Nvidia AI Chips to Get Micron’s 12-Tier HBM3E Memory Stacks — Samsung Still Lagging

While Samsung Electronics is struggling to supply Nvidia with advanced HBM3E memory, American Micron Technology’s success in this area allows it to claim the status of Nvidia’s second-largest supplier. Micron will soon begin supplying this customer with 12-tier HBM3E stacks, which it developed by September last year.

Image source: Micron Technology

Naturally, SK hynix is ​​ahead of both competitors in this area, but Nvidia itself is interested in having multiple HBM3E suppliers, as it allows it to increase the reliability of supply chains and gain some leverage for bargaining with suppliers. At the Wolfe Research event, as reported by Business Korea, Micron Technology CFO Mark Murphy spoke about the advantages of the company’s 12-tier HBM3E stacks. Compared to competitors’ 8-tier HBM3E chips, they provide a 50% increase in capacity and a 20% reduction in power consumption.

Let us recall that at this time Samsung only agreed to supply Nvidia with its 8-tier HBM3E stacks, which will be used to manufacture computing accelerators offered on the Chinese market. Samsung will send samples of 12-tier HBM3E memory to Nvidia by the end of this month, but it is not a fact that the corresponding products will satisfy the customer in terms of their characteristics and quality.

As previously reported, SK hynix and Samsung are racing to get HBM4 memory chips to market, aiming to offer them to Nvidia within a year. Micron Technology is set to begin shipping HBM4 next year, so competition for Nvidia’s HBM4 orders will be fierce.