandwidth beyond 2.8 TB/s-more than double that of Micron’s previous HBM3E generation. Alongside this, Micron also announced the volume production of the industry’s first PCIe 6.0 data center SSD and a new SOCAMM2 module, marking a notable milestone as the sole memory supplier ready with all three components for the Vera Rubin ecosystem at once.
Performance improvements of Micron’s HBM4 memory
Compared to its predecessor-HBM3E at the same capacity and stack height-the HBM4 delivers a 2.3x leap in bandwidth while boosting power efficiency by over 20%, according to Micron’s own power metrics. This improvement addresses long-standing challenges around balancing extreme data throughput with thermal and power limitations in high-performance GPU applications. With bandwidth nearing 3 TB/s on a single memory package, Nvidia’s Vera Rubin platform is set to benefit from substantial memory speed upgrades, critical for demanding workloads such as AI training, advanced simulations, and high-resolution rendering.
Micron’s PCIe 6.0 SSD and SOCAMM2 module for data centers
Micron’s simultaneous rollout of the PCIe 6.0 SSD signals a clear push to cover multiple data-hungry facets of Nvidia’s data center infrastructure, promising faster storage and improved memory performance in tandem. While other memory manufacturers have announced HBM4 designs, Micron’s achievement in scaling to high-volume production first gives them a significant edge in supplying key components for next-gen GPU deployments.
Power efficiency gains supporting Nvidia’s AI workloads
Since HBM4 significantly outperforms previous generations not just in raw bandwidth but also in power efficiency, it helps counteract the escalating power requirements typical of modern GPUs. This efficiency gain aligns with broader data center trends where energy consumption increasingly governs technology adoption. Given Nvidia’s focus on AI and data-intensive compute workloads, Micron’s advancements empower the Vera Rubin platform to achieve higher performance ceilings without proportionally increasing power draw.
Industry outlook for HBM4 memory and Nvidia integration
Following this announcement, industry watchers will keep close tabs on performance benchmarks and how quickly Nvidia integrates this memory into commercial GPUs and servers. The jump from HBM3E to HBM4 also raises questions about potential impacts on competition among memory vendors and how rivals like Samsung and SK Hynix will respond in their HBM4 production roadmaps.