Piecemakers Technology, Inc., a DRAM designer backed by Nanya Technology, is focusing on custom memory solutions for edge AI devices rather than High Bandwidth Memory (HBM). The company believes that inference memory needs will diverge from training memory, and proposes a solution of DRAM stacked directly onto processors using hybrid bonding. This strategy is reflected in their revenue model, which prioritizes design fees over chip sales, with custom AI design work comprising a significant portion of their income. Piecemakers aims to shift towards IP licensing and royalties by 2027, with their first major customer program expected to contribute financially around that time. AI
IMPACT This strategy could lead to more specialized and efficient memory solutions for edge AI devices, potentially reducing reliance on high-cost HBM for inference tasks.
RANK_REASON Company announcement of a new strategy diverging from industry trends (HBM) with a focus on custom edge AI memory solutions. [lever_c_demoted from significant: ic=1 ai=0.7]
- Cristiano Amon
- dynamic random-access memory
- Groq LPU
- High Bandwidth Memory
- Joseph Ting
- Lee Hsiao-wen
- Nanya Technology
- Nvidia
- Piecemakers Technology, Inc.
- Qualcomm
- Taiwan
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
