Everything That Happened in AI Today (Friday, November 13, 2026)

Nvidia's founder is out, its AI chip empire has an interim boss, and the companies counting on it suddenly have a leadership question they never expected to ask.

Welcome to The Neuron's Friday Around the Horn, where one executive departure managed to set off a chain reaction across AI.

Nvidia named an interim CEO, and the industry's open-model supporters started asking what its next chapter could look like. Elsewhere, SpaceX announced plans for two enormous computing campuses, while SoftBank disclosed what happened after it bought far more power equipment than it needed.

On the user side, researchers found that AI shopping assistants sometimes send customers to risky storefronts. Maybe don't let a chatbot choose where you enter your credit card number just yet. We have the details below, plus a crowded table of launches. Let's get into it.

🆕 NEW From The Neuron

• Jensen Huang Resigns. Nvidia Will Probably Be Fine - Eventually. Our latest analysis examines Nvidia's leadership transition and the business questions facing its customers.
• The AI Spending Boom Has a Problem: Most Companies Spend Very Little Corporate AI spending is growing, but Ramp's data suggests that a relatively small set of companies accounts for much of it.
• Fireworks AI Was Hacked. Here's How Criminals Make Money From Stolen AI Access Our explainer covers the earlier Fireworks breach and the market for stolen access to AI models.

Around the Horn: Friday, November 13, 2026

NVIDIA's longtime chief executive, Jensen Huang, stepped down amid reports about extramarital relationships and allegations involving a woman suspected by U.S. intelligence of passing information to China. Nvidia appointed its CFO, Colette Kress, interim CEO. The allegations remain allegations, and Nvidia has said it has found no evidence that corporate or national-security secrets were compromised.

Huang was more than Nvidia's public face. He was a major advocate for the company's investments across AI companies and infrastructure. A Semafor analysis of the end of the 'Nvidia put' argued that some AI companies may no longer assume Nvidia will keep backing the market on the same terms. Another Semafor report examined the potential consequences for Chinese AI developers, whose chip access already depends on U.S. export restrictions.

One immediate question concerns open models. Hugging Face CEO Clement Delangue pledged to continue defending open-source AI after Huang's exit. According to TechCrunch, Nvidia's planned $12.9 billion acquisition of Hugging Face is still expected to close in 2027. Nvidia says its product roadmap and customer commitments will continue. Whether the next CEO maintains Huang's dealmaking priorities remains to be seen.

🏆 TOP 5 NEWS (Around the Horn)

• SpaceX announced plans to invest more than $74 billion in AI supercomputing campuses in Texas and Arizona. The Arizona site alone would cost about $50 billion and still needs county approval; the Texas facility is designed to generate its own electricity.
• A coalition of nonprofits sued OpenAI, Anthropic, Google, Meta, and SpaceX, seeking a court order to halt development of models more capable than the companies' existing systems. The lawsuit cites the labs' own warnings about extreme AI risks.
• Notion confirmed that hackers accessed some customer chat transcripts through the breach at AI provider Fireworks. Fewer than 1% of Notion workspaces were affected, and most of the exposed chats could not be attributed to a specific user or workspace.
• An investigation found AI shopping assistants recommended risky retail websites. In brand-protection firm Podqi's tests, questionable sites appeared in 15% of Perplexity responses, 8% of ChatGPT responses, and 4% of Gemini responses. These are results from one firm's test, not universal error rates.
• SoftBank took a $3.7 billion charge after revising orders for AI data-center power equipment. The loss followed nonrefundable payments and contract changes, a costly example of the uncertainty around projected AI electricity demand.

Honorable Mentions

• A report claimed six Chinese AI models carry numerical traces of Nvidia Blackwell chips. The researcher says the patterns indicate which hardware generated starting weights, but they do not establish that export rules were violated. The supporting code had not been published when the report appeared.
• 40% of surveyed game developers now use generative AI at work, up from 36% last year, according to the GDC Festival of Gaming. Research and brainstorming were more common uses than final game artwork.
• The BBC investigated 'housefishing', in which AI-enhanced real-estate photos show rooms or features that differ from the property buyers eventually visit. Consumer authorities say undisclosed changes to important property features can be misleading.
• A study of human-AI creative work found that workflows where people simply approved AI-made decisions produced lower writing quality and satisfaction than other collaboration methods, even though completion times were similar.

