In the past 24 hours (December 15 to December 16, 2025), domestic AI hotspots have focused on video generation model upgrades, application gameplay sinking, and AI hardware supply rhythm adjustments; Overseas, controversies around the expansion of open source model ecosystems, infrastructure mergers and acquisitions, enterprise-level AI financing and copyright governance continue to heat up.
1. Alibaba released Tongyi Wanxiang 2.6: Cinematic video generation strengthens "role-playing"
Alibaba launched the Wanxiang 2.6 series of video models, adding role-playing, multi-camera narrative, sound drive and audio and picture synchronization capabilities, and simultaneously launched relevant platforms for call and experience. Video generation has moved from "can be generated" to "can control characters and narratives", which has a more direct impact on the industrial production of short dramas, advertisements and content.
2. Qianwen launched "AI Small Theater": one-click corner change co-production short film
Qianwen APP is connected to Wanxiang 2.6 and launched gameplay such as "AI Small Theater", focusing on "character co-production" and rapid generation of multi-shot short films. The closer coupling between model capabilities and productization means that video generation is spreading to socialization, templated, and low thresholds.
3. Doubao AI mobile phone sales resume: control the rhythm with F-code waitlist According
to multiple information, Doubao AI mobile phone-related models have reopened the waiting application channel and resumed purchase qualifications, but the quota and rules are more restrained. The popularity of AI terminals is still there, but channels, risk control and supply management are becoming key variables for whether products can continue to be released.
4. The Ministry of Industry and Information Technology has "conditional access" for L3 models for the first time: autonomous driving has entered the license stage
Accordingto reports, the Ministry of Industry and Information Technology has granted conditional access permits to some L3 conditional autonomous driving models in the latest batch of announcements, releasing a signal of coordinated promotion between supervision and industry. For the industrial chain of "intelligent driving large model + vehicle-end computing power + data closed loop", the compliance boundaries are clearer and more enforceable.
5. Cross-strait entrepreneur summit focuses on AI and industrial Internet: industrial collaboration is increased
Atthe relevant summit forum, experts advocated that the two sides of the strait work together to develop AI and the industrial Internet, emphasizing the promotion of intelligent upgrading on the basis of manufacturing, supply chain and 5G private networks. The industry side pays more attention to "replicable scene landing" rather than a single point of show-off.
6. NVIDIA Releases Nemotron 3 Open Source Model Family: For Multi-Agent and Efficient Inference
NVIDIAreleases Nemotron 3 open source models, data, and related libraries, covering different scales and emphasizing efficient and scalable agent development. The open source high-performance model superposition tool chain will further lower the threshold for enterprises to build industry agents and multi-agent systems.
7. NVIDIA acquires SchedMD: Incorporating Slurm scheduling capabilities into the AI infrastructure landscape
NVIDIA announced the acquisition of SchedMD to strengthen job scheduling and cluster management capabilities around Slurm, and said it will continue to maintain its open source attributes. For large model training and inference clusters, scheduling and resource orchestration are the key layers of "invisible but determining costs".
8. Databricks raised more than $4 billion and valued at $134 billion: enterprise-level AI platforms continue to attract money
Databricks disclosed that the latest round of financing exceeded $4 billion, with a valuation of $134 billion, reflecting that capital is still betting on "data governance + machine learning + enterprise AI landing" platform-based companies. Enterprise AI competition is shifting from model competition to systematic warfare of "data, engineering and governance capabilities".
9. Meta promotes "AI-first" internal office: competing tools and self-developed models are used in parallel
Accordingto the report, Meta internally promotes AI as a daily work infrastructure, introduces a number of AI tools and self-developed models to be used in parallel, and promotes adoption through incentive mechanisms. The trend for large organizations is more like a "multi-model, multi-tool assembly" rather than a single vendor taking all.
10. Escalation of copyright disputes over AI training in the UK: public consultation is more inclined to strong copyright protection
Accordingto the results of relevant consultations in the UK, most of the feedback opposes the "active withdrawal" training exception plan, requiring a clearer licensing and compensation mechanism between AI training and creative rights. The uncertainty of copyright rules is becoming one of the main external constraints on the expansion of generative AI in the content industry.
Frequently Asked Questions (Q&A)
Q: What are the most obvious product trends in the past 24 hours?
A: Video generation is shifting from capability display to "controllable, usable, and disseminated" product forms, with typical performances such as role-playing, multi-camera narratives, and rapid implementation of in-app gameplay.
Q: What does the open source model mean for enterprise implementation?
A: After the improvement of open source high-performance models and tool chains, enterprises can build industry agents at a lower cost, but they still need to invest in data compliance, evaluation, and operation and maintenance.
Q: Why is AI infrastructure M&A a hot topic?
A: The scale of training and inference is mainly subject to cluster management and resource utilization, and scheduling and orchestration capabilities directly determine computing power cost and delivery stability, so it has become a priority for large manufacturers to complete the territory.
Q: How will regulatory and copyright disputes affect the pace of the industry?
A: In the short term, it will increase the compliance cost and uncertainty of content applications, and in the long term, it may promote clearer licensing, identification and sharing mechanisms, promoting the industry from "grayscale expansion" to "sustainable operation".
