2025-05-16 23:20

블록미디어
![[All That China] "China's Import of Nvidia AI Chips to Decline from 63% to 42% This Year"](/_next/image?url=https%3A%2F%2Fwww.blockmedia.co.kr%2Fwp-content%2Fuploads%2F2024%2F05%2F%25EC%2597%2594%25EB%25B9%2584%25EB%2594%2594%25EC%2595%2584-%25EB%25B8%2594%25EB%25A3%25B8%25EB%25B2%2584%25EA%25B7%25B8%25EB%25A1%259C%25EA%25B3%25A0.jpg%3Fformat%3Dwebp%26width%3D600&w=1200&q=70)
출처: Block Media
# China's AI Server Market Set to Reduce Dependence on NVIDIA and AMD by 2025
China's AI server market is expected to largely diminish its dependence on foreign semiconductor suppliers, such as NVIDIA and AMD, by 2025, according to an October 16 report by Economic Daily, Licaiman, and United News Network.
# AI Chip Imports Decline Amid U.S. Export Controls
The latest report from Taiwan-based market research firm TrendForce indicates that U.S. export restrictions have led to a substantial drop in China's import share of AI semiconductors. From 63% in 2022, this figure is expected to fall to 42% this year.
TrendForce also noted that domestic Chinese semiconductor suppliers, bolstered by strong policy support, are increasingly capturing a larger market share. By the end of this year, their market presence is predicted to reach 40%, nearly equalling that of international competitors.
# Domestic AI Chip Development Gains Momentum
The report highlighted the rising importance of domestic AI application-specific integrated circuit (ASIC) development in the face of changing global trade dynamics and a revamped supply chain. Leading Chinese firms such as Huawei Technologies and Cambricon Technologies are central to this transformation.
Huawei has made significant strides with its Ascend (Shengteng) chips, designed for various applications including large language model (LLM) training, smart city infrastructure in smaller regions, and AI-driven cloud networks for major telecom operators. Initially focused on the local market, Huawei is now poised to challenge NVIDIA's dominance in China's AI sector due to growing LLM-based AI applications and government-supported national initiatives.
# Cambricon Targets Cloud Applications
Cambricon Technologies' Siyuan (MLU) AI chips are also gaining momentum, particularly for AI training and inference within cloud service providers (CSPs). The report mentions that these chips underwent initial testing with major Chinese CSPs in 2024 as part of a broader market expansion strategy. Cambricon has been scaling its Siyuan AI solutions in the cloud AI sector since the beginning of this year.
# Chinese Tech Giants Intensify AI Semiconductor Development Efforts
The report also highlighted significant advancements by China's leading CSPs in developing proprietary AI ASIC chips. Alibaba’s T-Head subsidiary has released its Hanguang 800 AI inference chip, while Baidu has moved from mass-producing Kunlun II chips to developing Kunlun III, designed for high-performance training and inference.
Tencent, the country's largest internet services provider, is advancing its in-house AI inference chip, Zixiao, while also integrating AI solutions from Enflame Technology, an IC design company in which Tencent has strategically invested.
With significant progress in domestic AI chip innovation and production supported by favorable policy initiatives and expanding AI applications, global giants like NVIDIA may face mounting competition in the Chinese market.
View original content to download multimedia: https://www.blockmedia.co.kr/archives/909380