AI Reading for Tuesday November 4

Amazon reported revenue of $180 billion and net profit of $21.2 billion as AWS revenue rose 20%, it added 3.8 GW of power, signed a $38 billion OpenAI GPU deal, and expanded same‑day grocery delivery to over 1,000 U.S. cities. - Stratechery by Ben Thompson

OpenAI signs $38 billion compute deal with Amazon, partnering with cloud leader for first time. - CNBC

Press release with big meaningless number, (e.g. option to buy up to $38b at a good price when GPUs are actually available) Amazon stock up 5%, everybody's a winner.

Noteworthy that both Nadella and Jassy say they hit the wall on power, have plenty of GPUs but nowhere to plug them in, seems not positive for certain high-priced chipmakers.

Nvidia agreed to invest up to $100 billion in OpenAI to finance at least 10 GW of data‑center capacity, while OpenAI signed a $300 billion cloud deal with Oracle (4.5 GW by 2027) and a 6 GW AMD deal. - The Register

How to build a trillion-dollar industry: 1) invest in your customers and make an announcement with big numbers, 2) their stock goes up. 3) profit! your investment went up, and you can sell them more stuff.

Alphabet sold $25 billion of bonds — $17.5 billion in the U.S. and €6.5 billion in Europe — in eight tranches maturing 3–50 years, with the longest tranche yielding 1.07 percentage points over Treasuries to fund AI expansion. - Yahoo

AI training firms are hiring freelance specialists—from physicians to art historians—to create problems, solutions and rubrics that teach models complex tasks, while Meta announced a $14.3 billion investment in Scale AI and hired its CEO to lead a Superintelligence lab. - Fast Company

Carnegie Mellon professor Zico Kolter will chair OpenAI's four-person Safety and Security Committee, created under California and Delaware regulatory conditions, with authority to request delays or halt AI system releases and full observation rights to for-profit board meetings. - Seattle PI

Beijing introduced grants cutting power costs by up to 50% for some of China’s largest data centres to support domestic AI chip development and incentivize onshore AI training, infrastructure investment and demand for domestically produced AI chips. - Financial Times

BlackRock portfolio manager Charles Lilford said an AI bust would not threaten sustainable energy equities, as electrification and decarbonization will continue regardless - Financial Post

Loop Capital raised its Nvidia price target to $350, implying an >$8.5 trillion market cap, as Nvidia recently passed a $5 trillion valuation and Loop forecasts Blackwell GPU unit shipments roughly doubling over 12–15 months amid generative AI demand. - Bloomberg

Big Short portfolio manager disclosed big bets against Nvidia and Palantir - Markets Insider

Jensen Huang says there would have been no Nvidia with Trump immigration policies - International Business Times

Tesla shareholders will vote Thursday on CEO Elon Musk's proposed pay package worth up to $1 trillion and a separate proposal to authorize investment in Musk's AI company xAI, with three directors also up for election. - Yahoo

Morningstar analyst Seth Goldstein said investors were "a little too carried away" with Tesla's robo-taxi and said he was skeptical of a full driverless, non-geofenced 2026 launch, while Macquarie said it never held Tesla due to volatility. - Yahoo

Google Cloud's Cybersecurity Forecast 2026 warns threat actors will fully embrace multimodal generative AI, increasing phishing, malware and impersonation and forecasting ransomware, data theft and multifaceted extortion as the most financially disruptive cybercrimes in 2026. - SiliconANGLE

Palo Alto Networks CEO Nikesh Arora warned enterprises are unprepared for AI agents accessing privileged resources, citing attackers exploiting smishing from 194,000 domains, and Palo Alto is integrating CyberArk tools and launching Cortex AgentiX trained on 1.2 billion playbook executions. - ZDNET

HackerOne leaders say agentic AI—autonomous agents forecast to appear in about one-third of enterprise applications by 2028 amid a projected 36.6% CAGR in AI adoption (2023–2030)—will automate SOC triage, scale pentesting, and create new attack vectors requiring governance. - TechRadar

Anthropic announces a deal with Cognizant, under which Cognizant will deploy Claude to its 350,000 employees and co-sell Claude models to its business customers - The Wall Street Journal

Anthropic researchers used concept injection to plant neural activation patterns in LLMs, finding models sometimes detected injected "thoughts"—Claude Opus 4.1 detected injections about 20% of the time—and more capable models showed stronger but unreliable introspective signals. - Hacker Noon

At London's High Court on Tuesday, Judge Joanna Smith largely dismissed Getty Images' copyright claims against Stability AI over Stable Diffusion, while finding limited trademark infringement for Getty watermarks after Getty dropped a key reproduction claim mid-trial. - Yahoo

Salesforce launched Agentforce under CEO Marc Benioff, reporting over 3 billion uses and customers including PepsiCo (helping 1.5 million stores, targeting 5 million by 2026), yet fewer than half of 12,500 customers pay and adoption is limited. - Business Insider

Analysis of ~180 million global job postings (Jan 2023–Oct 2025) found total postings down 8% in 2025 versus 2024, with creative and compliance roles falling up to 35% while machine learning engineers rose 40%. - Bloomberry

An IEEE survey of 400 senior tech leaders found demand for software-development skills in AI roles fell to 32% (-8 points), AI ethics rose to 44%, and 39% plan agentic AI for development while 47% plan it for real-time cybersecurity. - The Register

Personalized AI tutoring and automated grading reshape education but raise quality, privacy and equity concerns - eLearning Industry

Coca‑Cola refreshes its 'Holidays Are Coming' campaign with AI-generated creative - The Wall Street Journal

OpenAI Wants Brands to Allow Their Mascots to Appear in Gen AI Videos - The Wall Street Journal

AI in space: networks of weather balloons and satellites, with sensors that generate big data, can process it in place before beaming it to ground stations, speeding analysis. - IEEE Spectrum

AI to detect your mood and adjust your surroundings, maybe detect if a call-center employee is about to melt down and go postal - Ubergizmo

Startups like Known replace swipe-driven dating discovery with AI "matchmakers," exemplified by a user who spent a 20-minute AI chatbot call and received a paid in-person match offer for $25. - The New York Times

AI as a systemic problem. Sometimes makes nefarious collective action easier and positive collective action harder. - The New York Times

A prompt to build a press release … might do even a little better if you generate a few versions, merge them, add a critic-optimizer loop - Hacker Noon

These folks are deranged, there is really no way to interact successfully with deranged people acting with animus, I mean you can try to avoid a fight with a crazy guy in the street, but eventually you just have to defeat them and kick them out - The White House

Artificial intelligence song generation platform Udio said it would give its frustrated users 48 hours starting Monday to download their songs before the company shifts to a new business model to comply with a legal settlement. - ABC News

hat tip to a reader! AI assisted music mixes are getting very big and very good and possibly engendering more acceptance of AI in music.

Follow the latest AI headlines via SkynetAndChill.com on Bluesky