![[Sep. 21, 2026] Weekly News](https://www.kfab3d.com/data/file/news/thumb-3530929278_cJbakdSv_d45da3c3955667ca42ece1b389d9c0c97ec36d4e_480x270.jpg)
[Sep. 21, 2026] Weekly News
![[Sep. 17, 2026] onsemi Introduces the Embedded Power Platform, a Breakthrough Architecture for the AI Era](https://www.kfab3d.com/data/file/news/thumb-3698871488_xgyCfTAq_698df516be912052f1513e5ced547f1ea563bf1f_480x270.jpg)
[Sep. 17, 2026] onsemi Introduces the Embedded Power Platform, a Breakthrough Architecture for the AI Era
![[Sep. 14, 2026] Weekly News](https://www.kfab3d.com/data/file/news/thumb-3530929278_7DmUJAvy_a504241849bb8d3bc177f2e7737fdffdc5346ec3_480x270.jpg)
[Sep. 14, 2026] Weekly News
![[Sep. 11, 2026] d-Matrix Adopts NVIDIA NVLink Fusion for Rack-Scale XPU Deployment](https://www.kfab3d.com/data/file/news/thumb-3698871488_IfySUC7h_1d037a8de055b06a537a539d831e419476bb1218_480x270.jpg)
[Sep. 11, 2026] d-Matrix Adopts NVIDIA NVLink Fusion for Rack-Scale XPU Deployment
![[Sep. 10, 2026] A Stealth Startup Thinks It Just Hacked the Memory Shortage : Kepler Computing](https://www.kfab3d.com/data/file/news/thumb-3698871488_8lK2At3u_cfe117292bd68b036fa2cb9111caa9eab69102ad_480x270.jpg)
[Sep. 10, 2026] A Stealth Startup Thinks It Just Hacked the Memory Shortage : Kepler Computing
![[Sep. 10, 2026] 무라타, AI 반도체용 전력기판 'iPaS' 내년 양산](https://www.kfab3d.com/data/file/news/thumb-3698871488_0AzMdTug_37445c0f25ccfe836e49dcb1fed44c2544061f2b_480x270.png)
[Sep. 10, 2026] 무라타, AI 반도체용 전력기판 'iPaS' 내년 양산
![[Sep. 7, 2026] Weekly News](https://www.kfab3d.com/data/file/news/thumb-3530929278_iNQ46Jwf_430505058d7422c376304d3d1133474995e44719_480x270.jpg)
[Sep. 7, 2026] Weekly News
![[Aug. 31, 2026] Weekly News](https://www.kfab3d.com/data/file/news/thumb-3530929278_598mISFk_e58d54044b901c09ce64d085a1080f225bdc45d7_480x270.png)
[Aug. 31, 2026] Weekly News
![[Aug. 24, 2026] Weekly News](https://www.kfab3d.com/data/file/news/thumb-3530929278_TKLn3ANV_eea91a3eb25d141329b4c636ca4feb60930a3f8f_480x270.png)
[Aug. 24, 2026] Weekly News
d-Matrix는 DRAM 메모리 다이와 SRAM 연산 다이를 수직 적층한 ‘2층’ Raptor XPU를 NVLink Fusion을 통해 NVIDIA MGX 랙에 통합합니다.
GPU가 연산 집약적인 prefill을, Raptor가 지연시간에 민감한 decode를 맡는 이종 분리 구조이며, 연말 전 테이프아웃과 2027년 4분기 초기 공급이 목표입니다.
