Micron said it has begun shipments of sixth-generation high-bandwidth memory (HBM4). With competition heating up to supply HBM4 for Nvidia's next-generation artificial intelligence (AI) accelerator "Vera Rubin," set to launch in the second half, the company pushed back against projections that Samsung Electronics and SK hynix would split the volume between them.
Mark Murphy, Micron's chief financial officer (CFO), said at a semiconductor conference hosted on the 11th (local time) by U.S. research firm Wolfe Research, "We have already entered mass production of HBM4 and started customer shipments."
Addressing recent "HBM4 exclusion" rumors, CFO Murphy said, "I will correct inaccurate reports," adding, "HBM4 shipment volume in the first quarter of this year is scaling successfully. That is one quarter earlier than the timing mentioned at the earnings announcement in December last year."
He added, "HBM production capacity is expanding smoothly, and as we said a few months ago, 2026 HBM volume is already sold out," and said, "HBM4 Production yield is on plan. It delivers speeds of more than 11 Gb (gigabits) per second, and we have very high confidence in its performance, quality and reliability."
There had been assessments that Micron's HBM4 failed to meet Nvidia's required speed of more than 11 Gbps and had fallen behind Samsung Electronics and SK hynix in the supply race. Earlier, semiconductor analysis firm SemiAnalysis also said, "There are no signs that Nvidia will order Micron's HBM4," and cut Micron's HBM4 share in Nvidia's supply chain to 0%.
SemiAnalysis expected SK hynix and Samsung Electronics to account for 70% and 30%, respectively, in Nvidia's "Vera Rubin" supply chain.