In the past 24 hours (November 25 to November 26, 2025), the global artificial intelligence field has continued to play a high-intensity game: on the one hand, China's large models are accelerating to go overseas, and local governments are intensively launching application scenario lists; On the other hand, the United States is betting heavily on AI infrastructure with national plans, and technology giants are competing at the model, security and chip level.
1. China's Tongyi Qianwen has become the base of Singapore's national AI flagship model
Singapore's National Artificial Intelligence Program announced that the latest version of its flagship large model for Southeast Asia no longer uses the American open source model, but uses Alibaba Tongyi Qianwen as the technical base, and tops the Southeast Asian multilingual evaluation list. This is regarded as the first time that China's open source large model has entered the core territory of other countries' national AI strategies, strengthening the international influence of Chinese models in multiple languages and scenarios.
2. Alibaba Quark released a new generation of AI browser, Qwen and Qianwen App are deeply integrated
Alibaba's Quark launched an upgraded version of "AI Browser", which natively embeds Tongyi Qianwen and Qianwen App into the browser interactive interface, focusing on the "anytime" AI assistant experience. At the same time, the number of downloads of Qianwen App has recently exceeded 10 million, showing that the Chinese large model has entered the stage of large-scale competition in C-end applications, which has a squeezing effect on the domestic search, content and tool application pattern.
3. Baidu's architecture shock: Robin Li personally coordinated the research and development of basic and application models
Baidu made major adjustments to the technology research and development system, and established a new basic model research and development department and an application model research and development department, both of which directly reported to CEO Robin Li, and the role of the original CTO was significantly weakened. It is widely interpreted by the outside world that Baidu has "pulled the AI business to the height of the number one project", hoping to increase the pace on the two fronts of general large models and vertical applications to cope with fierce competition at home and abroad.
4. Hangzhou released two "AI application scenario lists", focusing on key infrastructure such as power grids
Hangzhou released two "artificial intelligence application scenario lists" around key areas such as power grid operation and equipment management, guiding enterprises to implement large models and embodied intelligent solutions by opening up specific scenarios such as power inspection and dispatching. The local power grid has built an intelligent inspection system of "air-space-ground integration", which greatly reduces the troubleshooting time, showing that AI is shifting from "conceptual hotspots" to engineering applications at the infrastructure level.
5. Huawei shows off four "black technologies", and Alibaba CEO is wary of AI bubbles
Domestic media reported that Huawei focused on displaying a number of new technologies related to AI computing power and terminal intelligence in the latest event, hoping to form differentiated advantages at both ends of the device and network. At the same time, Alibaba's CEO emphasized in public that there is a bubble in the current capital market's expectations for AI, calling on companies to pay more attention to sustainable business models and real productivity improvements, and pour a basin of "calm water" on the hot track.
6. The United States launched the "Genesis Project" to benchmark the "Manhattan Plan" in the AI era The
President of the United States signed an executive order to officially launch a national AI project called the "Genesis Project", led by the Department of Energy to integrate national laboratories, supercomputing centers and scientific research data to create a unified AI platform, focusing on breakthroughs in energy, quantum technology and national defense. Amazon also announced that it will build dedicated AI infrastructure for the government, with a planned investment of up to tens of billions of dollars, highlighting a new round of arms race between major powers at the level of computing power and basic scientific research.
7. Google released Gemini 3, emphasizing security governance in response to the 2026 AI Summit
Google launched a new generation of large model, Gemini 3, which strengthens security filtering, content tracking and enterprise-level permission management on the basis of multi-modal capabilities, and takes the initiative to "match" the upcoming 2026 Global AI Security Summit. This action is seen as a new round of differentiated competition around "compliance, security, and controllability" by US technology giants in addition to performance, and also warms up for the subsequent transnational regulatory framework.
8. OpenAI expects to have 220 million paying users in 2030, and the new valuation points to the $500 billion level
Some reports quoted internal forecasts that by 2030, at least 220 million people are expected to pay to use ChatGPT-related services; At the same time, OpenAI has re-reached an agreement with major shareholders, and a new round of valuation is said to be approaching $500 billion. Although the specific figures are still uncertain, it is enough to show the capital market's strong expectations for the commercialization space of general AI platforms, and also intensify the competitive pressure of global cloud vendors and large model companies.
9. Nvidia's "AI chip crown" was challenged by Google and Meta's self-developed chips
Overseas technology media analysis pointed out that with Google's large-scale adoption of its own TPUs and Meta's exploration of self-developed acceleration chips, cloud giants are trying to reduce their dependence on NVIDIA GPUs. The article believes that if self-developed chips continue to approach or even replace mainstream GPUs in terms of cost and performance, Nvidia's growth expectations in the AI computing power market may be repriced, which also means that the computing power supply pattern will be more diversified.
10. Samsung cooperates with Nota AI to do model compression and device-side optimization for Exynos 2500
South Korean startup Nota AI signed a cooperation agreement with Samsung Electronics to provide model compression and optimization technology for the latest mobile processor Exynos 2500 to improve the efficiency and power consumption performance of large models running on mobile phones. This shows that "device-side AI" is penetrating from flagship phones to larger-scale terminals, creating more space for developers to design lightweight, privacy-friendly AI applications.
Frequently Asked Questions (Q&A)
Q: What are the two most noteworthy changes in the global AI landscape in the past 24 hours?
A: First, the United States has upgraded AI to a national strategic project similar to the "Manhattan Project" through the "Genesis Plan" and large-scale government computing power investment; Second, Tongyi Qianwen was selected as the flagship model base by the Singapore National AI Project, marking the first time that China's open source large model has deeply entered the national strategic core of other countries, and the two lines jointly promote the new pattern of "China-US bipolar + multipolar catch-up".
Q: Has the international voice of Chinese AI manufacturers been substantially improved?
A: Yes. In addition to the Singapore national model using Tongyi Qianwen, reports show that Qwen is used as the basis for local models in Southeast Asia, Japan, South Korea and the Middle East, and the number of downloads and derivative models is among the top in the world. Coupled with the rapid increase in the volume of Quark AI browser and Qianwen App, Chinese manufacturers are increasing their voice in the three dimensions of "open source ecology + C-end application + international cooperation".
Q: What are the most realistic opportunities for enterprises and developers?
A: Domestically, Hangzhou and other places have structured and opened up specific issues such as power grid, manufacturing, and urban management through the "AI Application Scenario List", which is conducive to industry solutions and embodied intelligent teams; Internationally, AI browsers and device-side large models (such as optimizations on Exynos 2500) provide new entrances and new computing power forms for tools, efficiency, and privacy-sensitive applications. For developers, designing products around "browser entrance + industry scenario + device-side deployment" is one of the most down-to-earth directions in the short term.
Q: What are the main risks and hidden concerns facing the current development of AI?
A: On the one hand, capital and market expectations are highly concentrated in a few leading companies, combined with the high valuation of the US "Genesis Plan" and OpenAI, which is easy to amplify valuation fluctuations and the risk of "AI bubble", which is also a point repeatedly reminded by executives of large domestic companies; On the other hand, as models such as Gemini 3 put security and regulation first, coupled with the increase in abuse cases of generated content in fraud and cyberattacks, the constraints on privacy protection, content compliance and computing power security in various countries are expected to be significantly tightened, bringing compliance costs to product implementation.
