Today’s top line: Nvidia’s CEO resigned, rattling markets and the AI supply chain, while Google unveiled Gemini 4 Hydrogen and a Decisions API that promise faster, cheaper multimodal capabilities for search and developer workflows.
Nvidia CEO Jensen Huang resigns; CFO named interim CEO
Summary: Jensen Huang resigned as Nvidia chief executive; the company named chief financial officer Colette Kress as interim CEO. Nvidia shares fell more than 10% in pre-market trading, a move the article says would wipe off more than $400 billion in market capitalisation.
Digest: Huang led Nvidia for more than 33 years as the company became central to AI infrastructure, so his departure creates immediate leadership and market uncertainty at a firm critical to AI compute. Watch for shifts in strategy and investor reactions as the company settles on permanent leadership.
Source: Ars Technica
Google DeepMind releases Gemini 4 Hydrogen with 1M-token context
Summary: Google DeepMind introduced Gemini 4 Hydrogen, a model claiming higher scores on the Artificial Analysis Intelligence Index (51 vs 50 for Argon, reaching 54 at its highest setting) and a 1 million-token context window. Hydrogen is priced flat across that window at $1.50 per million input tokens and $7.50 per million output tokens and is positioned as a fast execution model for coding and agentic workflows.
Digest: Hydrogen aims to be a lower-cost, high-efficiency engine for day-to-day developer tasks and multi-model workflows; DeepMind also introduced Orchestrator to switch models automatically, which it says cut quota use by 24% and sped completion by 19% in benchmarks. Businesses that run heavy developer or agent workloads should test whether Hydrogen reduces token costs and latency in their pipelines.
Source: Google DeepMind
Google launches Gemini Decisions API for fast multimodal choices
Summary: Google released the Gemini Decisions API powered by the gemini-decision-1 model, which can evaluate up to 1,000 yes/no, choice, and rubric-scoring questions in a single pass across text, images, audio, video and PDFs. The API returns probabilities and confidence scores with end-to-end latency of about 150-250 ms and is billed at $0.10 per 1M input tokens with zero output tokens billed.
Digest: The Decisions API is built for bounded, automated decisions such as routing, policy checks and grading; Google says it is nearly 5x cheaper on shared benchmarks than Gemini 3.5 Flash-Lite. Expect adoption where fast, cheap, calibrated decisions can replace manual routing or screening steps in production systems.
Source: Google
TSMC to invest $1B in OpenAI and Anthropic
Summary: Taiwan Semiconductor Manufacturing Company (TSMC) approved plans to invest up to $500 million in OpenAI preferred stock and another $500 million to buy Anthropic common shares at the latter’s expected IPO price. The article notes TSMC reported more than $23 billion in profit for the July through September 2026 quarter.
Digest: The investments further entwine chip suppliers and AI labs and may calm investor nerves as the AI trade shows signs of strain, but analysts worry circular financing could obscure true market demand for AI hardware and services. Watch for more supplier-customer financing as a way to stabilise capacity and demand.
Source: Ars Technica
United Airlines blocks AI agents from booking flights, citing cybersecurity
Summary: United Airlines said it will restrict access for personal, third-party AI agents and assistants from booking flights and managing reservations, citing cybersecurity risks and the possibility of violating federal aviation rules that require human acknowledgement for certain checks. The airline will update booking, ticketing and distribution policies and work with travel agencies and aggregators.
Digest: Companies handling sensitive accounts are increasingly blocking autonomous AI assistants to prevent credential sharing and unintended actions. Expect similar restrictions from banks, travel platforms and other services that must meet regulatory or security obligations.
Source: TechCrunch
SoftBank posts $7.1B loss after over-ordering turbines for AI power needs
Summary: SoftBank announced a $7.1 billion quarterly loss, with over half attributed to losses on power-generating equipment after the company over-ordered turbines and paid nonrefundable deposits. The firm had placed deposits on more than 13 gigawatts of turbines, around 50% more than it needed, as it tried to secure capacity for AI-related power demand.
Digest: The episode highlights the risks in betting on dedicated natural-gas power to meet AI data centre demand and suggests operators may prefer grid connections plus batteries instead of tying up capital in oversized gas capacity. Expect more cautious energy procurement strategies from data-centre and AI firms.
Source: TechCrunch
MIT method trains climate emulators faster with less data
Summary: MIT researchers developed a training technique that produces climate emulators – fast surrogate AI models of earth system models – using a small amount of carefully selected synthetic data. The emulators run in seconds, adapt better to unseen scenarios, and the work appears this week in the Proceedings of the National Academy of Sciences.
Digest: Faster, more trustworthy emulators let scientists test many policy-relevant climate scenarios quickly and with lower compute cost. Governments and firms using climate projections should watch for tools that speed scenario analysis and improve decision support.
Source: MIT News
Quick hits
- Parents protest and sue over Gemini on school Chromebooks – Families and advocacy groups protested in cities including Chicago after Google said Search would always include AI; a lawsuit by a 13-year-old alleges harmful chatbot interactions on a school-issued device (Ars Technica)
- Gemini-powered smartglasses debut, starting at $395 – Google, Samsung and eyewear partners launched glasses from Warby Parker and Gentle Monster with Gemini voice and camera features, privacy controls and 12 MP cameras (Ars Technica)
- Hugging Face CEO pledges to defend open-source AI after Huang exit – The company says it will keep pushing for open-source models amid uncertainty over Nvidia’s future stance (TechCrunch)
- OpenAI commits $2M to research on AI and children’s safety – Funding will support shared tools to measure risks and opportunities for young people (OpenAI)