TL;DR
Samsung HBM4E was announced at NVIDIA GTC 2026 with 4.0 terabytes-per-second bandwidth and 16 gigabits-per-pin performance.
Samsung is mass-producing sixth-generation HBM4 memory for the NVIDIA Vera Rubin platform, with processing speeds of up to 13 gigabits per second.
The PM1763 SSD uses the PCIe 6.0 interface to reach read speeds of 28,400 megabytes per second and 61.44TB capacity.
New LPDDR6 memory brings 35 gigabits per second bandwidth to your mobile devices to support local AI processing without draining battery life.
Samsung HBM4E debuted at NVIDIA GTC 2026, where the company showcased a new range of hardware designed to handle the heavy processing demands of modern AI.
These components will appear in the next generation of data centers and personal gadgets as the industry moves toward more complex tasks.
This hardware rollout demonstrates how closely the two companies are working together to meet the needs of the NVIDIA Vera Rubin platform.
High bandwidth memory for large models
Samsung is currently in mass production of its sixth-generation HBM4 for use in NVIDIA hardware. This memory reaches processing speeds of 11.7 gigabits per second and can even reach 13 gigabits per second under specific conditions.
The Samsung HBM4E version pushes these limits by delivering 16 gigabits per pin and a total bandwidth of 4.0 terabytes per second.

To manage the thermal output of these 16-layer stacks, Samsung uses hybrid copper bonding which reduces heat resistance by 20 percent compared to older stacking methods.
NVIDIA CEO Jensen Huang visited the display and signed an HBM4 wafer with the words “Amazing HBM4” to acknowledge the hardware.
Storage solutions for data center efficiency
The PM1763 SSD uses the PCIe 6.0 interface to move data through a system at high speeds without creating bottlenecks, achieving sequential read speeds of 28,400 MB/s and write speeds of 21,000 MB/s.
Also Read: Samsung Galaxy S25 Ultra review: Still worth it 10 months later?
These drives are available in storage capacities that reach 61.44TB. Samsung also introduced the SOCAMM2 server memory module which uses low-power DRAM to provide the bandwidth needed for AI infrastructure.
These modules work with the PM1753 SSD inside the NVIDIA BlueField-4 architecture to help your system manage inference tasks efficiently.
Digital twins and factory automation
Samsung is using AI to improve its own manufacturing plants by creating digital twins of its facilities. The company uses NVIDIA Omniverse libraries to build a virtual model of its Pyeongtaek Plant 1 so it can simulate production before any physical changes occur.
This approach supports a plan to operate autonomous plants across the globe by 2030. According to the company, “As the industry’s only semiconductor company offering a total AI solution spanning memory, logic, foundry and advanced packaging, Samsung will exhibit its full suite of products and solutions that enable customers to design and build groundbreaking AI systems.”
Fast performance for your personal devices
AI hardware is also changing the components inside the phones and tablets that you use daily. Samsung showed LPDDR6 memory with bandwidth reaching 35 gigabits per second per pin while using adaptive voltage scaling to keep power consumption low.
The current LPDDR5X chips already hit 25 gigabits per second and reduce power use by 15 percent compared to previous standards.
These parts enable fast local AI processing on your mobile devices so you can run heavy applications without draining your battery quickly. Samsung also provides PM9E3 and PM9E1 NAND storage for the NVIDIA DGX Spark supercomputer to support these local workloads.





Leave a Reply