• Moonshot Minute
  • Posts
  • Anthropic's (ANT) $1.4 Trillion Question: Should You BUY Tomorrow?

Anthropic's (ANT) $1.4 Trillion Question: Should You BUY Tomorrow?

Three days. That's how long Anthropic says its AI took to get Claude running on AMD chips.

Premium Members: There's a short portfolio update at the top of the Premium section below, then a link to Friday's Buy Alert, exactly as it ran on November 13.

Three days.

That's how long Anthropic says it took its newest AI model to accomplish something that once required a team of engineers months of work.

What did it do?

It helped get Claude running on AMD processors.

Not Nvidia!

AMD.

Now, if that doesn't sound particularly exciting...

Consider the repercussions.

For years, Nvidia has enjoyed an extraordinary advantage in artificial intelligence.

Not simply because it makes the most powerful AI chips.

But because its chips are supported by a vast ecosystem of software that makes switching to competitors difficult, expensive, and time-consuming.

That software ecosystem is called CUDA.

And it's one of the primary reasons Nvidia has become the dominant supplier of AI computing hardware.

But during Anthropic's IPO roadshow last week, its Chief Compute Officer, Tom Brown, revealed something remarkable.

Anthropic had used its own advanced AI model, Mythos, to help move a Claude workload onto AMD hardware.

According to Anthropic, Claude wrote or adapted almost all the required computing kernels.

With very little human intervention. In approximately three days.

Now, the experiment involved prompt processing for one model on AMD MI355X processors.

It wasn't a complete migration of Anthropic's computing infrastructure.

Still...

What happens when AI can automate the engineering work that helps protect Nvidia's CUDA monopoly?

That's a question worth trillions of dollars.

And it brings us to this week's other enormous technology story.

Anthropic's IPO.

The company behind Claude is expected to enter the stock market under the ticker ANT.

Tomorrow, November 19.

Its proposed $70-to-$78 share price implies a valuation as high as approximately $1.4 trillion.

One of the largest IPOs in history.

And behind Anthropic is one of the most interesting figures in Silicon Valley.

Dario Amodei.

A man who helped build the AI revolution...

Yet worries deeply about what happens when it becomes too powerful.

Keep him in mind.

Because his beliefs could eventually matter as much to Anthropic shareholders as the technology itself.

So today, we're going to examine three questions.

Can Anthropic's technology dismantle Nvidia's competitive moat?

Can Anthropic turn its own technology into a profitable business?

And most importantly...

Should you buy Anthropic stock?

The Monopoly That AI Is Learning to Escape

In Tuesday's issue, I explained why I remain bullish on Nvidia over the next four or five years.

Despite Jensen Huang's extraordinary departure. Despite the Chinese export restrictions. And despite growing competition.

Nvidia's competitive advantages extend far beyond one executive or one processor.

Its network of hardware, software, developers, customers, and manufacturing partners is extraordinarily difficult to reproduce.

But here's what interests me about Anthropic's announcement.

That monopoly is becoming easier to work around.

For years, moving complex AI software from Nvidia hardware to competing processors required significant engineering effort.

Developers had to rewrite and optimize code. Test performance. Fix compatibility problems. And sometimes sacrifice speed or efficiency.

That gave Nvidia tremendous pricing power.

Customers couldn't simply replace Nvidia processors with cheaper alternatives.

At least, not easily.

But Anthropic's Mythos demonstration suggests AI may dramatically reduce part of that switching cost.

Think about what that means.

Instead of spending months making your software compatible with a new computer...

You assign an AI system to solve the problem.

And it does much of the work itself.

Anthropic now describes a strategy in which computing resources from Nvidia, AMD, Google, and Amazon can operate as a more interchangeable pool.

Not perfectly interchangeable.

The hardware still differs enormously.

But the goal is clear.

Make the software increasingly independent of the hardware.

And that could eventually change the economics of the entire semiconductor industry.

Because when buyers have more choices...

Suppliers lose some pricing power.

Does that mean Nvidia's CUDA monopoly is dead?

Absolutely not.

Nvidia's networking, hardware performance, software libraries, and installed base remain formidable advantages.

