Samsung has been striving to get its HBM3 and HBM3E chips certified by Nvidia since last year but faced obstacles due to heat and power consumption issues not meeting Nvidia's standards. It appears that Samsung has made slight modifications to its HBM3E chips to address these concerns.

Recently, the company's HBM3 chips received Nvidia's certification for the Chinese market. Now, Reuters reports that Samsung's HBM3E (fifth-generation HBM) chips have also received Nvidia's certification, marking a significant milestone.

According to a new report from Reuters, Samsung's 8-layer HBM3E chips have been certified by Nvidia.

This means the South Korean firm's memory chips can be used with Nvidia's AI accelerators. Although Samsung and Nvidia have yet to finalise a deal for these memory chips, it is anticipated that an agreement will be reached soon.

Samsung could commence supplying its chips in the fourth quarter (October-December) of this year. Given Nvidia's status as the world's leading AI chip firm, securing its certification is a crucial achievement for any component supplier.

The report also indicates that Samsung's 12-layer HBM3E chips have not yet received Nvidia's certification. Sources familiar with the matter declined to comment, citing confidentiality.

Samsung's main competitor, SK Hynix, is set to begin supplying 12-layer HBM3E chips around the same time (Q4 2024) that Samsung starts supplying its 8-layer HBM3E chips.

Despite being the world's largest memory chip maker, Samsung is lagging. By the fourth quarter, analysts expect 60 per cent of Samsung's HBM sales to come from HBM3E chips, whereas SK Hynix anticipates that HBM3E chips will account for 83 per cent of its overall HBM sales.

Micron, Samsung, and SK Hynix are already developing HBM4 chips, which are expected to be available for sale sometime in 2026.