๐จ๐ณcnBeta (Full RSS)โขFreshcollected in 12m
Intel-SoftBank ZAM Gets Japan Gov Funding Boost

๐กIntel's gov-funded next-gen memory poised to upgrade AI storage infra.
โก 30-Second TL;DR
What Changed
Intel partners with SoftBank subsidiary SAIMEMORY on ZAM
Why It Matters
Gov-backed ZAM could deliver high-density storage essential for AI workloads, enhancing data center efficiency.
What To Do Next
Track Intel ZAM specs for upcoming AI accelerator memory benchmarks.
Who should care:Enterprise & Security Teams
๐ง Deep Insight
AI-generated analysis for this event.
๐ Enhanced Key Takeaways
- โขThe ZAM technology utilizes a novel 3D-stacked ferroelectric material architecture designed to bridge the latency gap between traditional DRAM and NAND flash storage.
- โขNEDO's funding is part of Japan's broader 'Post-5G Information and Communication System Infrastructure Enhancement' initiative, aimed at securing domestic supply chains for advanced semiconductor components.
- โขSAIMEMORY, a SoftBank-backed venture, is leveraging proprietary AI-driven material discovery platforms to optimize the ZAM cell structure for higher endurance and lower power consumption compared to current MRAM solutions.
๐ Competitor Analysisโธ Show
| Feature | ZAM (Intel/SAIMEMORY) | MRAM (Everspin) | ReRAM (Various) |
|---|---|---|---|
| Latency | Near-DRAM | Near-DRAM | Medium |
| Endurance | Ultra-High | Very High | Moderate |
| Scalability | High (3D Stacked) | Moderate | High |
| Primary Use | AI/HPC Cache | Industrial/Aerospace | Embedded/IoT |
๐ ๏ธ Technical Deep Dive
- โขArchitecture: Employs a vertical 3D-stacking process that integrates ferroelectric capacitors directly onto logic wafers.
- โขMaterial Science: Utilizes a proprietary hafnium-based ferroelectric thin film to achieve non-volatile state switching at sub-nanosecond speeds.
- โขInterface: Designed to be compatible with CXL (Compute Express Link) 3.0 protocols to facilitate high-bandwidth, low-latency memory expansion in data center environments.
- โขPower Profile: Features a 'zero-leakage' standby mode, significantly reducing the idle power consumption typical of volatile DRAM.
๐ฎ Future ImplicationsAI analysis grounded in cited sources
ZAM will replace DRAM in AI training clusters by 2028.
The technology's ability to provide near-DRAM speeds with non-volatility addresses the memory wall bottleneck currently limiting large-scale model training efficiency.
Intel will integrate ZAM into its Xeon processor roadmap.
Intel's partnership suggests a strategic intent to create a proprietary memory-processor ecosystem that leverages ZAM's low-latency characteristics to outperform standard DDR5/6 configurations.
โณ Timeline
2024-09
SoftBank establishes SAIMEMORY as a dedicated venture for next-gen memory research.
2025-03
Intel and SAIMEMORY announce a strategic R&D partnership for Z-Angle Memory development.
2026-02
SAIMEMORY successfully demonstrates a 128-layer ZAM prototype in a laboratory setting.
๐ฐ
Weekly AI Recap
Read this week's curated digest of top AI events โ
๐Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: cnBeta (Full RSS) โ


