Samsung Electronics will unveil a physical chip of the next-generation High Bandwidth Memory (HBM), seventh-generation HBM (HBM4E), for the first time at Nvidia's developer conference (GTC) 2026 in the United States. By showcasing next-generation HBM technology ahead of rivals at the world's largest artificial intelligence (AI) ecosystem event, the move is seen as an effort to expand its presence in Nvidia's next AI chip supply chain.
Samsung Electronics said it will participate in Nvidia GTC 2026, held Mar. 16–19 (local time) in San Jose, California, and unveil HBM4E chips and core-die wafers. The exhibition also introduced the HBM4 product, which achieved the industry's first mass production and shipment in February.
HBM4E is the successor to HBM4 and is a next-generation High Bandwidth Memory (HBM) with significantly improved performance. Samsung Electronics is developing HBM4E that applies a 1c DRAM process and a 4-nanometer-based base die, supporting up to 16 gigabits per second (Gbps) per pin and up to 4TB/s bandwidth.
In the current AI HBM market, many assess that SK hynix is ahead in Nvidia's supply chain with HBM3E. However, at this GTC, Samsung Electronics preemptively unveiled the next-generation HBM4E technology, flaunting a faster development pace than competitors.
A generational shift in the AI Semiconductor market is also gaining momentum with the arrival of HBM4. Samsung Electronics launched mass production and shipment of HBM4 last month for the first time in the world, mounting a counteroffensive in the next-generation HBM race. HBM4 is a product with significantly improved speed and power efficiency compared with HBM3E, and is cited as a key memory to be installed in next-generation AI GPUs.
SK hynix is also preparing to transition to HBM4, but Samsung Electronics is narrowing the technology gap quickly by pulling forward its mass production timeline. Micron is targeting the AI memory market with HBM3E products, but is seen as somewhat behind Korean corporations in the HBM4 development schedule.
It is also notable that AI Semiconductor competition is expanding from GPU-centered to system-level rivalry that includes memory and packaging. In particular, Nvidia's next-generation AI platform "Vera Rubin" integrates GPU, CPU, memory and storage into a single system, and suppliers adopted for the platform are highly likely to secure stable demand in the next-generation AI data center market. There is analysis that Samsung Electronics, which possesses memory, foundry and packaging capabilities simultaneously, can secure competitiveness amid this shift.
At this event, Samsung Electronics showcased its semiconductor development capabilities that combine memory, foundry and packaging technologies through the "HBM4 Hero Wall" exhibit. It also displayed ▲ HBM4 for Rubin GPUs ▲ SOCAMM2 for Vera CPUs ▲ server SSD PM1763, unveiling memory solutions that make up Nvidia's AI platform.
On the second day of the event, Song Yong-ho, head of the Samsung Electronics AI Center, will take the stage as a presenter at Nvidia's special invitation. Song is expected to introduce Samsung's memory strategy and AI infrastructure technology direction supporting Nvidia's next-generation systems.
A semiconductor industry official said, "SK hynix had been ahead in the existing HBM3E market, but as Samsung Electronics followed its world-first mass production of HBM4 by unveiling HBM4E, the next-generation competition is heating up," adding, "As competition that had centered on GPUs expands into system competition that includes memory and packaging, Samsung's competitiveness is coming to the fore."