China’s leading domestic DRAM producer has announced a new step in the country’s effort to build a more capable memory-chip supply chain. ChangXin Memory Technologies, known as CXMT, says its fifth-generation DRAM technology platform has entered mass production and is now underpinning two new mobile-memory products. The key word is “says.”
Alibaba’s Qwen team has released a new visual model that tries to remove a common dividing line in image AI: the separation between generating a picture and editing an existing one. Qwen-Image-2.1 combines both functions in a single system, adding native transparent-image work and multi-reference editing to a relatively compact visual generation component.
StepFun is making its latest model announcement around a practical question: how much work can an AI system carry through before it loses the thread? Its new Step 5 Preview is designed for agentic tasks, software engineering, professional knowledge work, and finance, according to the company.
The first reported outcome from the New York meetings between U.S. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng is modest but consequential: a channel for continued discussion about artificial intelligence.
Building an AI agent is now relatively easy. Making one useful inside a real company is the harder part. Huawei Cloud is using that gap between a convincing demo and a dependable workplace tool as the center of its latest agentic-AI pitch.
A reported lawsuit in Zhejiang is putting an unusual form of AI reliance before a local court: a family’s use of a chatbot to choose an auspicious burial date. The case should not be treated as proof that a chatbot caused a later tragedy, or as a ruling that an AI provider is liable.
China’s listed-company earnings picture is improving in the aggregate, but the gains remain sharply uneven. A Nikkei Asia review of roughly 5,400 companies on mainland exchanges, excluding financial firms, found first-half net profit of about 1.9 trillion yuan, or US$283 billion, up 22% year on year.
Huawei has set out a campus-networking update aimed at workplaces where multiple AI agents, employee devices, business applications, and physical spaces must operate across the same network. The company calls the product the Stellar AI Campus Network Solution and says it combines superfast wireless, full-scope security, and network autonomy.
Artificial intelligence could strengthen China’s productive capacity while leaving a longstanding weakness in domestic demand unresolved, according to Huang Yiping, a Peking University professor and expert member of the People’s Bank of China Monetary Policy Committee. The significance of his warning lies in its institutional setting and its limits.
China has pushed back on calls from U.S. technology leaders to slow the advance of frontier AI, while publicly insisting that rapid development and safety governance must proceed together. The issue gained a new pre-summit frame in a September 18 Nikkei Asia report, published days before the expected Trump-Xi meeting.
Artificial intelligence has entered the immediate pre-summit conversation between Washington and Beijing, but only as a reported agenda item, not as an announced agreement. U.S.
SenseTime co-founder and chief scientist Lin Dahua has offered a deliberately conditional forecast: the “emergence moment” for multimodality could arrive within one or two years. KrASIA published its report of the interview on September 17, 2026, under the headline that a multimodal AI breakthrough could come within two years.
Shenzhen-listed RoboTechnik Intelligent Technology is planning a Hong Kong listing that could seek about US$800 million, according to The Economic Times. The report attributes the plan to two people with knowledge of the matter.
XPeng reportedly plans to offer technology to overseas automakers beyond Volkswagen, building on its Volkswagen alliance. TechNode reported the plan on September 18, citing a Reuters report based on two people familiar with the matter. The account describes an effort to find new revenue streams; it does not describe completed licensing transactions.
MiniMax Code is now publicly available as a terminal coding agent whose repository describes it as “An open-source coding agent for your terminal, powered by MiniMax.” That wording on the GitHub repository establishes the project’s own positioning: a command-line agent rather than a standalone model release.
Qwen’s September 18 announcement introduced Qwen3.8-LiveTranslate for real-time simultaneous interpretation. The company says its Interleave architecture connects streaming understanding, text output, and speech generation. Qwen says average lagging, or LAAL, fell from 2.8 seconds to 2.3 seconds. That is a company-reported comparison rather than an independently published benchmark result.
Reuters reports that ChangXin Memory Technologies, or CXMT, is preparing to enter the NAND flash market, citing three people with knowledge of the private plans. Reuters describes CXMT as a DRAM specialist, meaning that the reported initiative would extend beyond the company’s publicly identified memory focus.
Xiaomi is publicly showing reinforcement-learning training for two unreleased models, MiMo-V2.6-Pro and MiMo-V2.6-Flash, rather than limiting the project to a conventional announcement after training has concluded. TechNode reported on September 18 that the company’s MiMo team is operating a public dashboard for the two runs.
Ant International announced an AI-native product stack for global businesses spanning payments, accounts, foreign exchange, treasury, and growth operations. In a September 18 press release distributed through PR Newswire (news provided by Ant International), the company called it its largest-ever product upgrade, with close to 100 key products across four business pillars.
Beijing-based Spirit AI has set a target date for a proposed embodied-AI capability: converting a spoken request into physical actions a robot can reasonably attempt. In a September 18 report, Reuters said Gao Yang, the company’s co-founder and chief scientist, expects a “GPT-3.0 milestone” by mid-2027.
