Micron to Ship HBM3E Reminiscence to NVIDIA in Early 2024

Micron has reaffirmed plans to start out shipments of its HBM3E reminiscence in excessive quantity in early 2024, whereas additionally revealing that NVIDIA is considered one of its major prospects for the brand new RAM. In the meantime, the corporate pressured that its new product has been acquired with nice curiosity by the trade at giant, hinting that NVIDIA will doubtless not be the one buyer to finish up utilizing Micron’s HBM3E.

“The introduction of our HBM3E product providing has been met with robust buyer curiosity and enthusiasm,” stated Sanjay Mehrotra, president and chief government of Micron, on the firm’s earnings name.

Introducing HBM3E, which the corporate additionally calls HBM3 Gen2, forward of its rivals Samsung and SK Hynix is a giant deal for Micron, which is an underdog on the HBM market with a ten% market share. The corporate clearly pins a variety of hopes on its HBM3E since this may doubtless allow it to supply a premium product (to drive up its income and margins) forward of its rivals (to win market share).

Sometimes, reminiscence makers have a tendency to not reveal names of their prospects, however this time round Micron emphasised that its HBM3E is part of its buyer’s roadmap, and particularly talked about NVIDIA as its ally. In the meantime, the one HBM3E-supporting product that NVIDIA has introduced to this point is its Grace Hopper GH200 compute platform, which options an H100 compute GPU and a Grace CPU.

“We have now been working carefully with our prospects all through the event course of and have gotten a carefully built-in accomplice of their AI roadmaps,” stated Mehrotra. “Micron HBM3E is at present in qualification for NVIDIA compute merchandise, which can drive HBM3E-powered AI options.”

Micron’s 24 GB HBM3E modules are based mostly on eight stacked 24Gbit reminiscence dies made utilizing the corporate’s 1β (1-beta) fabrication course of. These modules can hit information charges as excessive as 9.2 GT/second, enabling a peak bandwidth of 1.2 TB/s per stack, which is a 44% improve over the quickest HBM3 modules out there. In the meantime, the corporate isn’t going to cease with its 8-Hello 24 Gbit-based HBM3E assemblies. The corporate has introduced plans to launch superior capability 36 GB 12-Hello HBM3E stacks in 2024 after it initiates mass manufacturing of 8-Hello 24GB stacks.

“We anticipate to start the manufacturing ramp of HBM3E in early calendar 2024 and to realize significant revenues in fiscal 2024,” added chief government of Micron.