Micron and Samsung compete for Nvidia’s attention in HBM3E memories

South Korea websites reports that both Samsung and Micron are accelerating the development and production of its HBM3E modules, aiming at Nvidia as the main client. But not only that, companies would also be seeking to ensure stocks as soon as possible due to the uncertainties caused by the current US tax policy.

The rise of the Generative AI has generated great demand for the high bandwidth memories of HBM technology. Since Nvidia is the largest AI company of the moment, it also becomes the main customer for component suppliers. There is information that the company wants to diversify its HBM3E supply and not depend so much on SK Hynix.

Samsung seeks to offer 8-HI HBM3E modules (layers), and it is rumored that the company can start its mass production next month, May. The company is still awaiting qualification from Nvidia itself, but recent comments from the company’s CEO Jensen Huang indicates that this should not take long.

Micron, in turn, pursues 12 -layer modules, and we have rumors that the company already has a contract with Nvidia for B300 accelerators, also known as Blackwell Ultra. The Sisa Journal website reports that the company’s production lines for HBM3E have already been running out on orders, and now the company seeks to expand its manufacturing capacity.

Demand for HBM3E should not decrease anytime soon

As mentioned, part of the reason for companies to accelerate HBM3E production is to ensure their profit margin at an unpredictable time in US tariff policy and how they can affect their supply chain.

Industry demand by AI accelerators, however, should not decrease anytime soon. Even with possible future price increases, Samsung and Micron seem safe that they will be able to vent their products.

Related News:

Meanwhile, SK Hynix is ​​not stopped. The company looks close to delivering the first components in HBM4 to try to get ahead of the new generation of memories technology.

Via: WCCFTech