🍪 TOP TREATS TO TRY

• Google released Gemini 4 Hydrogen, a faster, lower-cost model for coding and AI agents. It is available through Google AI Studio, the Gemini API, Search, and other Google products.
• Anthropic opened self-hosted Claude Managed Agents in private beta, allowing companies to run AI agents and store session history on their own servers while choosing where Claude requests are sent.
• Google and Samsung opened preorders for Gemini smart glasses, with styles from Warby Parker and Gentle Monster starting at $395 and shipping December 10. In The Verge's hands-on review, live translation and controls intended to protect bystanders stood out as alternatives to Meta's glasses.
• OpenAI introduced Daybreak Red Team and Defense Factory, AI security tools that test applications for weaknesses and propose fixes. Red Team is in early access; engineers review Defense Factory's suggested patches before deployment.
• Apple's new Apple TV 4K adds Siri AI and hands-free voice commands. The $299 device arrives November 13, although an early review found follow-up questions were still unreliable. Older Apple TV 4K models are expected to receive Siri AI in a future software update.
• World Labs opened Atlas to everyone, with apps and developer interfaces for reconstructing 3D scenes from images and producing videos or 3D scene assets for creative workflows.
• Krea introduced Krea Bots, AI collaborators designed to carry creative projects forward and adapt to a team's preferences. The company advertised beta access and $1,000 in credits.
• ElevenLabs launched its Marketplace, a directory of voice agents and apps built with ElevenLabs for sectors such as healthcare, financial services, and retail.
• Fish Audio introduced Fish Agent, a no-code voice-agent builder for incoming and outgoing calls. Its Live 3 model can listen while speaking and respond to interruptions; the company says it can deploy a basic agent in about 30 seconds.
• Thinking Machines made Inkling 2 Small's model weights downloadable on Hugging Face under the Apache 2.0 license. Developers can now modify and run the model, but its largest checkpoints require large GPU memory.
• IBM released LLMON preview variants of Granite 4.1, trained to distinguish instructions from outside data through dedicated markup. IBM reports better resistance to malicious instructions hidden in documents; the 3B and 8B previews are experimental.
• J.Crew launched a Claude-powered Style Advisor for U.S. shoppers. Customers describe an occasion and receive coordinated outfits assembled from the brand's own catalog styling, with purchase-history personalization planned for later.
• Nectar introduced Creator Agent, a brand tool that discovers influencers, contacts them, and manages campaigns. Nectar says it can search more than 300 million creator profiles.
• Millionways unveiled Thorsten-4, a model trained with psychologist-labeled interviews to infer motivations from text. The startup also disclosed $6 million in funding, although its claims about interpreting intent still need independent evidence.

🔬 AI Research, Models & Developer Tools

• Ricursive Intelligence scheduled an open-model hackathon for Friday, November 13, at its Palo Alto office, with builders invited to experiment with its Ripple coding model.

💰 Fundraising & Deals Roundup

• Norwest closed a $4 billion fund, its largest ever, to invest across technology, healthcare, and consumer companies, with AI among its priorities.
• Anthropic shared its proposed public offering. The company previously filed its registration statement and announced an expected price range of $70 to $78 per share, but the offering had not become effective at the time of its November 10 roadshow announcement.

Previous Around the Horn Digests

Catch up on the earlier issues:

• Thursday, November 12: United restricted AI booking agents; Databricks introduced a business assistant.
• Wednesday, November 11: Anthropic's IPO roadshow and an OpenAI quarterly-revenue report.
• Tuesday, November 10: Anthropic's model updates and early Fireworks breach reporting.
• Monday, November 9: AI power deals and new research-agent concerns.
• Friday, November 6: GPT-6 Cyber and financial accounts in Claude.
• Thursday, November 5: AI-found software vulnerabilities and research on mysterious cell particles.
• Wednesday, November 4: OpenAI security scrutiny and Google school AI policy.

That's a Wrap

We covered 32 unique source links today. If you made it this far, you're probably better informed about Nvidia's succession plan than anyone expected by Friday morning.

For the short daily version, subscribe to The Neuron. See you tomorrow.

P.S. Know someone who follows AI news? Send them this digest and invite them to subscribe.