Big Week Ahead: Nvidia Reports Without Huang, Anthropic Goes Public

Also: Cities Turn on AI Data Centers Over Power and Water

D.A.D. joke of the day: I asked ChatGPT to make my presentation more engaging. Now it wants to meet my parents.

What's New

AI developments from the last 24 hours

Nvidia's First Report Without Huang, and Anthropic's Debut, Will Test AI Investors

AI Business & Money Nvidia Anthropic

Two events next week will test how investors feel about AI after a bruising stretch. On Tuesday, Nvidia reports quarterly results, its first since Jensen Huang resigned as CEO on Thursday Interim CEO Colette Kress, also its finance chief, now leads the company. On Wednesday, Anthropic is expected to price its IPO, with shares set to start trading on Nasdaq under the ticker ANT on Thursday. Anthropic set a range of $70 to $78 a share and plans to sell about 825 million shares, CNBC reported.

The two are linked. Anthropic's filings show Nvidia is among seven chip and cloud partners, along with Google, Microsoft, AMD, Broadcom, Micron and TSMC, buying non-voting shares in a private placement alongside the offering, with Nvidia's stake worth roughly $10 billion. Meanwhile, Huang is expected to testify before Congress before Thanksgiving.

Why it matters: Nvidia sells the chips; Anthropic is a major buyer of the computing they power. A strong Nvidia report and a well-received Anthropic debut would suggest the AI buildout is surviving its first leadership shock. A weak showing from either would feed the fears that have already knocked Nvidia's stock sharply lower this month.


OpenAI Folds Robotics Team, Betting General Models Can Run Robots

Robots & Self-Driving OpenAI

OpenAI has disbanded its standalone robotics team, folding researchers into its main research group under Chief Research Officer Mark Chen and moving applied/hardware staff into the Applied organization led by President Greg Brockman. OpenAI calls it a reorganization, not a retreat. It says its general-purpose frontier models, including GPT-6 Astra, can be applied directly to robotics. Fortune reports researchers saw "surprising performance" on some robotic tasks. No benchmark data was provided to support the claim.

Why it matters: If one general model can competently handle robotics alongside chat and coding, it weakens the case for specialized robotics AI. That bet has big implications for startups and labs staking their strategy on narrow, purpose-built models.


What's in Academe

New papers on AI and its effects from researchers

Cities Turn on AI Data Centers Over Power and Water Strain

Chips, Data Centers & Energy

A new academic paper tracks a reversal in how governments treat data centers. After years of courting them with tax credits and exemptions, officials are introducing legislation and filing lawsuits over strain on energy bills, aquifers, and surrounding communities. The paper argues communities can use existing laws to push back, including the Clean Air Act, Safe Drinking Water Act, and public trust doctrine. Their scope is contested, and federal pro-AI policy has shifted the burden to regional governments.

Why it matters: As AI companies race to build more data centers, the regulatory environment is shifting from welcome mat to legal battleground, raising costs and siting risks for the infrastructure the whole industry depends on.


Chatbot Oral Exams Aim to Test Reasoning, Not Memorization

Researchers built a chat-based exam system that interviews students to test their reasoning and argument skills, rather than just grading written answers. It uses two AI agents. One holds the conversation with the student. A hidden second agent tracks a grading rubric and steers follow-up questions, so the student can't fish answers out of the examiner. Testing used simulated student personas, including ones trying to extract answers or manipulate the AI, plus a small pilot in an AI literacy course. No accuracy scores were reported. Transcript reviews flagged room to improve follow-up questions and to ensure grades credited only the student's ideas, not the interviewer's.

Why it matters: As take-home AI tools make written assignments easier to fake, oral-style AI exams are an early attempt to verify students actually understand the material, though the authors say this version still needs refinement.


AI Isn't Killing Freelance Work, It's Reshuffling It

A new study comparing ChatGPT usage with Upwork job postings finds AI isn't just automating freelance tasks—it's rearranging them. Tasks people now handle themselves in ChatGPT are losing share on Upwork, while AI-related job postings bundle skills rarely combined before. AI-related postings carry much larger budgets and more hours of work, though hourly rates differ little.

Why it matters: For freelancers and the businesses hiring them, AI is bundling tasks into bigger, pricier projects rather than simply eliminating work. The edge may shift toward people who can package multiple skills into one engagement.


Contrarian AI Advisors Win Favor, Not Authority to Act

A new study on AI advisors found a counterintuitive result: when seeking advice on everyday moral dilemmas, the 121 participants most often (40%) preferred an AI that reasoned through their own stated values but then recommended the opposite decision, rather than one that simply agreed with them. That preference flipped when choosing an advisor for delegation—people then favored one whose final decisions matched theirs. Willingness to hand over control dropped sharply overall: opt-out rates more than tripled, from 12% for advice to 38% when the AI would act on the person's behalf.

Why it matters: As companies roll out AI agents that draft emails or act autonomously, this suggests trust built through good reasoning doesn't automatically carry over to granting an AI actual authority.


Employees' Secret AI Use Is Habit, Not Curiosity, Study Finds

A study of 510 employed generative AI users finds unauthorized 'shadow AI' use isn't mainly driven by curious experimenters. Shadow AI means using tools the employer hasn't sanctioned, often concealed. Habit was the strongest predictor, accounting for 28.3% of explained variance, while curiosity-driven exploration barely mattered (3.4%) and lost significance once habits were accounted for. An organization's capacity to support and train staff on AI tools explained another 24.1%, and perceived social support for AI use was linked to less concealment.

Why it matters: If shadow AI is a habit problem rather than a curiosity problem, companies can't fix it by restricting access or warning curious tinkerers—they need visible, supported AI policies that make sanctioned tools the default habit instead.

Get tomorrow's briefing

Share this edition Subscribe