The story of Samsung’s attempts to arrange supplies of its HBM3E memory for Nvidia’s needs goes back to the first quarter of this year, since a photo from the stand of the Korean manufacturer began circulating in the media, where the autograph of the founder of Nvidia appeared next to the 12-tier HBM3E chip of this brand. However, it will begin receiving HBM3 type chips from Samsung no earlier than August.
This was reported by Reuters, citing its sources, adding that so far Nvidia has allegedly approved HBM3 chips manufactured by Samsung for use only as part of H20 accelerators for the Chinese market, which are not mass products on a global scale. Whether this means that HBM3 memory from Samsung, not to mention HBM3E, will require additional time to be certified for use in other Nvidia products is difficult to understand. It is noted that for the release of H20, the Korean supplier will begin supplying Nvidia with HBM3 chips in August of this year.
At the end of spring, there were rumors that HBM3E memory manufactured by Samsung could not pass certification to Nvidia requirements due to its high power consumption, but the Korean manufacturer later issued a refutation of this information. In early June, the head of Nvidia said that Samsung-branded HBM3E memory requires improvement. As the largest memory chip maker overall, Samsung lags far behind smaller competitors in the HBM segment. As a result, SK hynix is able to claim half of this market, while the American Micron Technology has also certified its HBM3E chips to Nvidia requirements and cooperates with this company. Samsung is one of the laggards in this regard. SK hynix has been supplying Nvidia with its HBM3 chips since June 2022, and in March of this year it began supplying HBM3E for the needs of this American client.