2026년 2월 27일 · Unknown · financial · 출처 Yahoo Finance
This article first appeared on GuruFocus.
Broadcom (NASDAQ:AVGO) is stepping further into the AI infrastructure race with the rollout of a new custom chip architecture that it is now shipping to Fujitsu Ltd., a move that could position it more directly against Nvidia Corp. (NASDAQ:NVDA) The company's 3.5D eXtreme Dimension System in Package stacks two chip dies top to top, diverging from an earlier design approach pioneered by Advanced Micro Devices Inc. that layered them bottom to top. According to Harish Bharadwaj, vice president of marketing in Broadcom's custom chip unit, the configuration allows more data to move across the chip with improved power efficiency a feature that may resonate with data center operators managing increasingly demanding AI workloads.
Warning! GuruFocus has detected 2 Warning Sign with AVGO. Is AVGO fairly valued? Test your thesis with our free DCF calculator.
The technology is the result of roughly five years of collaboration with Taiwan Semiconductor Manufacturing Co., during which Broadcom worked through issues tied to power delivery and thermal management. Fujitsu plans to deploy the chip in data centers and possibly supercomputers, Bharadwaj said, while the next phase could involve hyperscalers the largest data center operators adopting the stacked-die architecture. Bharadwaj added that across Broadcom's existing hyperscaler customers, each has selected the technology for its next generation of designs, and that about half a dozen additional custom programs are currently in development beyond Fujitsu.
The broader context matters. Broadcom has emerged as a key rival to Nvidia Corp., which continues to dominate AI accelerators, as customers spanning hardware vendors, AI labs and chip startups look for architectures that can extract more performance per watt. That demand dynamic helped lift Broadcom shares 49% last year. If hyperscaler rollouts materialize as described, the 3.5D packaging approach could become a meaningful lever in the next wave of AI-focused capital spending.
View Comments