Daily Highlights
- OpenAI cuts funding target to ~$20B
- Anthropic IPO filing imminent
- DeepSeek open-sources Super-agent architecture
- Tencent confirms Hunyuan Hy4 launch
- Chinese models lead global inference for 26 weeks
HEADLINES
[OpenAI Slashes Funding Target by $10B; Valuation Expectation Drops to $1.25T] According to the Financial Times, OpenAI has cut its latest fundraising target by roughly $10 billion to about $20 billion amid the tech stock rout, with its desired valuation dropping from $1.4 trillion to $1.25 trillion. Reuters also reported that OpenAI is seeking to accelerate its fundraising ahead of Anthropic's IPO, as the primary-market environment for AI capital tightens noticeably. (Source: FT/Reuters)
[Anthropic IPO Nears, Expected to File imminently] Anthropic is gearing up for what is expected to be one of the largest tech IPOs in US history, reportedly filing as early as Friday at a valuation estimated between $965 billion and $2 trillion. Its prospectus shows H1 2026 revenue of about $16.3 billion, up 13x year-on-year, alongside a net loss of $80.7 billion, with compute and infrastructure costs remaining its largest expense. (Source: Reuters)
NEW PRODUCTS & TECH
[Tencent Confirms Hunyuan Hy4 Full Version and Hy4 Flash, First to Add Visual Understanding] At the 2026 Tencent Global Digital Ecosystem Summit on Oct 29, Senior EVP Dowson Tang announced that Hunyuan Hy4 will launch soon: compared with the open-source Hy4 preview (770B total params, 49B active, 1M context), the full version strengthens long-horizon and real productivity tasks, improves chain-of-thought efficiency, and adds visual understanding for the first time; the cost-effective Hy4 Flash will follow shortly. Tencent Cloud also unveiled Buddy One, a cross-app AI collaboration system. (Source: Beijing News/Zhidx)
[DeepSeek Unveils Super-agent Architecture: AI Orchestrating AI] On the evening of Oct 31, DeepSeek's Harness team published a technical article titled "Super-agent Design Exploration: Toward More Autonomous Agent Systems," unveiling a new architecture for autonomous agents with the prototype open-sourced. The design targets the multi-task bottleneck: as single tasks need less human intervention, human attention becomes the limit, and Super-agent breaks it by letting AI orchestrate AI. (Source: Global Times)
IPO & FUNDINGS
[AI Capital Chill Spreads to Public Markets; Nasdaq Snaps Losing Streak Late in Week] Through late October, AI-bubble fears pushed the Nasdaq to consecutive weekly declines, wiping trillions of dollars in market value. The primary market cooled in tandem, with OpenAI trimming its target and reports that major banks could assign Anthropic's IPO a low credit rating. Sentiment recovered on Friday, however, as Anthropic's filing buoyed AI stocks and both the S&P 500 and Nasdaq closed higher. (Source: Reuters/Sina Tech)
OPEN SOURCE & RESEARCH
[DeepSeek Open-sources Super-agent: A Multi-agent Design for AI Orchestrating AI] DeepSeek released the Super-agent prototype to the open-source community alongside its technical writeup, directly addressing the ceiling of human multi-task attention and offering a new reference implementation for agent system design. (Source: DeepSeek/Global Times)
COMPANY & PEOPLE
[Tencent Cloud: 2M+ Servers Globally, Overseas AI Demand Grew Triple Digits] At the ecosystem summit, Dowson Tang revealed that Tencent Cloud added or expanded data centers in Frankfurt, Malaysia, Mexico, Osaka, Riyadh and Thailand over the past year, now runs more than 2 million servers worldwide, and saw overseas AI demand grow at a triple-digit rate, with further expansion planned across Asia, the Middle East, Europe and the Americas. (Source: Eastmoney/Zhidx)
[World Laureates Forum Opens in Shanghai; Nobel Laureate Geim on AI Safety] From Oct 30 to Nov 1, the 2026 World Laureates Forum gathered nearly 300 laureates and industry leaders in Shanghai. Andre Geim, 2010 Nobel Prize winner in Physics, said AI holds no magic: it is just a computing machine, without perception or life. (Source: China News Service)
POLICIES & DATA
[OpenRouter Data: Chinese Models Lead Global Inference for 26 Straight Weeks] According to OpenRouter data for the week of Oct 19-25, DeepSeek V4.1-Flash-1017 topped global usage with 33.6 trillion tokens in a single week. The global top five were all Chinese models, totaling 93.54 trillion tokens versus 22.47 trillion for US models, marking the 26th consecutive week Chinese models led in total usage. (Source: China News Service)
Disclaimer: For reference only; exercise caution for investment or decisions.
---
Powered by AOYII.com