But a crack may be forming in its moat.

And Anthropic is holding the chisel.

Why Anthropic Wants Cheaper Computing

Now consider something interesting about the timing.

Anthropic made this announcement while asking investors to value the company at approximately $1.4 trillion.

Coincidence? I doubt it.

Because buried inside its IPO prospectus is a number that should make every potential shareholder sit up.

$900.9 billion.

That's approximately how much Anthropic has committed to spend under agreements largely involving computing infrastructure.

(In October, I cited about $518 billion. That came from an earlier draft of the prospectus. The amended filing counts more agreements.)

Not all tomorrow.

Approximately $451.6 billion is expected before the end of 2030.

The remainder extends across subsequent years, substantially through 2045.

That's an astonishing amount of money...

And it helps explain why Anthropic is so interested in making computers interchangeable.

If it can run Claude efficiently across several hardware platforms...

It can choose more cost-effective resources.

Negotiate better agreements.

Reduce its dependence on individual suppliers.

And potentially improve profitability.

Investors certainly want to hear that.

Because Anthropic's computing costs are enormous.

During the first nine months of 2026, the company generated $33.2 billion in revenue.

Its computing and infrastructure expenses?

$18.8 billion.

That's approximately 57 cents of every revenue dollar.

And in the third quarter, computing expenses increased to 58% of revenue.

Anthropic chart of quarterly revenue against compute and infrastructure expenses, Q1 2025 to Q3 2026

Source: Anthropic

Even as Claude becomes more efficient, Anthropic keeps acquiring more computing capacity.

Why?

Because it's making a bet on Jevons paradox.

The Jevons Paradox

In 1865, English economist William Stanley Jevons noticed something peculiar about coal consumption.

As steam engines became more efficient... Britain consumed more coal.

Not less. Why?

Because cheaper, more efficient steam power created additional uses for steam engines.

Factories expanded. More machinery became economically practical.

And total demand for coal increased.

Today, economists call this the Jevons paradox.

Now apply the same principle to AI.

As algorithms become more efficient, the cost of performing a given computing task declines.

That enables more applications.

More AI agents. More automated research. More complex reasoning. More business automation.

And potentially millions of industrial processes that previously couldn't justify the cost.

Meanwhile, frontier AI laboratories aren't satisfied with producing today's intelligence more cheaply.

They're trying to create vastly more capable intelligence.

AGI.

And eventually ASI.

Every improvement in efficiency gives them additional resources to push toward those goals.

So cheaper computing doesn't necessarily mean less computing.

It could mean dramatically more.

The cost of intelligence falls. The consumption of intelligence explodes.

Not a mathematical certainty.

But a powerful economic possibility.

And one reason I remain bullish on Nvidia even as Anthropic works to reduce its dependence on Nvidia's hardware.

The total market can expand while competitors capture more of it.

So, how do we value Anthropic in this context?

What Business Is Anthropic Actually In?

Most investors think of Anthropic as the company behind Claude.

An AI chatbot competing with ChatGPT.

But that's a little like describing Amazon in 2005 as a company that sells books.

You miss the bigger opportunity.

Anthropic is developing several distinct businesses.

And I think investors need to understand them individually.

Because their economics, and their chances of success, are very different.

1. Claude Code: Rewriting the Software Industry

Claude Code helps programmers write, debug, test, and maintain software.

I use it every day in my work. In fact, the work I do in one day by myself used to take a team of four developers about two weeks to do.

But it increasingly does something more important.

It performs substantial engineering tasks autonomously.

That's a direct challenge to the existing software-development industry.

Coding assistants. Developer tools. Testing systems.

And potentially portions of the services businesses built around software engineering.

I consider this one of Anthropic's strongest near-term opportunities.

Developers have an obvious economic reason to adopt tools that help them produce more working software.

But Claude Code doesn't necessarily need to replace a developer's favorite tools.

It can operate inside those environments.

That distinction is important.

2. Claude Cowork: The Next Office Worker

Cowork extends similar capabilities to ordinary knowledge workers.

Instead of writing a program...

It can manipulate spreadsheets. Prepare reports. Research information. Navigate applications. And complete multistep projects.

