Nvidia CEO Jensen Huang stepped down, with CFO Colette Kress named interim CEO. Google released Gemini 4 Hydrogen for coding and agents, and Anthropic unveiled Claude Managed Agents that run inside a company's own firewall. We also cover a lawsuit demanding a halt to frontier AI development, SpaceX's $74 billion AI campus, and Korea's physical AI alliance.
🤖 Jensen Huang Steps Down as Nvidia CEO… CFO Colette Kress Named Interim CEO
Longtime Nvidia CEO Jensen Huang resigned, and CFO Colette Kress was appointed interim CEO. Reports cited an affair and allegations of involvement with a woman U.S. intelligence authorities suspect of passing information to China, though Nvidia said it found no evidence of corporate or national-security secrets being leaked. Huang has led investment and partnerships across the AI industry, so whether his successor continues his "deal-making" approach is a key question for the industry. The company said Nvidia's product roadmap and customer commitments, including the Hugging Face acquisition, remain unchanged.
⚡ Google Launches Gemini 4 Hydrogen… Fast, Low-Cost Model for Coding and Agents
Google released "Gemini 4 Hydrogen," the third model in the Gemini 4 family. A fast model focused on coding agents and everyday AI tasks, it is priced at $1.50 per million input tokens and $7.50 per million output tokens, and is available in Google AI Studio, the Gemini API, Search, the Gemini app, and more. In Google's own evaluations, it is about five times faster than previous models and scored 51 on an intelligence index. On the same day, Google and Samsung also began taking reservations for Gemini smart glasses in a Warby Parker and Gentle Monster style, starting at $395, with shipping expected December 10.
🏢 Anthropic Unveils Claude Managed Agents That Run Inside the Firewall… IPO Ticker 'ANT'
Anthropic began a private beta of "self-hosted Claude Managed Agents," letting companies run the full Claude agent service in their own cloud or on-premises. Servers, credentials, sandboxes, and session records stay inside the company, and model call routing can be chosen among Anthropic, AWS Bedrock, Google, and Microsoft. Meanwhile, Anthropic published its IPO page, listing on Nasdaq under the ticker "ANT" with an offering price range of $70–$78 per share. Pricing is expected around November 18, and Q3 revenue was reportedly $17 billion.
⚖️ Nonprofit Coalition Sues Five Big Tech Companies, Demanding a Halt to Frontier AI Development
A coalition of nonprofits led by Public Citizen sued the AI arms of OpenAI, Anthropic, Google, Meta, and SpaceX in California state court. The suit asks the court to order a halt to developing models more powerful than existing systems, citing the extreme AI risks the companies' own executives have warned about. OpenAI said it has already stopped training its most powerful model and will not resume until additional safety measures are in place.
🛡️ OpenAI Unveils AI Security Tools 'Daybreak'… Automated Penetration Testing and Patch Suggestions
OpenAI unveiled two AI-based cybersecurity tools called "Daybreak." "Daybreak Red Team" has a team of agents automatically penetration-test web apps, while "Daybreak Defense Factory" uses GPT-6 Cyber to scan codebases and suggest patches. The red-team tool is in early access, and suggested patches are deployed only after review by engineers. GPT-6 Cyber reportedly found 81 previously unknown high-risk vulnerabilities in Firefox.
🚀 SpaceX Pushes $74 Billion AI Supercomputing Campus
SpaceX announced plans to build AI supercomputing campuses in Texas and Arizona worth more than $74 billion. The Arizona site alone is about $50 billion and still needs county approval. The Texas facility is planned to have its own power generation. The project shows the AI infrastructure race expanding beyond chips into power and land. On the same day, SoftBank said it recorded a $3.7 billion loss as it adjusted power equipment orders for AI data centers.
🇰🇷 Korea's Physical AI Alliance Takes Off… Samsung, LG, and KakaoMobility Expand Nvidia Partnerships
Following "Nvidia AI Day Seoul 2026," held September 9–10 at COEX in Seoul, physical AI cooperation with Korean companies is taking concrete shape. Samsung Electronics is accelerating semiconductor design and process simulation with Nvidia's physics-based AI platform, and LG Electronics is applying an Isaac GR00T-based robot learning system in real factories. KakaoMobility signed an MOU with Nvidia to develop a Level 4 autonomous driving system. SK Hynix, SK Telecom, and Naver Cloud also shared cases in digital twins and large-scale GPU cluster operations, as the cooperation widens beyond HBM supply into manufacturing, robotics, and autonomous driving.