Dario wants to pace AI. Jensen says read the fine print. Karp says call Washington. Zuckerberg says hurry up. Meanwhile, the data centers keep getting built. 😂

Something strange happened in AI.

Dario Amodei, CEO of Anthropic, published “We Must Pace the Frontier,” arguing that AI capability is advancing so quickly—partly because AI is increasingly helping build better AI—that safety work may not be keeping up.

Sam Altman agreed. Elon Musk replied: “Dario is right.”

Then Jensen Huang of Nvidia basically walked into the room carrying a fire extinguisher and said: don't let “doomsday narratives” become an excuse to escape existing laws.

Alex Karp of Palantir then asked the nastiest question of all:

Who pays if frontier AI creates enormous liability?

Suddenly, this isn't just an AI story.

It's a capital-allocation story.

The first mistake: believing “pace” means “stop”

Amodei isn't calling for AI to be abandoned. His proposal is to slow the rate of capability advancement enough to allow safety work and independent evaluation to catch up. Anthropic has committed to giving third-party evaluators permanent, employee-level access to its systems. Altman said OpenAI would do the same.

That's important.

Because investors heard:

“AI slowdown.”

What Amodei actually said was closer to:

“AI continues moving very fast—but let's build brakes before we need them.”

And there is now evidence that this isn't entirely theoretical. Reuters reported that OpenAI shelved a planned model release after internal safety testing identified concerning behavior. Yet days later OpenAI launched new “always-on” autonomous agents.

Translation?

The race didn't stop. The race got more complicated.

🥊 Six CEOs. Six incentives. One very expensive machine.

This is where the story gets interesting.

Don't ask who is telling the truth. Ask what each person has to gain if their version wins.

That's not cynicism.

That's investing.

🛡️ Jensen's argument deserves more credit than the headlines give it

Huang's criticism is not simply:

He explicitly says companies need to devote researchers and compute to safety and that unsafe products should face existing laws. His objection is to catastrophic claims being used to justify relieving companies from laws already on the books.

That creates a fascinating investment question:

Could “AI safety” become an expensive new moat for incumbents?

Independent evaluators.

Audits.

Security.

Testing.

Compliance.

Monitoring.

Documentation.

A giant AI company can afford this.

A tiny startup?

Maybe not.

So regulation could make AI safer and make the industry more concentrated.

Two things can be true at once.

💣 Then Karp drops the liability grenade

Karp's argument is even more provocative.

When asked what an AI company's massive potential liability would look like in an S-1—the U.S. registration document for an Initial Public Offering (IPO)—he replied:

“You're assuming that there will be an S-1.”

He then argued that if liability became effectively unlimited, government involvement could become the only practical solution, even floating nationalization as the extreme outcome.

Important distinction:

Karp did not reveal a secret OpenAI nationalization plan.

He presented a scenario.

But the question underneath it is excellent:

That's a question investors should care about.

🧱 Here's the part I care about most

Forget the speeches for a moment.

Look at the concrete.

Data centers take years.

Power infrastructure takes years.

Networking infrastructure takes years.

Chip supply agreements aren't cancelled because someone posts a thoughtful essay on Saturday.

That's why I prefer one simple test:

Does this headline change the capital budget?

“Let's be safer.”

❌ Interesting.

“Let's use independent evaluators.”

⚠️ More safety spending.

“Let's delay a model.”

⚠️ Potentially important.

“We're cancelling data centers.”

🚨 Now we're talking.

“Hyperscaler CapEx is being cut.”

🚨 Pay attention.

“GPU orders are being cancelled.”

🚨 The thesis has changed.

That's the difference between noise and signal.

🤖 And Zuckerberg may be telling us where the race is actually going

While everyone argues about whether frontier models are moving too fast, Meta is pushing AI agents and consumer hardware.

OpenAI is doing the same.

The race is shifting from:

Who has the smartest chatbot?

to:

Who can make AI actually do things?

That means the next bottleneck may not be intelligence.

It may be:

autonomy → reliability → security → trust → economics.

And that creates new investment opportunities.

Cybersecurity.

Identity.

Monitoring.

Networking.

Power.

Cooling.

Data centers.

Enterprise software.

AI evaluation.

The AI economy is becoming much bigger than the model itself.

🎯 Your AI Investor Checklist

Before buying or selling an AI-linked stock, ask:

☐ Did hyperscaler capital expenditure change?

☐ Are GPU or networking orders being delayed?

☐ Are data-center projects being cancelled—or merely delayed?

☐ Is AI inference demand increasing?

☐ Are enterprises actually paying for AI?

☐ Are AI agents becoming useful enough to justify deployment?

☐ Is safety spending creating new recurring businesses?

☐ Does the company generate free cash flow—or just promise future greatness?

☐ Would I still own the stock if the words “AI” disappeared from the investor presentation?

