普通外文研报
J.P. Morgan Asia Technology Tracker 10 Jun 26 China Home Appliance; Delta Electronics, Inc. and More
研报英文原文证据摘录
J.P. Morgan Asia Technology Tracker 10 Jun 26 China Home Appliance; Delta Electronics, Inc. and More
Asia Pacific Equity Research AC Asia Pacific Equity Research
10 June 2026 J P M O R G A N
aprr.publications@jpmorgan.com
Europe Reports/Notes
Tech Snippets
Nvidia building chips tailored for radio units in mobile networks is not a negative for the equipment
suppliers (Sandeep Deshpande)
The News: Nvidia is reportedly extending its AI-RAN strategy beyond central unit/distributed unit (CU/DU)
compute to the radio unit (RU), developing a GPU-based chip for the 6G to replace the custom ASICs that
handle low-PHY functions like beamforming in Massive MIMO radios. The article argues that the technical
driver is Massive MIMO: as radios scale from basic 4T/4R toward 128T/128R (implying ~32x more low-PHY
compute) and potentially “ultra-MIMO” concepts up to 1,024T/1,024R, more PHY processing shifts to the
RU to preserve performance (e.g., beamforming). Nvidia believes this creates an opening to replace RU
ASICs with a GPU-based approach as it makes programmable silicon increasingly attractive versus fixed-
function ASICs that lack scale economics in a non-growth equipment market. Key risks discussed remain
power consumption (RUs drive ~90% of network energy use) and performance gap versus purpose-built
ASICs. Nvidia could likely counter the power dissipation issue with an automotive-like sub-100W embedded
GPU design. The piece also highlights Nvidia’s recent investments in Marvell ($2bn) and Nokia ($1bn) and
notes Marvell’s intent to enhance its Octeon base-station processors to work directly with Nvidia GPUs,
aiming to run radio workloads and AI applications on the same platform.
Key Rating, Price Target & Estimate Changes for Asia tech
companies
None
本摘录由系统从所标注的 PDF 证据页直接提取并保留英文原文,不做批量翻译;登录后在阅读器切换中文时才按需翻译。
打开研报阅读器