SkeenaOneData-driven decisions, made simple

AI briefing ·

US tightens rules on AI chips abroad as Alibaba launches its Qwen4-Max flagship

Short answer

The US Commerce Department issued a rule requiring licences for advanced AI chips in five countries and for overseas data centres serving Chinese companies. Alibaba launched Qwen4-Max, its new top model, OpenAI signed a large computing deal built on AMD chips, and two studies looked at how scientists and patients use AI.

Watch the one-minute video
Policy

US requires licences for AI chips sent to data centres in five countries

The US Commerce Department issued an interim rule on November 16 that requires a licence to export advanced AI chips from Nvidia and other American makers to data centres in Singapore, Malaysia, Thailand, Indonesia and Brazil. It also requires data centre operators anywhere to obtain a licence before giving companies based in China, or other restricted countries, access to large amounts of computing power. Reports say operators already serving such customers will have 90 days to get one. The rule is aimed at a gap that let companies use American AI chips through data centres abroad.

Why it matters: Where AI computing can be built and who may rent it is now set by export rules, which affects cloud prices and data centre investment around the world.

Models

Alibaba launches Qwen4-Max, its new flagship AI model

Alibaba made Qwen4-Max, its new top model, generally available on November 16. Alibaba describes it as a 2.4-trillion-parameter model that can read images and handle very long documents, and prices it at $1.50 per million input tokens. On its own internal test, Alibaba says the average cost of completing a task fell by about 60% compared with its previous flagship; outside reviewers note this has not been independently checked. Unlike earlier Qwen flagships, Qwen4-Max's weights are not being released for download. Alibaba also says it removed offensive hacking material from the training data.

Why it matters: Strong models at lower prices from China add competition, which keeps pushing down what businesses pay for AI.

Chips and hardware

OpenAI signs a six-year deal for 457 megawatts of AMD-based computing

Cloud provider TensorWave announced on November 16 a six-year agreement to supply OpenAI with 457 megawatts of computing capacity built on AMD's Instinct MI455X chips, across several US data centre sites. TensorWave says capacity will start coming online in the first half of 2027, with full deployment expected by early 2028. It describes itself as the first independent cloud provider publicly named as a supplier under OpenAI's 2025 partnership with AMD. The price was not disclosed.

Why it matters: It shows AI labs spreading their computing across more chip makers and smaller cloud providers, beyond the largest cloud companies.

Research

OpenAI report finds scientists say ChatGPT nearly halves the time some work takes

OpenAI's economic research team published a report on November 16 on how scientists use ChatGPT, based on activity from 1,819 scientists in a pilot programme and a survey of 251 of them. It found that 68% used ChatGPT to develop research ideas. Survey respondents estimated that the same work, done to the same quality without AI, would have taken about 1.9 times as long. OpenAI notes these are the scientists' own estimates and that the findings do not prove cause and effect. It also expanded its free ChatGPT programme for academic researchers.

Why it matters: Research teams in universities and companies are starting to measure what AI actually saves them, which helps decide where it is worth paying for.

Research

Apple and OpenEvidence launch a study of how people use AI for health questions

Apple and medical AI company OpenEvidence have opened a study that aims to recruit up to 200,000 iPhone users in the United States. Participants ask health questions in a dedicated app, get an AI answer with links to the published research behind it, and rate the answers for accuracy and usefulness. Sharing health data is optional, and Apple says the study data will not be used to train AI models. The app is designed to suggest following up with a doctor when questions raise concerns.

Why it matters: Before AI health answers reach everyday phones and assistants, companies are testing whether people trust and understand them.

How this feed is made. The AI feed is researched and written each morning by Claude, an AI model made by Anthropic, and published automatically without human editing. Every story links to the reporting it is based on. AI can make mistakes, so check the source before relying on a detail, and treat nothing here as professional advice.