Realtime AI News
Alibaba shares slide 7.3% as Nvidia controversy revives China chip-access fears
Alibaba Group shares fell as much as 7.3% on Thursday, weighing on the Hang Seng Tech Index, while Chinese AI peers Z.ai and MiniMax each dropped at least 8% as investors feared tighter US scrutiny of how Chinese firms obtain advanced Nvidia chips. The trigger was the widening controversy around Nvidia chief executive Jensen Huang, which investors worry could revive legislation letting Washington police remote access to restricted computing.

Alibaba Group Holding shares fell as much as 7.3% on Thursday, dragging on Hong Kong's Hang Seng Tech Index, as investors weighed the possibility that Washington will tighten its scrutiny of how Chinese companies obtain advanced AI chips. The sell-off was not confined to Alibaba: Chinese model developers Z.ai and MiniMax Group each dropped at least 8%, according to Quartz.
The trigger, according to Bloomberg, is the widening uproar around Nvidia chief executive Jensen Huang. Alibaba rents Nvidia chips through Singapore-based Megaspeed International, and investors fear Washington could clamp down harder on how cloud providers obtain high-end chips as that controversy unfolds.
The affair centres on Alice Huang, who is not related to Jensen Huang. CNBC reported that US intelligence officials suspect she is working as an asset for Chinese intelligence, an allegation China has denied, according to The Japan Times; Megaspeed has rejected claims that it unlawfully rerouted chips to China. Jensen Huang acknowledged having cheated on his wife after reports tied him to Alice Huang.
Corporate records add a financial thread. The Japan Times reported that Alice Huang held investments in, and did business with, top Chinese tech firms including Alibaba and Tencent, and Singapore filings indicate that Apex Enterprise Solutions, an Alibaba unit, put money into Speedmatrix, a Malaysian business with ties to Megaspeed.
The Information has alleged that Alice Huang assembled a web of companies spanning China and Singapore that moved roughly 240,000 Nvidia chips into China, part of the hardware supporting Alibaba, Tencent and several smaller AI labs. Reuters said it could not independently verify those claims.
What worries investors is the policy tail. According to The Japan Times, Chinese AI firms are already seeking to lease Nvidia-powered computing capacity in data centres located abroad, with one large cloud provider internally debating whether to secure such capacity in advance. The companies are concerned the affair may give fresh impetus to the Remote Access Security Act, which would empower the US Commerce Department to oversee how restricted technology is accessed remotely through the cloud; the House passed it 369 to 22 in January and the Senate has yet to act.
Analysts see a delay rather than a cliff. Julia Pan of UOB Kay Hian told Bloomberg that Chinese cloud providers have been building chip inventories since January, but that research and development timelines will inevitably be postponed. Beijing on Thursday urged Washington not to use the spying allegations as grounds for technology curbs, and said the two governments remain in touch ahead of a leaders' meeting.
Two near-term dates matter. Alibaba and Tencent report September-quarter results next week, the first earnings calls since the controversy escalated. Nvidia's board has opened an independent investigation, and House committee leaders have asked Jensen Huang to appear at a Nov. 18 briefing, according to CNBC. For Chinese AI companies, the practical question is less whether the chips exist than whether they can still be reached from a data centre that Washington considers remote.
Why it matters
The sell-off prices geopolitical risk directly into Chinese AI valuations: the near-term hit is delayed research timelines, while the medium-term question is whether the Remote Access Security Act clears the Senate and whether offshore leased capacity falls under it.
Nearby Updates
All11/12, 20:58
NFON launches AI Agent Studio for building agents from plain-language descriptions
NFON AG has launched AI Agent Studio, which turns a natural-language description of a business task into a working AI agent that can be tested and refined. The product combines a conversational Agent Mode with a visual Flow-Based Mode and is available now through NFON's European partner network.
11/12, 21:49
OpenAI's Control Room can cut off a rogue research agent in 35 seconds
OpenAI has built and deployed Control Room, a system that halts models working on research tasks when a safety threat appears, blocking further output within about 35 seconds of a stop command. The mechanism does not shut down the model or its servers, cutting output only for the job that was stopped, and it is designed to fail closed if the service itself breaks.
11/12, 22:02
Insight Enterprises Launches Global AI Agent Store Across 14 Countries
Insight Enterprises has launched a global AI Agent Store covering 14 countries, packaging prebuilt AI agents as catalog products rather than one-off projects. The move tests whether enterprise agents can be sold through a repeatable channel, though details on catalog size, pricing and suppliers were not disclosed.
11/12, 22:06
American Heritage Credit Union's SoundHound AI Agent Passes Half a Million Member Conversations
American Heritage Credit Union's AI agent, deployed with SoundHound AI, has handled more than half a million member conversations in under a year. The milestone shows a credit union-scale institution running a conversational agent as a mainstream service channel, though the report gives no detail on resolution rates or human escalation.