Samsung could adopt SK Hynix’s manufacturing technique for HBM memory amid production challenges

A highly concerning rumor has emerged from South Korea, with troubling statements from a company like Samsung denying everything and non-optimal production rates. The issue revolves around a critical sector such as memory for AI GPUs, and there is no time to waste. Because of this, industry sources have put forth a very plausible argument, claiming that Samsung will use its biggest rival SK Hynix’s manufacturing technology to boost its HBM memory success rate.

The rumor and argument are on the table, the affected party denies everything, and a performance problem that curiously remains unsolved falls behind schedule compared to the industry. Will Samsung really maintain its pride intact, despite market demand for more chips, knowing that they cannot produce them due to various issues? Place your bets on who is telling the truth, but first, the arguments.

Samsung reportedly gave in to SK Hynix’s technique and plans to use it to improve performance

The problem lies in production, not the HBM3e memory technology itself. It is a headache on how to enhance the success rate of this type of memory, as dozens of leading companies are knocking on your door, asking for every single chip you produce, and you are in a time race with failure rates too high compared to your competition, which also hails from your own country.

Five anonymous industry officials from South Korea have exclusively stated that Samsung Electronics plans to use rival SK Hynix’s HBM3e memory chip manufacturing technology. The goal is not only to compete by swallowing their pride but also to prevent Micron, which has accelerated its plans, from arriving first and better.

Currently, Samsung manufactures HBM3e in 12H with NCF (nano-core fiber), allowing the same height as if there were eight layers. This height improvement is Samsung’s main defensive stronghold for its NCF technology, which also improves thermal performance in the stacks, thereby increasing overall performance.

The downside is that it is challenging to produce, lagging 10% to 20% behind SK Hynix in this regard, fueling rumors that Samsung may use SK Hynix’s MR-MUF for HBM3e instead of NCF.

Samsung denies the information while purchase orders for SK Hynix are being discussed

The sources claim that Samsung has recently issued purchase orders for SK Hynix chip manufacturing equipment using its MR-MUF technology, according to three of the five informants. Moreover, they claim that Samsung is very concerned about performance and success rates, similar to what happened with its SF3, where they had to seek help from an American company to increase production. According to rumors, history is repeating itself with SK Hynix.

However, when asked, Samsung released an official statement denying everything:

“The NCF technology is an ‘optimal solution’ for HBM products and will be used in the new HBM3E chips. We are conducting our HBM3E product business as planned. The rumors that Samsung will apply MR-MUF to its HBM production are not true.”

This denial must be taken with a grain of salt, considering Samsung’s history of denying issues before finally admitting and addressing the reality. For example, with the GAA SF3 node’s first generation, they claimed a 70% success rate, while rumors mentioned an unsolvable issue and a maximum 40% production rate.

Eventually, Samsung admitted everything, but only after they had resolved most issues with external assistance. So, has history repeated itself? Will Samsung use SK Hynix’s MR-MUF technology to avoid losing the HBM3e volume opportunity against Micron and its biggest rival? More answers will come soon.