AI Technology Observations | 17 November 2026, 07:00

This observation covers AI technology changes formally released or reliably verified around the period from 7:53 am on 16 November to 7:53 am on 17 November 2026 in Australia/Sydney. It includes three items.

Qwen4-Max moves from a time-limited preview to a production API

Alibaba formally released Qwen4-Max on 16 November under the model identifier qwen4-max. It is callable through QwenCloud's Token Plan and API and is also available in QwenWork and QwenStudio. The Qwen4-Max-Preview opened on 11 November only through the Singapore Token Plan and is scheduled to end on 25 November; the new fact in this release is that a separate production model is now live. Alibaba describes it as a sparse mixture-of-experts model with 2.4 trillion total parameters and 71 billion activated per token. It natively accepts text, image and video input and provides a one-million-token context window, tool calling and structured output. The flagship model's weights have not been released; Alibaba has only announced plans for smaller open-weight Qwen4 models. Performance, cost, safety and long-running agent results are developer reports. Reuters did not independently verify them, and no complete public third-party reproduction is available.

The United States implements licensing controls on remote access to advanced AI compute

The US Department of Commerce's Bureau of Industry and Security issued the interim final rule Addressing Remote Access to Advanced Semiconductors on 16 November, and it took effect immediately. The rule adds section 744.24 to the Export Administration Regulations. A licence is required when an exporter knows that covered advanced-computing items will be used in a data centre anywhere in the world and that one or more linked data centres will provide more than 12,000,000 aggregate TPP of direct or indirect access to a company headquartered, or whose ultimate parent is headquartered, in Macau or Country Group D:5. Foreign data centres must examine direct and indirect customer ownership or control. Compute used exclusively for commercial recommender systems is excepted when reporting conditions are met. The rule also adds destination-based licence requirements for specified advanced-computing items sent to Indonesia, Malaysia, Singapore, Thailand and Brazil. Existing remote services may continue for 90 days but must obtain a licence within that period. The rule is in force as an interim final rule; licence applications are reviewed case by case, and implementation and any later amendments remain incomplete.

TensorWave and OpenAI sign a contract for 457 MW of AMD-powered compute

TensorWave announced on 16 November that it had signed a six-year contract to provide OpenAI with 457 MW of dedicated, single-tenant AI capacity across multiple data centres. The plan uses AMD Helios rack-scale systems incorporating Instinct MI455X GPUs, sixth-generation EPYC processors, ROCm software and Pensando networking. Compared with OpenAI's previously announced relationship with AMD, the new facts are the named cloud infrastructure provider, contract term, power capacity and deployment schedule. Data Center Dynamics independently confirmed the contract structure and reported that operations are expected to begin in the first half of 2027, with full deployment targeted for early 2028. None of the contracted capacity is reported as operational yet. The parties have not disclosed sites, price, rack count, GPU count or delivered capacity, and no operating performance data is available.

Sources

Alibaba Cloud, Alibaba Unveils Qwen4-Max: New AI Teammate for Work, Creativity, and Discovery, 16 November 2026. https://www.alibabacloud.com/en/press-room/alibaba-unveils-qwen4-max

QwenCloud, Model releases, 16 November 2026. https://docs.qwencloud.com/changelog/models

Reuters, Alibaba details new AI safeguards as Chinese developers step up test disclosures, 16 November 2026. https://www.reuters.com/business/retail-consumer/alibaba-unveils-new-flagship-ai-model-with-safety-upgrades-2026-11-16/

US Bureau of Industry and Security, BIS Issues Rule Addressing Remote Access to American Super Intelligence Technology, 16 November 2026. https://www.bis.gov/press-release/bis-issues-rule-addressing-remote-access-american-super-intelligence-technology

US Department of Commerce, Addressing Remote Access to Advanced Semiconductors, 16 November 2026. https://www.bis.gov/media/documents/rin-0694-ak50.pdf

Associated Press, Trump administration moves to restrict China from indirect access to AI technology, 16 November 2026. https://apnews.com/article/nvidia-export-controls-chips-china-remote-access-60f868d3bb430b85a8657b27d421096e

TensorWave, TensorWave to Provide OpenAI with 457 Megawatts of AMD-Powered AI Infrastructure, 16 November 2026. https://tensorwave.com/blog/tensorwave-to-provide-openai-with-457-megawatts-of-amd-powered-ai-infrastructure

Data Center Dynamics, OpenAI signs on for 457MW AI capacity with TensorWave, 16 November 2026. https://www.datacenterdynamics.com/en/news/openai-signs-on-for-457mw-ai-capacity-with-tensorwave/


Discover more from Geoffrey Chen

Subscribe to get the latest posts sent to your email.