Nam Kim

Nam Kim asked about the long-term growth potential and early market sizing for High Bandwidth Flash (HBF). He also inquired about Sandisk's competitive approach, considering competitors' greater TSV stacking experience from HBM, and what competitive advantages Sandisk believes it brings to HBF.

Answer

Chairman and CEO David Goeckeler deferred on providing specific market potential or TAM numbers for HBF, reinforcing that it is an inference-based solution not targeting model training or replacing HBM. He expressed belief in a large opportunity, especially in the device market due to Flash density. Goeckeler also deferred on detailing Sandisk's competitive advantages, mentioning that the company has focused on design innovations for NAND to achieve higher bandwidth and durability, which they are not yet ready to discuss publicly.

Nam Kim inquired about the long-term growth potential and early market sizing for SanDisk's High-Bandwidth Flash (HBF) technology. He also asked about SanDisk's competitive advantages in HBF, particularly in light of competitors' experience with TSV stacking in HBM.

Answer

Chairman and CEO David Goeckeler deferred on providing specific market potential and TAM numbers for HBF, stating that more details would be shared later. He reiterated that HBF is an inference-based solution, not targeting model training or replacing HBM. He hinted at design innovations in NAND that focus on higher bandwidth and durability, which SanDisk believes will make HBF a compelling technology, but was not ready to disclose these publicly.

Nam Kim asked for Rambus's view on the evolving CXL market, considering CXL 3.1, competitor aggression, and MRDIMM's potential impact on memory capacity constraints. He also sought clarity on Rambus's strategy regarding CXL controllers or other engagements in this space.

Answer

Luc Seraphin (CEO) explained Rambus has two CXL plays: silicon IP (CXL controllers across generations, a growth driver) and product development. He noted the CXL product market is fragmented with bespoke solutions, making it uneconomical for Rambus to roll out a general product, thus preferring the FIP (IP) side. He views memory expansion as the most promising CXL usage model, which MRDIMM addresses by leveraging existing server infrastructure to double capacity and bandwidth. Rambus has chosen the MRDIMM option for now but remains open to pivoting as the market develops.

Nam Kim asked for Rambus's outlook on the CXL market, considering CXL 3.1's expected release, competitor aggression, and MRDIMM's role in addressing memory capacity constraints. He sought clarity on Rambus's strategy regarding CXL controllers or other engagements in this space.

Answer

Luc Seraphin (CEO, Rambus Inc.) explained Rambus's two CXL plays: silicon IP (CXL controllers) which is a growth driver, as many customers require CXL interfaces. However, he noted the CXL market is fragmented with bespoke solutions, making a CXL product (chip) development not economically viable for Rambus at this time. He views memory expansion as the most promising CXL usage model, which MRDIMM effectively addresses by leveraging existing server infrastructure to double capacity and bandwidth. Rambus has prioritized the MRDIMM option but remains open to pivoting if the market evolves differently.