All articles tagged with #hbm4
Micron (MU) fell about 5% as investors weigh durable AI-memory demand against rising capital spending and potential peak-cycle risk; JR Research downgraded to Sell citing margin concerns, while the broader market remains constructive with a Strong Buy consensus and sizable upside driven by hefty capex and strong free cash flow.
Micron has begun high-volume production of HBM4 36GB 12H for NVIDIA Vera Rubin, delivering over 2.8 TB/s bandwidth and more than 20% power efficiency versus HBM3E; it also mass-produces the PCIe Gen6 data center SSD (9650) with up to 28 GB/s sequential read and 5.5M IOPS, and ships 192GB SOCAMM2 memory for Vera Rubin platforms—all showcased at NVIDIA GTC 2026 as part of an AI-optimized memory and storage portfolio.
Micron, a key Nvidia memory supplier, could lose share as Nvidia's Vera Rubin reportedly sources HBM4 from SK Hynix and Samsung, sidelining Micron in the AI data-center memory supply; while demand for Micron's other memory chips may persist and memory prices could rise on persistent shortages, the hit may be temporary as Vera Rubin ramps.
Samsung is set to supply HBM4 memory for NVIDIA’s Vera Rubin AI servers, with shipments expected as early as June and customer deliveries starting in August. Samsung’s HBM4 delivers 11 Gbps+ pin speeds using a 4nm logic base die from its own foundry, giving it a supply advantage over rivals and signaling a strong comeback in high-end memory for Vera Rubin, which will be showcased at GTC 2026.
Samsung plans to start producing its next-generation HBM4 memory chips for Nvidia as early as next month, a move that could ease memory constraints for Nvidia’s AI processors and support its Vera Rubin line. Samsung’s stock rose about 2% on the Reuters report, as the company aims to narrow the gap with SK Hynix in AI memory; HBM4 qualification tests reportedly passed for Nvidia and AMD, with shipments to Nvidia possibly beginning next month.
Samsung Electronics' customers have praised the competitiveness of its next-generation HBM4 chips, with the company aiming to catch up with rivals like SK Hynix in AI chip markets, while also expanding its foundry business through major deals such as with Tesla. Despite positive market performance and record highs, Samsung faces challenges from rising component costs and global trade risks as it seeks to maintain its competitive edge.
High Bandwidth Memory (HBM) is crucial for AI performance, powering the fastest GPUs and accelerators. Industry leaders like Micron, Samsung, and SK hynix are advancing HBM technology with HBM4 and HBM4E, which promise higher capacities and speeds to support next-generation AI models. HBM4 is expected to arrive around 2026, with HBM4E following in 2027, enabling even more powerful AI hardware.