Price hikes suppress mobile demand, flattening contract prices while spot market stays weak. NAND controllers are evolving into AI compute layers. Driven by CMX architecture, customized NAND will integrate with AI processors across three phases to break memory bottlenecks.

Key Highlights

• Storage Market Dynamics: Secondary price increases by smartphone vendors have intensified consumer resistance and weakened end demand, leading to flat Q4 contract prices. Meanwhile, spot market transactions remain sluggish with ongoing downward price adjustments.
• Evolution of NAND Controllers: Leveraging advanced manufacturing processes, control chips are taking over traditional CPU/GPU tasks—such as CXL communication, data compression/encryption, and expanded cache—transforming SSDs from static storage into active compute participants.
• Architecture Shift & Ecosystem Expansion: Industry leaders are pushing new frameworks like CMX to offload KV cache onto SSDs, significantly boosting inference speed and energy efficiency. Key SSD controller and storage suppliers are launching dedicated products in alignment.
• Three-Phase Roadmap:
  • Phase 1 (Short-term): Offloading cache management to SSDs to alleviate HBM/DRAM capacity constraints and high cost burdens.
  • Phase 2 (Mid-term): Utilizing dedicated controllers and advanced interfaces to achieve low-latency near-memory computing.
  • Phase 3 (Long-term): Overcoming thermal and packaging limits to achieve 3D heterogeneous integration of NAND and AI compute chips.

Table of Contents

1. Market Update
2. TrendForce’s View
   • NAND Controller Customization from Conventional Architecture to NVIDIA’s CMX

<Total Pages: 3>