In the past 24 hours (October 24), North America and Asia-Pacific simultaneously released major AI developments: the European and American markets focused on computing power investment and application security, while Asia focused on top-level policy design and industry implementation.
1. Anthropic and Google reach a large TPU computing power order, with the scale reaching one million units
- Disclosure on October 23-24: Anthropic will expand its use of Google TPU computing power, with the scale reaching 1 million TPUs, with a value of "tens of billions of dollars".
- The corresponding computing power capacity exceeds 1GW, which is expected to be available in phases starting from 2026 and will be used for training and inference of subsequent versions of Claude.
- Google uses this to strengthen the availability of computing power ecosystem in the "non-NVIDIA" route, alleviate the shortage of AI GPUs, and take a step towards diversifying the industry landscape.
2. Microsoft releases Copilot "Fall Edition" and Edge "Copilot Mode"
- The human-oriented update launched on October 24th strengthens cross-scenario personalization and usability improvements.
- Edge has added a new "Copilot Mode", which can sense tabs and perform in-browser actions such as summaries and ticket booking; emphasizing safety guardrails and controllability.
- In contrast to this week's browser-level AI race, enterprise-side collaboration and compliance will become the differentiating points.
3. Anthropic expands its presence in Asia: announcing the opening of an office in South Korea
- Confirmed on October 24: A branch will be established in South Korea to expand corporate customer service capabilities in core markets such as Japan, South Korea, and India.
- The company said that three of the world's top five customers are from Asia, and the commercialization growth rate in Asia-Pacific is obvious.
- The cooperation with TPU and the Asian layout demonstrate its "multi-cloud + multi-regional" operation strategy.
4. AI security false alarm sparks controversy: Student mistakenly identifies potato chip bag as a gun
- Report on October 24: An AI gun detection system at a high school in Maryland, USA, gave a false alarm. A student was handcuffed and taken away on the spot. It was later clarified that the gun was a snack bag.
- Highlight the false alarm costs and governance responsibilities of computer vision in public safety scenarios.
- Local education department and police procedures will be reviewed, with supplier identification and testing standards receiving particular attention.
5. Bloomberg: AI data center boom changes US manufacturing investment structure
- An analysis on October 24th stated that factory construction spending has fallen by about 2.5% year-on-year so far in 2025, while data center spending has increased by nearly 18%.
- AI infrastructure absorbs capital and electricity allocation, or squeezes the pace of traditional manufacturing expansion.
- Structural redistribution of industrial policies, energy and tax incentives becomes the focus.
6. During the 15th Five-Year Plan period, AI top-level design and the "AI+" initiative will be intensified and promoted.
- Authoritative interpretation of the press conference on October 24: We will continue to strengthen the top-level design and systematic deployment of artificial intelligence.
- Key areas include: basic research and core technology development, high-end computing chips, AI governance regulations and ethics, and international collaboration and global governance.
- It also proposes to build an open, shared and secure national integrated data market and fully implement "artificial intelligence +".
7. Baidu Smart Cloud: Ranked first in the number of large-scale model projects won in the financial industry from January to September
- Third-party statistics: From January to September 2025, there were 358 successful bids for large-scale model projects in the financial industry, a significant year-on-year increase; the disclosed amount was approximately RMB 955 million.
- Baidu Smart Cloud leads in both the number and amount of winning bids; its services cover multiple institutions including banks, insurance companies, and securities companies.
- The financial industry has shifted from "pilot exploration" to "large-scale deployment", and AI application scenarios have been accelerated.
8. CPIC Technology and Baidu Smart Cloud have reached a strategic partnership to jointly build a generative AI insurance laboratory
- Signing of the contract on October 24: The two parties will jointly tackle key issues such as computing power operation, domain large models, knowledge management and multimodality.
- Create "insurance intelligence" applications for policy circulation, claims risk control, customer service operations and other aspects.
- The goal is to achieve "independent and controllable" core AI capabilities and promote industry-level implementation models.
9. Shanghai West Bund AI Technology Consumer Carnival Opens: 300+ Companies Participate, 40+ New Products Launch
- Opening on October 24: Covering scenes of creation, entertainment, life, etc., the immersive exhibition area exceeds 10,000 square meters.
- More than 40 new products will be launched one after another, and more than 30 types of robots will be displayed. The "Robot Kingdom" will become a highlight.
- More than 100 AI consumer products can be ordered on-site, verifying the trend of "AI consumption" transforming from experience to transaction.
10. iFLYTEK's 1024 Developer Festival kicked off online, revealing the "four-layer architecture" of the Star Intelligence entity.
- The online agenda will be launched on October 24th, and the offline main forum will be held in Hefei on November 6th, when the Spark model upgrade will be released.
- Today's technology sharing revealed the four-layer architecture of the "Xingchen Intelligent Body" platform, emphasizing the systematic foundation of tool calling and multi-modal interaction.
- The official investor document also emphasized that Spark will enhance memory and multi-mode interaction, positioning itself as an upgrade route on the "national computing power base."
Frequently Asked Questions (Q&A)
Q: What are the key numbers and timelines for the Anthropic-Google TPU collaboration?
A: The scale will reach 1 million TPUs, with a value of "tens of billions of dollars". It is expected to be supplied in phases starting from 2026, and the goal is to be used for training and inference of subsequent Claude models.
Q: What is the most direct value of this round of Microsoft Copilot updates to enterprises?
A: Stronger human-centric experience and task automation capabilities; Edge "Copilot Mode" performs summaries, ticket booking and other actions within the browser, while emphasizing compliance and safety guardrails, which facilitates rapid implementation in office scenarios.
Q: What are the four key words in China’s 15th Five-Year Plan’s top-level AI design?
A: Basic research and key technologies, high-end computing power and data elements, AI governance and ethical standards, international collaboration and global governance; and supporting the construction of a national integrated data market and the comprehensive implementation of "artificial intelligence +".
Q: Why is AI bidding and tendering in the financial industry growing rapidly and Baidu Smart Cloud leading the way?
A: Financial scenarios are highly standardized, compliance is auditable, and ROI is quantifiable, driving large-scale adoption; Baidu has an advantage in combining full-stack capabilities with industry know-how, leading the way in both the number and amount of winning bids.
Q: What are the business signals that Anthropic is increasing its investment in Asia?
A: Setting up an office in South Korea and claiming that three of the world's top five customers are from Asia, combined with Google TPU and multi-cloud strategy, demonstrates its long-term investment in the Asia-Pacific enterprise market.