☐ What observable event would prove my thesis wrong?

That last question is the killer.

If you don't know what would make you wrong, you're not investing.

You're joining a fan club. 😂

🟢 How I would position the themes

🟢 Higher conviction

AI infrastructure, networking, power, cybersecurity, inference and enterprise adoption.

🟡 Selective

Frontier-model companies and highly valued AI software businesses.

🔴 Dangerous

Companies whose entire valuation depends on a future AI breakthrough, endless funding or government protection.

And I would not automatically buy Nvidia, AMD, TSMC or every data-center name simply because AI remains strong.

Great business ≠ great price.

The second half of the equation is always:

What am I paying?

🚨 My “AI Story Actually Changed” Alerts

GREEN

CapEx rising + AI revenue rising + utilization rising.

Stay constructive.

YELLOW

CapEx rising but ROI questionable.

Become selective.

RED

Hyperscalers cut CapEx + chip orders weaken + data-center cancellations appear.

Now reassess the thesis.

And one more:

BLACK SWAN

Government creates enormous liability protection or ownership structures for frontier AI.

Don't blindly celebrate.

That could protect the companies.

It could also socialize the downside while privatizing the upside.

That's a very different investment proposition.

The real lesson

The AI debate has become too binary.

“AI is going to save humanity.”

“AI is going to destroy humanity.”

“AI is slowing down.”

“AI is accelerating.”

Meanwhile, companies are doing something much less exciting:

spending money.

And money leaves footprints.

So when the next billionaire tells you AI is slowing down, speeding up, saving humanity or destroying it before lunch, don't immediately trade.

Ask:

Because speeches are cheap.

Chips aren't.

Data centers aren't.

Electricity isn't.

And eventually, neither is liability.

💌 Why Wealth Builder matters

The biggest pain point isn't a lack of information. It's too much information with no decision framework. One CEO says “slow down,” another says “speed up,” a chipmaker says “keep building,” and suddenly retail investors are trying to trade six conflicting headlines before breakfast.

Newsletters like Wealth Builder can help by turning that noise into repeatable frameworks: what changed, what didn't, which numbers matter, what risks are hiding underneath the story, and what would actually invalidate an investment thesis. Passive income and long-term investing become much easier when you have a process instead of constantly reacting to headlines.

If you want more practical, plain-English investing ideas and like-minded newsletters, explore them here: Wealth Builder & like-minded newsletters. Don't chase every headline. Build a system that helps you see through them.

The 3-word takeaway

FOLLOW THE CAPITAL.

Notes & sources

• Dario Amodei, CEO of Anthropic — “We Must Pace the Frontier,” September 2026. Amodei argues that AI capability is accelerating partly because AI is increasingly helping develop AI, and proposes independent evaluators and coordinated safety measures.
• Sam Altman, CEO of OpenAI — agreed that the frontier needs pacing and backed independent evaluators; he also said OpenAI would not pursue an IPO in 2026.
• Elon Musk, CEO of xAI/SpaceX — responded to Amodei's essay: “Dario is right.” Musk's team subsequently conducted a 72-hour AI-agent startup-building experiment.
• Mark Zuckerberg, CEO of Meta — has pursued increasingly autonomous AI agents and consumer AI products; Meta's strategy illustrates the continued push toward deployment rather than simply larger frontier models.
• Jensen Huang, CEO of Nvidia — criticized catastrophic AI narratives while arguing that existing liability laws should be applied and that safety work should accompany rapid development.
• Alex Karp, CEO of Palantir — questioned whether frontier AI companies could ultimately carry the potential liability privately and raised government involvement/nationalization as an extreme scenario. This is Karp's argument, not evidence of an OpenAI nationalization plan.
• Recent development: OpenAI reportedly shelved a planned model release after safety concerns emerged during internal testing, while simultaneously expanding autonomous-agent products.
• Recent development: Altman has now said the benefits of AI justify accepting some risks, while arguing against excessive centralization of AI power.
• AI = Artificial Intelligence.
• AI CapEx = Artificial Intelligence capital expenditure—the money companies spend on long-lived infrastructure such as data centers, servers, networking and power equipment.
• GPU = Graphics Processing Unit, a processor heavily used for AI computation.
• IPO = Initial Public Offering, when a private company first sells shares to public investors.
• S-1 = U.S. Securities and Exchange Commission registration statement used for a public offering.
• RSI = Recursive Self-Improvement, where AI systems increasingly contribute to developing more capable AI systems.
• FCF = Free Cash Flow, cash generated after necessary capital expenditure.

#AIInvesting #AIStocks #WealthBuilder #ArtificialIntelligence #NVIDIA #AIInfrastructure #Cybersecurity #PassiveIncome #RetailInvesting #FollowTheCapital #NoHypeJustResearch