Nvidia Paid for the AI Boom. Who Pays Now? (Nov 13, 2026)

🎧 Listen ad-free:


🌐 Daily Briefing: https://djamgamind.com/daily
📚 Research Hub: https://djamgamind.com/pdfs

🔍 Keywords: Nvidia, Jensen Huang, Colette Kress, AI Financing, Export Controls, OpenAI, ChatGPT Wallet, Agentic Commerce, GPT-6 Cyber, Linux Security, Microsoft 365 Outage, Gemini 4 Hydrogen, AI Costs, AI Agents

Today’s question: Nvidia paid for a large part of the AI boom. Who pays now?

Summary:
The day after Jensen Huang resigned, the question is money. Nvidia has made $530 billion of commitments and guarantees that hold up data centers, startups and open models, according to SemiAnalysis, and Axios reports the industry “remains highly dependent on its checkbook.” OpenAI rolled out ChatGPT Wallet, so its agents can pay for things with your approval. OpenAI’s GPT-6 Cyber found and patched 948 Linux kernel flaws at about $1,280 each. A Microsoft 365 outage hit Outlook, Teams and Copilot worldwide. And Google released Gemini 4 Hydrogen, where you choose how much reasoning to pay for.

Important Topics:

  • The checkbook: what Nvidia finances, and what changes without Huang.

  • The wallet: OpenAI’s agents can now pay; United says not on its site.

  • The repair bill: 948 Linux flaws at $1,280 each, and a worldwide Microsoft 365 outage.

  • The price list: Gemini 4 Hydrogen and paying by reasoning level.

  • Today’s corners: Cert Corner, Red Flag or Green Light?, and the Boardroom Call.


💵 The checkbook: Nvidia after Huang

Axios AI+ · The Rundown AI · Techpresso · Nvidia press release

  • Where things stand: Huang resigned Thursday. Colette Kress, CFO since 2013, is interim CEO and holds both roles while the board searches. Nvidia closed down 4% Thursday after falling more than 14% earlier, and reports earnings November 17 (The Rundown AI).

  • Inside the company: a suggestion-box post from a longtime engineer drew over 1,600 comments, many urging the board to reinstate Huang, according to internal messages seen by WIRED (Techpresso).

  • The checkbook: Nvidia “has made $530 billion of off-balance-sheet commitments and guarantees,” including deals to buy memory and backstops for data centers, according to SemiAnalysis (Axios AI+).

  • Why that matters more than chips: the industry “remains highly dependent on its checkbook,” Axios writes. It expects Nvidia to become less aggressive about startup funding, acquisitions and backstops (Axios AI+).

  • The alternatives are real: custom chips from Google, Amazon, Microsoft and OpenAI, plus AMD. Anthropic says its Mythos model helped run certain workloads on AMD chips in about three days, compared with six to 12 months in the past (Axios AI+).

  • Open models: Nvidia has a pending acquisition of Hugging Face and builds the Nemotron open models. Without it, Axios says, it’s not clear who funds open-weight alternatives (Axios AI+).

  • Export controls: researcher Will DePue says patterns in published model weights show four Chinese labs trained on restricted Blackwell chips, about 12,000 GPUs in the runs he examined. The data cannot show where the chips sit or prove smuggling (Techpresso, The Rundown AI).

Why it matters: Axios’ bottom line: Huang’s departure “leaves a hole in the ecosystem that will be harder to fill than any shortage of Nvidia chips.” If any part of your AI plan depends on a vendor’s credits, guarantees or prepaid capacity, find out who is behind that money.

🤝 PARTNER SLOT
This space is reserved for one enterprise AI partner each day: AI security, governance and agent controls for the leaders who read this briefing. Sponsorship options and media kit →

💳 The wallet: agents can now pay

Techpresso · The Next Web · TBPN

  • ChatGPT Wallet: OpenAI’s “dots” agents can now buy things for users, after each purchase gets an approval and a security check. It was built with American Express, Mastercard and Visa (Techpresso).

  • The design: the agent never sees your real card. It passes a per-transaction “payment token,” and card rewards and protections stay in place (Techpresso).

  • The limits: it starts with US Pro subscribers aged 18 and over, and each purchase is approved with Face ID or a fingerprint (The Next Web).

  • The business model: OpenAI won’t take a cut of purchases at first. “Our goal here is to really focus on merchants and our network partners,” said OpenAI’s Chaya Nayak (The Next Web).

  • The card networks: “It just makes sense to watch it, start it in a contained environment, and then keep rolling from there,” said American Express’s Luke Gebb (The Next Web).

  • The other side: United Airlines’ revised Contract of Carriage lets it cancel bookings made by unauthorized AI agents, void tickets and erase MileagePlus miles. Bookings made before November 11 won’t be canceled (Techpresso).

Why it matters: The payment networks have now built the controls for agent spending: approval per purchase, a token per transaction, a small start. Apply the same three to your own agents before you give them a card.

🔧 The repair bill: bugs and outages

The Batch · OpenAI · Techpresso · Generative AI

  • GPT-6 Cyber: in a mid-October scan of the Linux kernel, OpenAI’s security model addressed 948 flaws that let ordinary users take full control of a machine, and wrote a patch for each (The Batch).

  • The price of a bug: the run cost around $1,280 per vulnerability. In July, the same setup on GPT-5.6 Sol addressed 166 for around $3,100 each (The Batch).

  • Who gets it: selected customers of OpenAI’s Daybreak Red security program, who must use single sign-on, multifactor authentication and managed devices (The Batch).

  • Microsoft 365: an outage hit Outlook, Teams and Copilot starting Thursday. Downdetector tracked complaints across 74 countries, with Outlook reports peaking near 138,000 (Techpresso).

