Prediction: Micron and Sandisk Stocks Will Both Plummet After July 30

Micron Technology (NASDAQ: MU) and Sandisk (NASDAQ: SNDK) person been premier beneficiaries of the AI representation roar acknowledgment to their enactment successful DRAM, NAND flash, and high-bandwidth representation (HMB). So acold this year, Sandisk and Micron person surged 503% and 210%, respectively -- making them the apical 2 performers successful the Nasdaq-100.

Missed Nvidia successful 2009? This Rare Signal Is Flashing Again. In 2009, a "Double Down" awesome flashed for a little-known chipmaker called Nvidia. For the archetypal clip successful years, that aforesaid "Total Conviction" awesome is flashing for a institution 1/100th the size of Nvidia. Continue »

While it whitethorn beryllium tempting to travel the momentum, I deliberation a harsh sell-off could beryllium successful store aboriginal this month. Read connected to find retired why.

Why representation is becoming the bottleneck of AI workloads

AI systems cannot relation efficiently without high-performance memory. Frontier models shuttle billions of parameters and contextual information betwixt processors astatine blistering speed. Insufficient DRAM bandwidth oregon retention capableness rapidly becomes a limiting origin careless of however almighty the underlying GPU clusters are.

Micron specializes successful DRAM and HBM for grooming and real-time inference. Sandisk focuses connected high-density NAND flash and enterprise-grade solid-state drives (SSDs) -- the retention instauration for immense information sets and exemplary weights that powerfulness AI astatine scale. Both companies person ridden almighty tailwinds arsenic AI hyperscalers contention to grow capableness -- creating a structural shortage that has supported robust pricing powerfulness and net expansion.

Apple is pushing for much ratio connected its devices

At the moment, determination is nary denotation that Apple (NASDAQ: AAPL) volition physique its ain representation fabs. Instead, the institution continues to trust connected outer suppliers for its DRAM and NAND needs. However, caller reports suggest that Apple is exploring a collaboration with a institution called PrismML, which specializes successful representation compression.

The thought present is simple: Apple could usage bundle optimizations to shrink AI models truthful they tally much efficaciously connected iPhones. By doing so, Apple tin usage representation quantization to slash representation outgo requirements connected prime hardware portion deploying the astir demanding features connected higher-end devices wherever constituent costs tin beryllium passed on to consumers successful the signifier of terms hikes.