Neo Semiconductor develops 3D memory products that combine NAND‑like structural simplicity with DRAM‑class performance, delivering dramatically higher density, bandwidth, and lower power consumption for AI and high‑performance computing.
Funding
Funding not disclosed
Founders
Product
Problem
Current memory technologies for AI accelerators and high‑performance computing face a trade‑off between density, bandwidth, and power consumption. Conventional DRAM and HBM cannot scale cost‑effectively to meet the bandwidth demands of modern AI workloads, leading to memory bottlenecks and excessive system power.
Solution
Neo Semiconductor creates 3D memory products that merge NAND‑like structural simplicity with DRAM‑class performance. Their X‑DRAM line uses a 3D NAND‑based process to achieve eight‑fold higher density and the lowest power consumption among DRAM solutions. X‑HBM provides up to 16× the bandwidth and 10× the density of traditional HBM, directly addressing AI and HPC memory bottlenecks. X‑AI integrates AI inference logic into the 3D memory stack, delivering up to 100× acceleration of AI chip performance while cutting power use by 99%. Together, these technologies enable chipmakers and system builders to build higher‑density, lower‑power AI platforms without waiting for next‑generation memory nodes.
Target Audience
Primary customers are AI chip designers, server OEMs, and high‑performance computing system integrators that require high‑density, high‑bandwidth, low‑power memory solutions for next‑generation AI workloads.
Features
- 3D X‑DRAM™ with a NAND‑like 1T1C cell architecture and IGZO channel for high density (≈8× DRAM) and ultra‑low power operation
- X‑HBM™ offering up to 16× bandwidth and 10× density compared with conventional HBM, optimized for AI and HPC workloads
- X‑AI™ embedding AI processing directly in the memory stack, achieving ~100× inference acceleration and ~99% power reduction
- X‑NAND™ architecture delivering 3× faster random read/write and 20× higher sequential throughput for AI‑centric storage
- Manufacturing‑friendly 3D NAND‑based process that reduces cost and simplifies scaling of memory capacity