  • The cause: Microsoft blamed an overwhelmed process that handles secure server connections, “not a cyberattack or traffic spike,” in Techpresso’s words. The outage also broke OpenAI’s and Anthropic’s Microsoft connectors (Techpresso).

Why it matters: Finding a serious flaw now costs about as much as a laptop, and that price will fall for attackers too. Shorten your patch cycle, and list which of your AI tools stop working when one provider goes down.

🏷️ The price list: pay by reasoning level

Techpresso · The Rundown AI · Generative AI

  • Gemini 4 Hydrogen: Google’s new model for coding assistants and agents costs $1.50 per million input tokens and $7.50 per million output tokens, with a 95% cache discount (Techpresso).

  • The dial: adjustable reasoning levels “trade higher benchmark scores for longer task times and bigger charges,” Techpresso writes. At high reasoning it scores 57% on Terminal-Bench 4.1, against 18.2% for Gemini 3.8 Flash.

  • The reach: Google put Hydrogen straight into Search for everyone (Generative AI). At its highest thinking level it outperforms Gemini 4 Argon (The Rundown AI).

  • Decision models: Google’s Gemini Decision 1 answers only yes-no, multiple choice and rating questions, at $0.10 per million input tokens, nearly five times cheaper than Gemini 3.5 Flash-Lite (Techpresso).

Why it matters: The reasoning level is now a line on your bill. Set it per task type, and track the cost of each completed task, not the cost per token.


🎯 Today’s corners

🎓 Cert Corner: Your team’s coding agent runs every task at the model’s highest reasoning level. Costs are climbing and simple tasks are slower. What’s the best fix? A) Switch every task to the lowest reasoning level

B) Set the reasoning level per task type, and measure pass rate and cost per task

C) Raise the output token limit

D) Turn off caching

🚩 Red Flag or Green Light? To save time, a manager turns on auto-approve for every agent purchase under $500, and all the team’s agents share one corporate card.

🚩 or ✅?

🏛️ The Boardroom Call (cost): Your biggest AI supplier also finances part of your AI plan: credits, prepaid capacity, a guarantee behind your data center lease. Its CEO has just resigned. What do you do with next year’s AI budget?

A) Nothing; contracts are contracts

B) Re-price the plan as if the financing ends at renewal, and qualify a second supplier

C) Sign a longer deal now, before the terms change

D) Pause AI spending until a permanent CEO is named

Answers below. Tell us yours in the comments.

Strategic C-Suite Takeaways

  1. List what your vendors finance for you: credits, guarantees, prepaid capacity. Price next year without them.

  2. Qualify a second chip or cloud path. Switching has become faster.

  3. Give each agent its own spending identity, limit and approval rule.

  4. Shorten your patch cycle. Finding flaws is getting cheap for everyone.

  5. Set reasoning levels per task, and track cost per completed task.

What Else Happened in AI on November 13, 2026?

  • Google’s Gemini glasses are up for pre-order from $395, made with Warby Parker, Gentle Monster and Samsung, and reach stores December 10. Generative AI

  • Four nonprofits asked a San Francisco court to stop Anthropic, Google, Meta, OpenAI and SpaceXSI from developing or deploying new AI models; the companies haven’t commented yet. Generative AI

  • Andrew Ng pointed readers to the new open source OpenWorker Agent Harness SDK, and encouraged engineers to build an agent harness themselves. The Batch

  • China’s AI firms are renting Nvidia’s best chips in data centers in countries like Malaysia and Thailand; the controls limit where a chip ships, not who logs in. Techpresso

  • GPT-6 Cyber’s pipeline first flagged 2,308 candidate flaws, confirmed 1,724, and produced 1,169 working exploits before the 948 patches. The Batch


✅ Our take on today’s corners

🎓 Cert Corner: B. Reasoning level is a cost setting. Gemini 4 Hydrogen’s levels trade higher scores for longer task times and bigger charges, so match the level to the task and measure the result. The lowest level everywhere (A) fails the hard tasks, and turning off caching (D) raises the bill. Practice thousands more in DjamgaTech.

🚩 Red Flag. One shared card means you can’t tell which agent bought what, and auto-approve removes the check that ChatGPT Wallet keeps on every purchase. Give each agent its own token and limit, and keep a human approval. 📘 Securing AI systems is a full domain of the AWS AI Practitioner exam: our 2026 guide.

🏛️ B. Supply won’t stop, but the financing may get more cautious, so budget as if it ends at renewal and have a second supplier tested. A ignores the change, C locks you in at the moment of most uncertainty, and D stops work that doesn’t depend on the supplier. Your org does it differently? Tell us. 📘 AI strategy and business value are core topics of Google’s Gen AI Leader exam: our 2026 guide.

🎓 The Cert Corner is open for one sponsor. Training and certification platforms: djamgamind.com/partners


🤝 Advertise with AI Unraveled & DjamgaMind

Put your enterprise AI brand in front of the CTOs, CISOs and engineering leaders who make AI buying decisions, daily, in English and French, on podcast apps, YouTube and Substack. Start by sponsoring one daily corner for four weeks.
Sponsorship options & media kit → · Email: etienne_noumen@djamgamind.com

🔗 Tools from the show

🎧 Prefer it commercial-free? Subscribe on Apple Podcasts.

⚗️ PRODUCTION NOTE: We Practice What We Preach.
AI Unraveled is researched and produced by Etienne Noumen, P.Eng. (Calgary, Canada) using a hybrid “Human-in-the-Loop” workflow.



Read original on Substack

Related articles

Level up your Tech Career with DjamgaTech PRO

Download the DjamgaTech PRO App on the Apple App Store to access hundreds of interactive prep questions, flashcards, and quizzes cleanly without ads.

Djamgatech PRO App
Download DjamgaTech PRO on the App Store 🚀

← All articles