Daily AI digest, November 12

Felix Laurent 13 Nov 2026 4 min read

News

The biggest AI news of November 12 came from Nvidia. Jensen Huang resigned as CEO of the company he co-founded 33 years ago, and CFO Colette Kress became interim CEO. Google released Gemini 4 Hydrogen, its first Gemini 4 model for everyone. Meta, Replit and Anthropic gave agents more real-world work, and SpaceX put more than $74B into two AI data center campuses. Below are the big stories of Thursday, November 12, plus the day's rumours.

Jensen Huang resigns as Nvidia CEO

Nvidia's board announced that Huang has resigned as president, CEO and board member, effective immediately. CFO Colette Kress is interim CEO and keeps her finance role. Longtime director Harvey C. Jones becomes the board's first-ever chair, and the board will consider internal and external candidates.

Nvidia shares opened down more than 13% and closed down 4.3%, CNBC reports. The company has lost more than $740B of market value since last Friday. Its first earnings under Kress are due on November 17. Our breakdown of the change is here.

The exit also hangs over Nvidia's role in AI financing. In August, Nvidia signed preliminary agreements with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to raise over $500B for AI infrastructure. Those deals still need final agreements, CNBC notes.

Google releases Gemini 4 Hydrogen

Google released Gemini 4 Hydrogen, a fast model for coding, agents and everyday work. Google says it finishes tasks 2.5 times faster than Gemini 3.8 Flash and scores higher than Gemini 4 Argon on the Artificial Analysis Intelligence Index.

Hydrogen is rolling out in the Gemini app, Google Search, the Gemini API and Google AI Studio, mostly without a paid plan. The API costs $1.50 per million input tokens and $7.50 per million output tokens. Benchmarks and availability are here.

Agents get phones, wallets and servers

Meta's Muse can now make phone calls for paid US subscribers, to insurers, cable companies or restaurants. Alexandr Wang announced it; more in our article.

Replit Agent can now buy supplies through Stripe's Link wallet and run Google Ads for a business. Replit says the owner stays in control of every dollar; details are here.

Anthropic now lets companies run Claude Managed Agents entirely on their own infrastructure, by request through their account team. We covered the setup here. ElevenLabs also opened a marketplace of voice agents built by partners such as Salesforce, IBM and Twilio, as we explain here.

SpaceX plans $74B of AI data centers

SpaceX said it plans to invest more than $74B in two AI data center campuses. About $24B goes to Starbase in Texas and roughly $50B to Cochise County, Arizona. Both still need local approvals, and the Arizona site already faces opposition. Our breakdown is here.

Nonprofits sue to halt frontier AI

Four nonprofits, including Public Citizen and Heat Initiative, sued Anthropic, Google, Meta, OpenAI and SpaceX in San Francisco Superior Court. They claim the companies violate California's unfair competition law and ask for an emergency injunction on next-generation models.

The groups want work to resume only once the companies prove the technology is under human control, Al Jazeera reports. The article reports no response from the companies.

Smaller updates

Seven smaller stories came out on Thursday.

Cisco's quarterly revenue rose 24% to $18.5B, and it expects $9B of AI infrastructure revenue from hyperscalers this fiscal year, CNBC reports.
Naver signed a deal worth about $1.5B to sell AI computing to an unnamed US inference company, the first customer of its AI factory with Nvidia and Brookfield.
Agile Robots raised $500M at an $8.5B valuation, led by SoftBank with Nvidia and Hitachi, dpa reports.
Edge0 released Edge0-145B, an open model with about 10B active parameters that it says runs on a laptop with 29GB of peak memory.
Norwest closed a $4B fund, its largest, with a focus on physical AI, voice AI and agentic cybersecurity, TechCrunch reports.
Perplexity's Personal CFO, which connects bank accounts through Plaid, expanded to 17 European countries.
We also published a guide to prompting Claude Opus 5.5 and Fable 5.2, which is here.

Three of the seven are about money flowing into AI hardware and robots.

Rumours

Five more stories circulated on November 12, but we could not read the primary report or confirm them with the companies.

Underwriters reportedly expect Anthropic to price its IPO above the $70 to $78 range, which could lift its value past $1.5 trillion, Bloomberg reports.
ByteDance is reportedly preparing a Muse-like AI agent that could launch in the coming weeks, Bloomberg reports.
Anthropic's updated usage policy reportedly blocks Claude use from mainland China and China-controlled entities as of November 12, TokenPost reports. We could not find the change on Anthropic's pages.
Sam Altman reportedly told Samsung executives that AI demand "has been double-counted", according to meeting notes cited by Korea's Financial News.
An unverified analysis by Unemployment Labs claims six Chinese open models show signs of training on banned Blackwell GPUs.

Treat all five as unconfirmed until the companies announce them.

What the day was about

Thursday ended the Huang era at the company that sells most of the industry's chips. Nvidia now has an interim CEO, a first-ever board chair and earnings due in five days.

The rest of the industry kept building. Google shipped a faster Gemini for free users, agents gained phones and payment tools, and SpaceX committed $74B to new data centers. On November 12 Nvidia's leadership changed overnight, while demand for AI compute kept growing around it.