Ricardo

Nvidia's board just chose $13.6 billion over national security.

Every director judging Jensen Huang's affair with a suspected Chinese spy gets poorer if they find something serious.

In the same statement that opened the investigation, the board said it has confidence in Jensen and his leadership team.

So I went through Nvidia's own proxy filing to see who's doing the judging.

Here's who's sitting on that board:

- Mark Stevens owns about 34.1 million Nvidia shares, worth roughly $6.2 billion.
- Tench Coxe owns about 30.6 million shares, worth roughly $5.5 billion. He's been on the board since 1993, the year Jensen co-founded Nvidia.
- Harvey Jones owns about 7 million shares, worth roughly $1.3 billion. He's also been there since 1993.

At Thursday's close of $181.60, I count about $13.6 billion of Nvidia stock held by the 9 independent directors. Jensen himself owns about 871 million shares, or 3.58% of the company, worth roughly $158 billion.

So picture the choice in front of them:

If they find something serious and push Jensen out, Nvidia's stock probably takes a hit, and so does their own net worth.

If they find nothing, everything stays exactly the way it is.

And the board isn't the only referee with money on the line:

The President called Jensen an honorable guy who probably has an excuse, and said he hopes Jensen keeps running Nvidia. Washington's own plan to let Nvidia sell H200 chips in China gives the US government a 25% cut of those sales. That plan is now reportedly frozen.

So almost every referee in this story has a reason to hope there's nothing to find.

The one exception might be Congress, which wants Jensen at a closed-door briefing on November 18. That's the day after Nvidia reports earnings.

Jensen Huang reached the final stage of wealth: The Chinese Spy.

He's way too influential to face any serious consequences, and this immunity goes beyond just an affair...

AI was the perfect excuse to fire people, and companies just stopped using it.

In May, US employers blamed AI for 38,579 job cuts.

In October, they blamed it for 1,265.

That's a drop of about 97% in 5 months.

But companies still announced 43,994 job cuts in October, according to Challenger, Gray & Christmas. AI was the reason given for fewer than 3% of them.

I think that tells us what AI was really doing in all those layoff announcements...

It let executives present a cost cut as a bet on the future.

And Meta is the best example I've seen:

In January 2025, Mark Zuckerberg told Joe Rogan that Meta would have an AI that could work like a mid-level engineer that year.

Then Meta drew up a plan called Project OT that considered shrinking some teams by as much as 60% and handing their work to AI agents, according to Reuters. On May 20, it cut about 8,000 jobs, roughly 10% of its staff. The night before, Zuckerberg cancelled a second wave that was planned for November.

So what happened when the agents took over more of the work?

Here's what Meta's own internal numbers showed:

- Code changes to internal systems rose 220% from a year earlier.
- Changes that turned into new or better features for users rose just 36%.
- Major technical and security incidents rose 40%.
- Time spent fixing those incidents rose 70%.

One internal post warned that the agents were taking huge, disruptive actions that a human would almost never take. In July, Zuckerberg told staff at a town hall that agent-driven work hadn't sped things up the way Meta expected.

And this was a company with internal leaderboards, some of them unofficial, ranking employees by how many AI tokens they used.

So I'd say Meta measured how much AI its people used, and the AI produced a mountain of activity and a lot more cleanup. Now The Guardian reports Meta is rehiring managers.

Meta isn't the only one walking it back though:

- In a Robert Half survey of 2,000 US hiring managers, 32% said their companies brought back the exact roles they had cut because of AI.
- An Atlanta Fed paper surveying nearly 750 executives found little evidence of near-term job losses across the economy due to AI.

Meanwhile, 121,401 job cuts this year were announced with AI as the reason. That's about 1 in 5 of all the cuts Challenger has counted in 2026.

Those people paid for the AI story, whether or not the AI could do their jobs.

Companies should report revenue per employee 12 months later. If it didn't go up, explain what the AI did instead.

How many of this year's AI layoffs were ever really about AI?

Amazon's profits are about to be held hostage by Anthropic's stock price.

Millions of retirement accounts own Amazon through index funds. And this year, a huge chunk of the profit those funds are paying for came from one private company getting more expensive on paper.

