Nvidia's China Spy Crisis
· The Fluency Briefing
Welcome to this week's AI
This Week in AI
Hey there. Here's a number to start with: companies blame stalled AI projects on cost overruns three times as often as on compute shortages venturebeat.com, Nov 10. So the thing slowing you down probably isn't chips. It's the bill. Meanwhile, the man who sells most of those chips spent the week in a scandal that reads like a spy novel, and voters turned against data centers at the ballot box. Here's what mattered.

📰 The Big Story
First, Nvidia waved off a report about an alleged spy plot as routine customer meetings. Then Jensen Huang admitted to an affair therundown.ai, Nov 10. On Sunday, Nvidia's board launched an independent investigation after Huang told staff he'd been unfaithful to his wife techcrunch.com, Nov 9.
The spy angle makes this more than a personal scandal. According to the New York Times, US officials believed Alice Huang (no relation) had been recruited by China's top intelligence agency. Her current whereabouts are unknown bbc.co.uk, Nov 9. Nvidia underperformed the Nasdaq by roughly 7% on the news lesswrong.com, Nov 10. Kalshi bettors put his odds of surviving the week at a low techcrunch.com, Nov 12, while Yale's Jeffrey Sonnenfeld calls it "about a 50-50." Huang has already pulled out of the ElevenLabs Summit sifted.eu, Nov 10.
The timing makes it worse. Investigators found that two Chinese companies cleared to buy Nvidia's H200 chips list addresses that lead to empty offices tomshardware.com, Nov 8. That deepens what CNBC calls Nvidia's China problem cnbc.com, Nov 10.
Why this matters going forward: Axios is openly asking whether the AI boom can survive without Huang axios.com, Nov 9. A trillion-dollar industry turns out to rest on one person's judgment.

📋 5 Stories That Shaped the Week
Beyond the headlines, here's what shaped the week...
The builders kept building as if nothing happened. UK travel company Holiday Extras says OpenAI's Codex is changing what it can build. By the company's account, it rebuilt its app in about two weeks, down from six months openai.com, Nov 12. Let's be real: speed like that shrinks the moat for every agency billing you by the month.
Agents are also getting their own infrastructure. Mount Thor raised $45 million to build Mac data centers that rent machines to AI agents 9to5mac.com, Nov 12. Reason Machines open-sourced Aspen, a Git server (the system that stores and tracks code) designed for agents rather than humans marktechpost.com, Nov 9. Translation: agents are becoming customers in their own right, with their own landlords.
Models got cheaper and faster, too. Anthropic's Claude Fable 5.2 launched under a "pacing, not pausing" banner therundown.ai, Nov 10. Amazon Bedrock added Moonshot AI's Kimi K3.1 Pro Highspeed aws.amazon.com, Nov 11. Prices dropping means your next contract renewal is a negotiation, not a formality.
Speed isn't the same as judgment, though. Nine AI agents attempted open research questions from US national labs, and none of the 135 submissions met the experts' bar the-decoder.com, Nov 7. Meanwhile, a report that Anthropic is using AI to build better AI has researchers alarmed theverge.com, Nov 7.
The public is pushing back. Midterm voters delivered a stinging rebuke of data centers just as Anthropic prepares to go public cnbc.com, Nov 7. Tech giants are also pushing AI into schools while fighting child safety lawsuits cnbc.com, Nov 7.
🔗 The Pattern We Noticed
A week ago, we treated the AI stack's foundation (chips, capital, permission to build) as solid ground. We worried about agents misbehaving on top of it, not about the floor itself.
This week the floor shifted twice. Nvidia's governance, which every model depends on, turned out to hinge on one man's private life and a suspected intelligence operation bbc.co.uk, Nov 9. Then voters turned down data center measures cnbc.com, Nov 7, so permission to build is now contested at the ballot box.
Here's the updated read. The biggest risk to your AI plans now sits upstream of any agent: in a supplier's boardroom and a county's zoning vote. That means price and availability shocks can arrive from causes that have nothing to do with technology. For you, any tool you can't swap within a week is now a governance bet, not just a software choice.

🔮 On the Horizon
These stories are still unfolding — here's what to track:
- Nvidia: Jensen Huang is no longer CEO, or is on formal leave, by November 27, 2026. We'd know from an Nvidia press release or 8-K filing.
- Anthropic: The company publicly files its S-1 by November 30, 2026. We'd know from SEC EDGAR.
- US Commerce Department: It publicly suspends or reviews the H200 export approvals for the two empty-office Chinese buyers by December 4, 2026. We'd know from a Commerce statement or major press report.
📚 Term of the Week

Going deeper on one concept that shaped this week's AI conversation.
"Model Routing"
What it is: Model routing means automatically sending each AI request to whichever model fits it best. Cheap, fast models handle easy tasks, and expensive ones handle hard tasks. Think of a dispatcher assigning jobs to the right worker instead of sending the senior partner to fix the printer.
Why it matters this week: Fireworks argues you're wasting money unless you use many models theregister.com, Nov 8. Cost overruns are the top reason AI projects stall.
The bigger picture: Cheaper releases like Fable 5.2 and Kimi K3.1 make routing more valuable each month. It also cuts your dependence on any single vendor, which looks smarter after this week's supplier drama.
Try this: Run one routine task, like summarizing an email, through a cheap model and a premium one. Then decide whether the premium answer was worth the price.
📬 That's a Wrap
This was the week the AI boom learned its weakest link might be a person, not a processor. Your move: last week you checked your main AI vendor's price-change clause and named a backup tool. You probably found that they can change prices on short notice. Now run one real weekly task through that backup tool and note what breaks (20 minutes).
Fluently yours, The My AI Fluency Team
What We're Working On
✨ AI Can Do That? Free 3-Day Live Workshop — Join us and Dylan Stewart live, November 27–29 at noon Pacific, as we take real ideas from attendees and work through them with AI. No experience needed, and one registration covers all three days. → Reserve My Free Spot
✨ Free 30-Minute AI Consultation — Discover how My AI Fluency can help your business unlock the potential of AI. We'll discuss your goals, explore practical AI opportunities for your industry, and outline clear next steps. → Schedule Free Call
✨ How AI-Fluent Are You? — Test your AI fluency with our interactive quiz. See how you stack up and discover what to learn next. → Take the Quiz
📞 Book a Consultation | 🌐 Website
