← Feed Deep Dive Matrix Subscribe

Piecemakers bets edge AI devices will diverge from reliance on HBM

tomshardware.com 2026-09-16 Shane Downing
Entities
Tags
DRAM DesignAI InferenceHBMHybrid BondingEdge AINVIDIATSMCSemiconductor Supply ChainCustom MemoryChip DesignDRAM StackingSRAMGroqNanyaFormosa Advanced TechnologiesWafer-on-WaferTSV3D StackingMemory BandwidthAI AcceleratorChip PackagingSemiconductor InvestmentTaiwan SemiconductorChip ManufacturingMemory TechnologyAI ChipChip ArchitectureSemiconductor Market
News Summary
PieceMakers, a DRAM designer backed by Nanya, has entered Taiwan's Emerging Stock Board with a strong market reception, seeing its stock price rise 23.6% on its debut. Unlike traditional HBM vendors, ... Read original →
Industry Analysis
PieceMakers’ emergence signals a paradigm shift in AI memory architecture, moving beyond traditional HBM toward more efficient, low-power stacking solutions. Their HBLL and HiBaLL technologies, leveraging hybrid bonding to directly stack DRAM on processors, outperform HBM in bandwidth and latency—particularly for edge AI inference. This innovation pressures major players like NVIDIA and SK Hynix to accelerate 3D stacking and high-bandwidth memory development. TSMC and other foundries face increased demands for EUV and TSV capabilities. Nanya’s strategic move through PieceMakers allows it to avoid direct HBM competition while targeting edge-AI growth. However, yield control and packaging complexity remain key hurdles. Within the next 12 months, if production scalability is achieved, this approach could redefine the AI memory landscape, accelerating the convergence of edge and datacenter computing.
Read Original Article →
Related
This page displays AI-generated summaries and metadata for research purposes. Original content belongs to the respective publishers.