That company is Anthropic. And its $1.5 trillion IPO price will likely have to come down, maybe closer to $1 trillion.

Amazon depends more on that number than almost anyone.

In the second quarter, Amazon made $27.5 billion of operating income.

Its net income was $62.6 billion.

Most of that gap was Anthropic. After Anthropic's May round valued it at $965 billion, Amazon marked up its Anthropic shares by about $50.5 billion. The first quarter had already added about $16.8 billion of Anthropic gains.

That's roughly $67 billion of pre-tax gains in 6 months from revaluing one stake.

On June 30, Amazon carried the whole position at about $190 billion. Its first $8 billion went in between 2023 and 2024.

And the money runs in a circle:

In April, Amazon put another $5 billion into Anthropic, with up to $20 billion more tied to milestones. In the same deal, Anthropic agreed to spend more than $100 billion on AWS over the next 10 years.

So the loop goes: Amazon invests in Anthropic -> Anthropic commits to buying Amazon's cloud -> Anthropic raises money at a higher price -> Amazon books the higher price as profit

None of this is illegal. The accounting rules tell Amazon to update the value when a new funding round sets a new price.

But the third quarter shows what Amazon looks like with no new round to mark: Net income fell to $20.7 billion. Free cash flow over the last 12 months came in at minus $42.4 billion, with $77.3 billion of capex in the quarter alone.

And the IPO changes the game completely.

Once Anthropic trades publicly, Amazon's stake gets valued off a stock price that moves every single day. That price could start moving Amazon's reported profit far more directly.

Investing .com ran an illustration: If Anthropic trades at $750 billion, Amazon's stake would be worth about $42 billion less than its June value.

That's about TWICE what Amazon earned in the entire third quarter.

To be clear, a $1 trillion price still sits above the $965 billion Amazon marked to in May. So the real risk starts after the first trade.

Alphabet is in a similar spot. It carried $124.3 billion of private company stakes on June 30, and reports say most of that is Anthropic.

If Anthropic falls it takes the big guys with it.

This little auction matters more than everything the Nasdaq 100 lost the same day.

Stocks trade every second, so everybody knows what Nvidia's worth at any moment.

Private data center loans almost never get priced in public like this. And for 2 years, the whole AI boom has been funded on the assumption that this paper's safe.

In 13 days, real buyers have to name a real price.

ING is seeking offers on a dozen loans through a bids wanted in competition process, with bids due November 19 and a $10 million minimum.

The pile includes debt tied to Vantage Data Centers and CyrusOne.

And it includes a $25 million piece tied to the New Mexico site in Stargate, the venture backed by OpenAI, SoftBank and Oracle.

Bloomberg calls this a rare auction of only data center debt. No price has been reported, and ING declined to comment.

$25 million is pocket change on a campus reported at $165 billion, and I won't pretend otherwise.

But just look at which campus it is:

Oracle sent a force majeure notice to the developer of that site in September. That's the legal notice saying events beyond a company's control may excuse it from its obligations. Oracle says the schedule hasn't changed, and it calls these notices commonplace on projects this size.

New Mexico regulators have until November 23 to decide the air permit for the site's power plant.

That's 4 days AFTER the bids are due.

The FT has already reported the project's roughly $18 billion construction loan trading at stressed levels.

So anybody bidding on that slice is putting a price on a permit nobody has yet.

And ING isn't alone:

Startup Fortune, citing Bloomberg, says JPMorgan, Morgan Stanley, SMBC and MUFG have been shopping pieces of AI debt they helped underwrite.

Andrzej Skiba of RBC's BlueBay unit told Bloomberg he's seeing a "massive hedge fund capitulation" in the space.

CoreWeave's 9.625% notes due 2032 last traded around 76 cents on the dollar, for a yield of about 16%.

The cost of insuring CoreWeave's debt was reported around 1,164 basis points this week, close to its October record.

SoftBank's shares fell more than 9% in Tokyo on Friday, and its default insurance hit a 4-year high.

Meanwhile Apple went out for about $7 billion of bonds this week and drew about $31 billion of orders.

That's the tell for me, because lenders still have plenty of money. They're just getting pickier about AI paper.

