Huawei executive Wang Tao gives an interview explaining how the company builds AI compute competitiveness through a 'super node plus cluster' approach. The piece focuses on Huawei's technology strategy and product positioning in AI infrastructure, without disclosing specific investment figures or financial data.
According to 10jqka.com.cn, another A-share listed company has signed a computing power (AI compute) order worth more than 1.9 billion yuan. It is the latest in a string of Chinese listed firms disclosing large compute-related orders amid expanding domestic demand for AI computing capacity.
Sina Finance reports that domestic GPU maker Xiyun's C-series GPU became the first to complete Day-0 adaptation for China Telecom's Xingchen Xing 4.0 large model, described as a new milestone for domestic compute. The item highlights hardware-software co-development between Chinese chips and models, but discloses no orders, shipment volumes or investment figures.
This is an evening-brief style roundup from Xueqiu whose headline focuses on intensifying competition in the compute industry and restructuring moves by China's state-owned telecom enterprises. The excerpt is extremely limited, offering no specific companies, figures or project details, indicating only that compute infrastructure and reform of state-owned telecom players were the day's topics.
Sohu published a report visiting the Karamay Cloud Computing Industrial Park in Xinjiang, describing the development of local computing infrastructure and its new momentum. It is an on-the-ground feature about a local industrial park, disclosing no investment figures, compute scale, or corporate financial data.
Sohu reports that a Chinese province has built 88 large-scale compute centers, with its intelligent compute capacity surpassing Inner Mongolia and Zhejiang to rank first in the country. The piece focuses on regional rankings of compute infrastructure and discloses no specific investment or revenue figures.
Hebei Radio and Television reports that Langfang is accelerating construction of computing-power infrastructure to attract AI and digital-economy companies. The item is part of a local 'high-quality development' research series, emphasizing how a computing foundation drives regional industrial clustering.
Shennan Circuits disclosed that its Wuxi-based AI computing power electronic circuit (PCB) project is under construction, with a one-year build period. The item is a routine capacity-expansion disclosure and contains no specific investment amount, order, or revenue figures.
China Telecom released and open-sourced its new-generation Xing (Star) model Xing4.0-29B-A4B on September 17, with 29B total parameters, only 4B activated, native 256K context extendable to 512K. It is billed as China's first 10B-plus-parameter model trained entirely on a domestic stack — Ascend training chips, domestic frameworks, in-house architecture — and scored 93.52 on SuperCLUE's agent benchmark, third place and less than one point behind two leading Qwen models. After 4-bit quantization its VRAM footprint drops to 15GB, letting consumer GPUs like the RTX 3090/4090 run long-context tasks locally.
Sina Finance reports that the "Pearl Constellation" strategy has been put in place, with computing power accelerating its push into space. The report discloses no investment figures, participating companies, or specific timelines.
On September 19, Qwen officially released its new simultaneous interpretation model, Qwen3.8-LiveTranslate. The model rebuilds real-time interpretation on an Interleave architecture, improving accuracy, fluency and conciseness, while cutting average latency per character (LAAL) from 2.8 seconds to 2.3 seconds.
Alibaba's Qwen released the simultaneous interpretation model Qwen3.8-LiveTranslate on September 19. Built on an Interleave architecture with a Hybrid MoE Thinker–Talker dual-module design, it cuts average latency per character (LAAL) from 2.8 seconds to 2.3 seconds and adds real-time speaker diarization, same-frame bilingual output, and long-context disambiguation. The company says it supports 60 languages and leads the previous generation and mainstream real-time interpretation systems on translation quality, latency, speech recognition and speech synthesis across the Omnilingua-MSpeaker and FLEURS benchmarks.
CITIC Securities issued a research note saying Huawei's Ascend 960 chip will launch earlier than expected, and argues the entire domestic compute supply chain in China will benefit. The view concerns China's homegrown AI chip and compute ecosystem, and carries no specific investment figures or financial data.
Sina Finance reports that two major ETF issuers are in a fierce marketing battle over compute-themed ETFs, offering freebies such as KIMI memberships and Zhipu monthly cards as promotions. It reflects a crowded pipeline of compute/AI-themed fund products and intense distribution competition in China.
According to chnfund.com, China AMC and E Fund have rolled out a marketing gimmick in which investors who chat about compute-themed ETFs receive free tokens. The report focuses on how Chinese mutual funds are marketing AI-compute-themed products rather than on the scale or returns of AI investment itself.
A CITIC Securities research note argues that diamond-based thermal management is poised for its industrialization debut year, driven by three factors: sharply rising AI chip power consumption, traditional metal cooling approaching physical limits, and improving diamond fabrication processes. The report recommends positioning around two themes — companies first to commercial delivery and those with significant production capacity — highlighting firms already shipping heat sinks in small batches with in-house equipment and leading scale, as well as those with large-size manufacturing and full-supply-chain capabilities that are accelerating expansion.
An article from 21st Century Business Herald discusses coordination between compute demand and green power supply, examining how data center electricity needs could combine with renewable energy investment to form a new green investment track. It focuses on the power strain created by compute expansion and the green energy investment opportunities that follow.
Zhu Zhaosheng, head of Huawei's computing strategy and business development, said Ascend has crossed its ecosystem inflection point and can build a new AI ecosystem in the agentic era. The Ascend CANN open-source community averages over 5,200 monthly active users and has been China's most active open-source community since June, with more than 30,000 lines of code merged daily. Over 40 large language and multimodal models have been pretrained on Ascend, and Huawei laid out an annual roadmap through Ascend 980.
Alibaba stock was upgraded to Strong Buy after a 16% drop, trading at a forward P/E of 16.5x. Its AI cloud business grew 45% year-over-year while the company continues heavy AI capital expenditure.
Guangzhou Municipal Press and Publication Bureau released the public notice for 2026 game and esports industry support projects, with 35 projects proposed for support, covering seven directions: technological innovation, original premium game R&D, game operations, game going global, esports events, esports leagues, and esports clubs. Supported projects include NetEase's Messiah game engine, Wuthering Waves, Tale of Immortal, etc., with a maximum support of 5 million yuan per project.