Samsung announced on Tuesday that its sixth-generation HBM4 memory chips, developed for Nvidia’s Vera Rubin platform, can reach speeds of 11.7 gigabits per second, with the potential to climb as high as 13 Gbps.
These performance levels significantly surpass the current industry benchmark of 8 Gbps. In addition, Samsung revealed an enhanced HBM4E version capable of operating at speeds of up to 16 Gbps, further strengthening its position in high-bandwidth memory innovation.
The announcement was made during Nvidia’s GPU Technology Conference, where Samsung presented its latest advancements in AI computing and emphasized its ongoing collaboration with the U.S. chip giant.
This showcase highlights Samsung’s growing role in the emerging HBM4 market, driven by increasing demand for AI infrastructure. The company recently stated that it became the first to mass-produce and deliver HBM4 products. It also plans to distribute HBM4E samples to clients in the second half of the year.
Although Samsung remains the world’s largest memory chip manufacturer, it had previously lagged behind competitors such as SK Hynix and Micron in supplying earlier-generation HBM3 and HBM3E chips to Nvidia.
During the conference, Samsung reaffirmed its focus on delivering a complete AI computing ecosystem. The company noted that it is currently the only semiconductor provider offering a fully integrated AI solution that includes memory, logic chips, foundry services, and advanced packaging.
Samsung stated that it will showcase its full portfolio of technologies and solutions designed to help customers develop next-generation AI systems.
Meanwhile, Nvidia CEO Jensen Huang introduced several new hardware and software innovations at the event, projecting that the company could generate up to $1 trillion in revenue from its Blackwell and Rubin chips by the end of 2027.
