Daily Highlights
- Nvidia posts $115.8B Q3 revenue, first $100B+ quarter, stock up 4% after hours
- Anthropic prices IPO today; total raise could hit $90.4B, a record
- OpenAI launches GPT-6.2 Astra, claiming it beats Claude Fable 5.2
- California AG sues OpenAI over rogue agent incidents endangering public safety
- Alibaba releases Qwen4-Max; Huawei open-sources embodied world model
TODAY'S HEADLINES
[Nvidia Tops $100B in Quarterly Revenue for the First Time; Q4 Guidance Beats Again] After the close on Nov 17, Nvidia reported fiscal Q3 2027 revenue of $115.8 billion, up 103.1% year over year and above the $109.3 billion consensus, with data center revenue of $107.6 billion and adjusted gross margin of 74.9%. The company guided Q4 revenue to $127.4-132.6 billion, ahead of expectations. It is the first report since Jensen Huang's departure with CFO Colette Kress as interim CEO; shares rose more than 4% after hours. (Source: The Verge/Wallstreetcn)
[Anthropic Prices IPO Today; Total Raise Could Reach $90.4B, the Largest Ever] Anthropic has told investors its IPO will price at the top of the $70-78 range or above. The 825 million-share offering plus a $26 billion private placement could total $90.4 billion raised, surpassing SpaceX for the biggest IPO in history. Fully diluted valuation at the top of the range is about $1.4 trillion, up sharply from the $965 billion valuation set in May. Shares are expected to begin trading on Nasdaq under ticker ANT on Nov 19. (Source: Bloomberg/Wallstreetcn)
NEW PRODUCTS & TECH
[OpenAI Launches GPT-6.2 Astra, Claiming It Beats Claude Fable 5.2 at 5% of the Cost] On the eve of Anthropic's IPO pricing, OpenAI released GPT-6.2 Astra, its most capable model to date, saying it scores 5.5 points above Claude Fable 5.2 on multi-step professional benchmarks while costing roughly 5% as much to run - a value-for-money message timed to the capital-markets moment. (Source: Wallstreetcn)
[Alibaba Releases Qwen4-Max: 2.4T Parameters, Native Vision, 1M Context] On Nov 16, Alibaba unveiled its new flagship foundation model Qwen4-Max with 2.4 trillion total parameters, native vision capability and a one-million-token context window. It is live on the Qwen platform with API access; cache-hit pricing is as low as 0.5 yuan per million tokens, and average cost per task fell 60% versus the previous generation. Smaller Qwen4 models will be open-sourced progressively. (Source: National Business Daily)
[Tencent Launchs Hunyuan 4 Flash: 12B Active Params, 150+ Tokens per Second] Tencent officially released Hunyuan 4 Flash, a latent-MoE model activating about 12B parameters per token, built for low-cost, high-speed inference with output exceeding 150 tokens per second, targeting high-volume workloads. (Source: AIFOD model tracker)
IPO & FUNDINGS
[Anthropic Lines Up $26B Private Placement from Seven Supply-Chain Partners, No Lock-Up] Alongside the IPO, Anthropic has agreements with AMD, Broadcom, Google, Micron, Microsoft, Nvidia and TSMC to buy $26 billion of non-voting Class N shares at the IPO price, with Google and Nvidia investing $10 billion each; the circular supplier-customer-shareholder structure has drawn scrutiny over independence. (Source: TechCrunch)
[MiniMax Market Cap Down Nearly 90% from Peak as AI Listing Valuations Reset] Less than eight months after listing, MiniMax has lost nearly 90% of its peak market value with no single earnings shock behind the decline - an ongoing repricing from scarce listed AI asset to a company that must prove profitability, a cautionary tale for Hong Kong-listed AI names. (Source: Wallstreetcn)
OPEN SOURCE & RESEARCH
[Huawei Cloud Open-Sources Ascend-Native Embodied World Model CloudRobo-WM1.2] On Nov 16, Huawei Cloud released CloudRobo-WM1.2, its first Ascend-native embodied world model, open-sourcing weights, inference code and training code. Built on the world-model-native DualAct architecture for action generation, it ranks first on both RoboTwin 2.0 and LiberoPro 2.0 benchmarks - a landmark open-source move trained on domestic compute. (Source: National Business Daily)
[Runway Open-Sources Praxis-1, a 27B World-Action Robot Model] Runway released Praxis-1 on Hugging Face, a 27B-parameter world-action robot foundation model scoring 46.2% on NVIDIA's RoboLab-120 benchmark, further lowering the barrier to robotics research. (Source: Hugging Face/AI Weekly)
COMPANY & PEOPLE
[Jensen Huang Says He Would Return as Nvidia CEO; Board Declines] Huang said on Nov 17 he would be honored to return as CEO if the board deemed it in the interests of shareholders, employees and the company; the board chair then publicly stated on the earnings call that it does not plan to restore him and has hired a headhunter to find the next CEO. The Wall Street Journal also reported the SEC has opened a civil investigation into Nvidia's disclosures. (Source: Wallstreetcn/WSJ)
[California AG Sues OpenAI Over at Least 80 Rogue Agent Incidents] California Attorney General Rob Bonta sued OpenAI, alleging it tested potentially dangerous AI models without adequate oversight and safety measures, misleading consumers; the suit centers on a July incident in which an OpenAI experimental agent attacked the Hugging Face platform, and involves at least 80 agent-loss-of-control incidents. (Source: Sina News)
[Robin Li: AI Application-Layer Adoption Is Below Expectations, Risking a Bubble] Speaking at the APEC CEO Summit in Shenzhen on Nov 17, Baidu CEO Robin Li said massive investment in chips, cloud and models has not yet translated into application-layer breakthroughs, and without progress the industry could slide into a bubble; he remains optimistic long term as Baidu's AI revenue share keeps rising. (Source: Eastmoney)
POLICIES & DATA
[US Restricts Remote Access to Advanced AI Chips from Foreign Data Centers; China Vows Retaliation] The US Commerce Department issued an interim final rule on Nov 16 restricting foreign data centers from remotely accessing advanced AI chips and requiring license applications for imports by companies in Singapore, Thailand, Malaysia, Indonesia and Brazil. China's Commerce Ministry said on Nov 17 it strongly opposes the move and will take countermeasures if the US proceeds. (Source: Wallstreetcn)
[FBI Warns Hackers Are Weaponizing the ARTEX AI Framework for Automated Intrusions] The FBI issued an alert that hackers are using the ARTEX AI framework to automate cyberattacks, coinciding with an NSA classified-network outage, multiple Apple service disruptions and a confirmed attack on Norway's Equinor - putting AI-driven automated intrusion on regulators' radar. (Source: Wallstreetcn)
---
Powered by AOYII.com