Over the past three days, AI industry news has clustered around capital markets and model releases, with Anthropic's oversubscribed IPO and Google's rapid-fire launches drawing the most attention, while domestic computing build-out and industry applications moved in parallel. Here are 12 selected highlights for November 15.
Anthropic IPO about 6x oversubscribed — order books have reached roughly six times the shares offered, according to people familiar with the matter. The company is raising up to 64.4 billion dollars in the IPO plus a 26 billion dollar private placement, which would make it the largest IPO in history. Filing documents show Nvidia plans to buy about 10 billion dollars of shares, annualized revenue hit 81 billion dollars as of October 31, and the stock is expected to list on Nasdaq under ticker ANT on the 19th. (Sina Finance)
Google ships Gemini 4 Hydrogen. The new mainline model arrived on November 12, and the Flash naming line is retired. Against Gemini 3.8 Flash, it completes tasks 2.5x faster at the default high-thinking setting, lifts Terminal-Bench 4.1 scores from 18.2% to 57%, and cuts fabricated claims by more than 5x. It is live in AI Studio, the Gemini API, the Gemini app and Google Search. (NetEase Tech)
DeepMind releases Genie 4 world model. Genie 4 supports real-time interactive simulation, where user actions get instant environmental feedback — another iteration in Google's world-model line. (Sohu Tech)
Musk signs open letter urging Huang's return. On November 14, Elon Musk signed an open letter calling on Nvidia's board to bring back Jensen Huang. By 21:00 Beijing time signatures had passed 85,000, including current and former Nvidia employees and shareholders. Nvidia, responding to successor rumors, said it has not formally contacted any candidates. (Ifeng, East Money)
HK AI stocks slide, Zhipu down over 11%. Hong Kong tech fell on November 13, with the Hang Seng Tech Index down more than 3%. AI large-model names led losses, and Zhipu dropped over 11%; CLS reported a "valuation squeeze" on the two model champions. (21jingji, CLS)
Eli Lilly CEO on AI curing cancer. Asked about claims that AI could conquer cancer within a decade, Eli Lilly's chief executive said the vision is beautiful but unlikely to happen. (Ifeng Tech)
Yanggao computing center enters pre-construction. The Yanggao intelligent computing center in Shanxi has completed deployment planning for 845 standard racks and entered the pre-construction phase. (SeeDao)
Chen Fan-Chang on US-China AI competition. The HKUST former president said Chinese AI should not obsess over model rankings, nor focus merely on dodging export controls. (Caixin)
Xiaopingguo Pediatrics closes Pre-D round. The pediatric care provider announced a Pre-D financing round led by a major AI company as strategic investor. (Sina Finance)
Caixin cover on money flooding computing centers. The latest cover story examines the capital flows and return logic behind the computing-center construction boom. (Caixin)
Yao, Yang and co-authors publish AI reasoning paper. Turing Award winner Yao Qizhi, Yuan Yang's team and Moonshot AI founder Yang Zhilin jointly published a paper exploring how AI could "think like Newton", moving from exploration to proof. (BAAI Community)
Beihang professor on LLM hallucination limits. Professor Sha Lei said hallucinations cannot be fully eliminated — the more AI knows, the harder its boundaries are to draw. (Sohu)
Capital-market swings and model iteration keep pulling at each other. Per Wall Street insights outlet WSJ CN, Nvidia's earnings and Anthropic's debut land back to back next week — the next checkpoint for the AI trade. (Compiled from public reports)