And the AI trade has needed the opposite. The money behind these campuses has to keep rolling from one lender to the next, and that only works if someone's always willing to take the paper at a good price.

If the next buyer wants a big discount, the bank thinks twice about the next loan. Then the next campus costs more to finance, and eventually somebody's forecast for 2028 falls apart.

Now, a smart person would say I'm making a mountain out of a molehill.

$500 million is a rounding error in a market this big. Banks sell loans every day to manage their books. ING may have decided this months ago for reasons that have nothing to do with AI. Investment grade credit held up fine on Thursday, with LQD up 0.15%. And Oracle reportedly can't terminate the New Mexico lease, which should comfort the lenders.

If ING gets close to full price, I'm wrong and the panic was overdone.

But I think that's the point of watching this, and I'd bet the result leaks.

CoreWeave reports November 10. Nvidia reports November 17. Then the ING bids land on November 19.

I'd pay more attention to the third one.

So who wins this one: ING getting close to full price for those Stargate loans, or a buyer demanding a serious discount?

Jensen Huang says engineers should burn $250,000 of AI a year, and Apple just cut its engineers' AI budget to about $50,000.

And Apple plans to cut it again.

Last month, some Apple engineers saw their daily AI spending cap drop from $300 to $200, Bloomberg reported. Some engineering teams could be cut to $100 a day next. Many groups outside engineering already sit at $100, and Apple has considered going as low as $25 for some of them.

In March, Jensen said on the All-In podcast that he'd be deeply alarmed if a $500,000 engineer used less than $250,000 of AI tokens in a year. He also said Dario Amodei's forecast of hundreds of billions in AI revenue was too conservative.

So Jensen's rule is simple: AI spending should run about half of what an engineer gets paid.

By my math, at about 250 workdays a year, $200 a day comes to roughly $50,000. The planned $100 cap comes to roughly $25,000.

By Jensen's rule, a $100 daily cap only makes sense for an engineer paid about $50,000 a year.

And Apple still wants its people using MORE AI. It's steering teams toward cheaper models and AI that runs on the device, and Bloomberg's Mark Gurman has said Apple's internal work runs heavily on Anthropic's models.

I think Apple just showed where the real limit on the AI boom sits now: The finance department.

And that's just ONE company:

- Accenture capped Claude at $29 a month per employee on October 1, after some staff had been spending up to $1,000 a month, 404 Media reported.
- Uber burned through its entire 2026 AI budget in about 4 months, then capped AI coding tools at $1,500 a month per tool.
- Microsoft has cut allowances too, and its AI chief Mustafa Suleyman told Bloomberg that Anthropic is extremely expensive and that Microsoft wants to eventually get rid of that cost.
- Meta had internal leaderboards ranking staff by how many tokens they used, and now it has introduced token budgets.

The usage is clearly there. Uber's engineers used the tools so much they spent a full year's budget in 4 months.

So the people saying stop are the ones paying the bill.

That's a huge problem for the AI labs, for one reason: Their spending is already locked in.

Anthropic has committed to $895 billion, mostly for computing power. OpenAI has signed more than $1.4 trillion of compute deals. And paying for all of it depends on companies spending more on AI per worker every year.

Apple, Accenture, Uber and Microsoft all pay for Anthropic's models, and Anthropic could start its IPO roadshow as early as next week. Its own filing shows revenue grew about 47% from the second quarter to the third, down from about 144% the quarter before, and price cuts took about $600 million out of revenue.

Plenty of people will say this is just housekeeping, and they have numbers:

Bloomberg says few Apple engineers ever used the full $300, so the savings may be small. Cheaper models can do the same work for less. Apple's R&D still rose 32% to $45.8 billion in fiscal 2026, partly because of AI. And more companies start using AI every month, so total spending can keep growing even with a cap on each worker.

But Jensen's pitch depends on spending per worker going UP. Apple, Microsoft, Meta, Uber and Accenture are all writing rules that push it down.

So the first question for Anthropic's roadshow should be simple: How much are its biggest customers spending per employee compared with a year ago, and how many of them have put caps in place?

Who's right about what an engineer's AI is worth, Jensen Huang or Apple's finance team?