24-hour AI news: Baidu disclosed AI revenue for the first time, and Amazon bet $15 billion on AI infrastructure

In the past 24 hours, China has mainly focused on Baidu's AI business revenue disclosure, Beijing's "2025 Artificial Intelligence + Conference" and trusted AI security practices; On the North American and global enterprise side, the AI industry presents a pattern of "revenue realization + security and credibility + infrastructure expansion" in the release of enterprise AI data platforms, the implementation of AI factory cluster solutions, and large technology companies continue to increase AI infrastructure through bond issuance.

1. Baidu disclosed AI business revenue for the first time, and AI native marketing increased by more than two times

1. On the evening of November 18, Baidu released its financial report for the third quarter of 2025, with a total quarterly revenue of about 31.2 billion yuan, of which Baidu's core revenue was about 24.7 billion yuan, and for the first time, it disclosed AI business revenue separately, with a year-on-year increase of more than 50%, AI cloud revenue increased by more than 30% year-on-year, and AI application revenue was about 2.6 billion yuan. The revenue of AI native marketing services was about 2.8 billion yuan, a year-on-year increase of more than twofold.
2. The financial report also emphasizes the fully unmanned operation and globalization progress of Radish Run, and mentions that Wenxin Model 5.0 and Kunlun Core new products will continue to form a "model + chip + application" synergy, which is conducive to more directly transforming large model capabilities into revenue and profits.

2. The "2025 Artificial Intelligence + Conference" discusses the "AI Next Decade" and the direction of AGI

1. From November 15th to 17th, the "2025 Artificial Intelligence + Conference" was held at the Zhongguancun International Innovation Center in Beijing, with the theme of "AI Next Decade: Scenario-Driven × New Quality Engine".
2. Academicians and experts generally believe that artificial general intelligence (AGI) will become a key direction in the next ten years, and China has advantages in application scenarios and industrial bases.

3. Shanghai Jiaotong University demonstrates trusted artificial intelligence and large model security shooting range at the Wuzhen Summit

1. During the Wuzhen Summit of the World Internet Conference, Shanghai Jiao Tong University systematically displayed the latest achievements of trusted artificial intelligence and active security protection at the School of Cyberspace Security Achievement Exhibition, including multi-modal large model security shooting range, AI security vulnerability library and AI security open source community.
2. Relevant technologies have been piloted in key industries such as financial risk control and industrial Internet, forming a complete innovation chain of "offensive and defensive evaluation, vulnerability management, community co-construction, and industry implementation", and has reached further cooperation intentions with a number of enterprises and scientific research institutions, emphasizing the strengthening of security and compliance while developing generative AI.

4. OpenText released AI Data Platform, focusing on "trusted data + AI agent"

1. On November 18, OpenText launched OpenText AI Data Platform (AIDP) at its annual conference, integrating data management, compliance governance and AI capabilities into a unified platform, supporting multi-cloud, multi-model, and multi-application scenarios.
2. On top of its Aviator agent system, the platform provides data compliance, privacy protection, and threat detection capabilities, emphasizing "starting from data rather than models" to help large enterprises deploy business process-oriented AI agents under the premise of ensuring audit and security.

5. Supermicro released an AI factory cluster solution based on NVIDIA Blackwell

1. Supermicro announced the launch of a new generation of AI Factory cluster solution at the Supercomputing Conference held in St. Louis, USA, based on NVIDIA enterprise reference architecture and Blackwell series GPUs, providing from 4 nodes 32 Standardized configuration of GPUs to 32 nodes and 256 GPUs.
2. The solution is pre-integrated with software and hardware components such as NVIDIA AI Enterprise, Omniverse, and Spectrum-X Ethernet, and adopts factory-level pre-integration and multi-rack testing, positioning it as an "out-of-the-box" AI factory to help enterprises build large model training and inference infrastructure faster.

6. Amazon issued $15 billion in bonds to bet on AI infrastructure, and the AI bubble debate heated up

1. According to comprehensive reports from a number of technology and financial media on the same day, Amazon issued about $15 billion in corporate bonds on November 18, with a subscription demand of about $80 billion, and the funds raised were mainly used for AI infrastructure investment, mergers and acquisitions, and share buybacks.
2. Market analysis believes that technology giants such as Amazon, Google, Meta, and Oracle have intensively raised funds through bond issuance to build data centers and AI computing power, which is regarded as part of the "AI super capital expenditure cycle" and has further heated up the debate on whether AI has appeared in a bubble.

Frequently Asked Questions (Q&A)

Q: What is the real key signal of Baidu's disclosure of AI revenue this time?

A: First, for the first time, the AI business was separated separately, proving that large models and AI applications have become business sectors that can be independently evaluated by the capital market; Second, the growth rate of AI application and AI native marketing revenue is significantly higher than the whole, indicating that "AI functions" are truly transforming into subscription and advertising revenue, not just storytelling.

Q: What is the main judgment of the "AI Next Decade" conference on China's AI industry?

A: The main line can be summarized as "AGI direction + scenario driven + new quality productivity": that is, aiming at general artificial intelligence in the medium and long term, and driving the implementation through high-value scenarios such as finance, manufacturing, and urban governance in the short term, and closely binding large model capabilities with industrial upgrading and new quality productivity construction.

Q: What is the real value of trusted artificial intelligence and large model security ranges for enterprises?

A: It transforms "security" from an abstract slogan into a testable and quantifiable capability: enterprises can simulate attack and overreach scenarios in the shooting range to verify whether the model will leak privacy, generate illegal content, or be maliciously exploited, thereby completing a "physical examination" before going live and reducing compliance and reputational risks.

Q: How will OpenText and Supermicro's overseas actions affect the deployment of AI by enterprises?

A: The former starts with data security and AI agents to solve the problems of "dare not use" and "cannot be used", while the latter starts with computing power and cluster engineering to solve the problems of "not fast to build" and "unstable luck"; The combination of the two means that global enterprises are accelerating from "trial AI tools" to "systematically building AI bases and AI factories".

Q: Is Amazon's $15 billion bond issuance a signal of an AI boom or a bubble?

A: In the short term, this is the normal operation of large technology companies to lock in low-cost funds and seize AI infrastructure opportunities; However, in the context of the continued high valuations of companies such as NVIDIA, such intensive and huge bond issuance will also amplify the market's concerns about "whether AI capital expenditure is excessive", and the realization of future performance will become a key verification point.