(This is the white-collar wipeout I've been writing about.)

That places Anthropic in competition with Microsoft, Google, Salesforce, and a vast number of smaller enterprise software companies.

The potential market is enormous.

But there's a problem.

Microsoft and Google already own the environments where much of the world's office work happens.

Replacing them outright won't be easy.

Both are also investing heavily in their own AI systems.

So I don't assume Anthropic will displace their software franchises.

It may be more profitable to become the intelligence operating inside existing applications.

Which brings us to my favorite part of Anthropic's business.

3. Claude Platform: Intelligence Inside Everything

Consider a company like Siemens.

The German industrial technology giant makes the software and automation systems that help operate factories around the world.

Its customers rely on sophisticated engineering tools, industrial controllers, specialized machinery, and decades of accumulated expertise.

Now imagine an engineer troubleshooting a production problem.

Instead of manually reviewing hundreds of diagnostic messages, equipment logs, and engineering documents...

An AI assistant analyzes the information.

Suggests possible causes.

Identifies relevant engineering changes.

And helps develop PLC code for an engineer to review and test.

All inside software the company already uses.

Siemens industrial automation software showing PLC programming blocks

Source: Siemens

That's potentially enormously valuable.

And Anthropic wouldn't have to reproduce Siemens' engineering products, installed base, or customer relationships.

It could provide intelligence.

Siemens could provide the application and industrial expertise.

This is a hypothetical example.

Industrial control applications would also require rigorous validation and appropriate human oversight...

But imagine similar arrangements across thousands of established monopoly companies.

Financial software. Insurance platforms. Medical systems. Engineering applications. Manufacturing software. Logistics. Cybersecurity. Legal research. Scientific instruments.

Suddenly, Anthropic isn't simply selling chatbot subscriptions.

It's supplying intelligence to the existing software economy.

And I think that could become its most important long-term business.

So... If Anthropic supplies the intelligence while Siemens or another software company owns the customer relationship...

Who captures the profits?

That's a question investors should be asking...

The Product Mix Nobody Should Ignore

Anthropic's prospectus provides an interesting clue.

During the first nine months of 2026...

79% of revenue came from consumption-based services.

Only 21% came from subscriptions.

In other words, much of Anthropic's business already depends on charging for the intelligence customers consume.

But the company doesn't provide a detailed revenue breakdown for Claude Code, Cowork, Security, healthcare, finance, and its other applications.

I'd like to see those numbers.

Because Claude Code has different competitors and economics from healthcare AI.

And selling intelligence to industrial software developers is a different business from selling $20 monthly subscriptions to individuals.

Anthropic says its Claude Partner Network includes more than 4,000 organizations with certified practitioners.

And 44% of its revenue during the first nine months of 2026 came through cloud-provider marketplaces.

That suggests a meaningful distribution strategy is already operating.

Anthropic chart of customers paying at least $100,000 a year, rising to more than 11,900 by September 30, 2026

Source: Anthropic

But cloud distribution isn't the same as deep integration into customers' most important software moats.

Until this happens...

If another AI model becomes better or substantially cheaper, a software vendor could switch.

Which makes Anthropic's own competitive moat an interesting question.

It wants computing hardware to become interchangeable.

But what happens if AI models themselves become interchangeable?

Anthropic might help weaken Nvidia's monopoly...

Only to find its customers doing the same thing to Anthropic.

So, Should You BUY Anthropic's IPO?

Here's my take.

I'm extremely bullish on Anthropic's technology.

But at the proposed IPO valuation, I'm not ready to recommend buying its stock.

Not yet.

And the reason comes down to a little mathematics.

Anthropic's revenue growth has been extraordinary.

Its third-quarter revenue reached approximately $17 billion.

By the end of October, the company reported an annualized revenue run rate of roughly $81 billion.

That's remarkable for a business founded in 2021.

But an $81 billion revenue run rate isn't the same as $81 billion in annual profit.

Or even $81 billion in completed annual sales.

And at a $1.4 trillion valuation, investors are paying approximately 17 times that run-rate revenue.

Let's use a five-year investment horizon.

