analiza aiqbench
Goldman Sachs analysis suggests that high-bandwidth flash (HBF) technologies could effectively address the AI "memory wall" by alleviating storage bottlenecks during inference workloads. This shift is projected to unlock a market between $40 billion and $150 billion, with significant implications for NAND manufacturers like Sandisk and SK Hynix over DRAM-focused competitors.
Goldman Sachs: High-Bandwidth Flash Could Crack AI 'Memory Wall', With Potential Market Size Up to $150 Billion
Goldman Sachs released a research report on the 12th stating that as demand for AI inference grows rapidly, insufficient storage capacity is becoming an important bottleneck constraining computing power efficiency. New NAND technologies such as high-bandwidth flash (HBF) are expected to alleviate this 'memory wall' problem and significantly reduce AI infrastructure costs. The bank estimates that if GPU adoption rates reach 15% to 50%, the potential market size for HBF could range from $40 billion to $150 billion. Goldman Sachs maintains a 'buy' rating for Sandisk and SK Hynix due to their leading advantages in HBF technology; Samsung Electronics is also expected to benefit. In contrast, Micron's direct benefits are lower as its business focus leans towards DRAM and HBM, so the bank maintains a 'neutral' rating.
Citește articolul complet pe 36KrSubiecte conexe

[AINews] not much happened today
Latent Space · 10 sept. 2026

[AINews] Meta Muse Wonder & Wonder Full: MSL’s Claude 5.5-tier Frontier Model
Latent Space · 13 oct. 2026

GEM Training: How Meta Doubled the Efficiency of Its LLM-Scale Ads Foundation Model
Meta Engineering · 3 aug. 2026

My Assessment of Plan A's Compute Verification Strategy (+ open questions)
LessWrong · 30 iul. 2026