Suppose Anthropic eventually achieves a 25% net profit margin.

And suppose investors are willing to value it at 25 times annual earnings in 2031.

What would Anthropic need to earn to deliver a 10% annual return from a $1.4 trillion starting valuation?

The answer is approximately $90 billion in annual net income.

At a 25% margin, that requires roughly...

$360 billion in annual revenue.

That's more than four times Anthropic's current reported revenue run rate.

It would require approximately 35% annualized growth over the next five years.

And that's assuming a substantial net profit margin and a healthy terminal valuation.

Neither is guaranteed. Consider the risks.

Competition from OpenAI, Google, Meta, and other model developers.

Falling prices for intelligence. Growing infrastructure commitments. Potential customer switching.

And a computing bill that could continue increasing even as individual workloads become cheaper.

And, having read the prospectus for you...

There's one more risk that deserves attention.

He has a name.

Dario Amodei: The CEO Who Might Pull the Plug

Dario Amodei isn't your typical Silicon Valley entrepreneur.

A former Google Brain researcher and vice president of research at OpenAI, he helped develop some of the technology that made today's AI revolution possible.

He left OpenAI in 2020 and co-founded Anthropic with his sister Daniela and several colleagues.

One important reason?

Concerns about how increasingly powerful AI should be developed and controlled.

Dario has spent years thinking deeply about the extraordinary possibilities of artificial intelligence.

And its dangers.

His ideas overlap with the AI safety concerns often associated with the effective altruism movement.

Although he doesn't identify as an effective altruist.

To me, his caution is understandable.

In fact, I admire it.

But now he's about to become CEO of a publicly traded company valued at approximately $1.4 trillion.

And that creates an extraordinary tension.

What happens when doing the right thing for humanity conflicts with doing the profitable thing for shareholders?

In fact, we've already seen a preview...

Earlier this year, Dario clashed with the administration over the Pentagon's use of Claude.

Anthropic refused to remove restrictions involving two applications.

Mass domestic surveillance. And fully autonomous weapons.

The Pentagon argued it needed to use AI for lawful military purposes and that the government, rather than a private contractor, should determine those uses.

Dario wouldn't agree to the requested terms.

The dispute threatened a defense agreement worth up to $200 million.

It escalated into legal action, and the Pentagon moved to exclude Anthropic from defense work.

In September, a federal appeals court upheld the Pentagon's supply-chain-risk designation, although Anthropic prevailed in a separate California case involving other government actions.

That's a serious business consequence.

And it isn't the only example.

Consider Anthropic's latest cybersecurity research.

Remember Mythos?

The same AI system that helped Anthropic move its software onto AMD chips?

It has also demonstrated an extraordinary ability to discover software vulnerabilities.

Anthropic has shown that Mythos can do more than discover vulnerabilities.

In some cases, it can develop working exploits autonomously.

By November 2, Anthropic reported that its models, including Mythos, had disclosed 5,087 vulnerabilities across 680 open-source projects.

Now imagine what those capabilities might mean in the wrong hands.

Anthropic initially restricted access to Mythos Preview, working with selected organizations on defensive cybersecurity rather than making the model generally available.

And its IPO prospectus reveals another remarkable decision.

The company recently delayed pretraining a new frontier model for more than a month to improve its defenses against theft of the model's weights.

Anthropic is already sacrificing speed to manage AI risk.

But here's something even more significant.

In September, Dario published an essay titled We Must Pace the Frontier.

He argued that AI capabilities are advancing so quickly that safety measures risk falling behind.

And that the industry may need to deliberately slow its rate of advancement to keep those safeguards effective.

Not stop AI development forever.

Slow it enough to understand and manage the risks.

That's a remarkable position for the CEO of a company spending hundreds of billions of dollars to develop increasingly powerful AI.

And it leads to an uncomfortable question.

What happens if Dario decides Claude has become too dangerous to release?

Can he simply turn it off?

Not unilaterally, as far as Anthropic's disclosed governance arrangements establish.

But the company has deliberately given itself the ability to delay training, restrict deployments, and prioritize safety over short-term commercial opportunities.

And here's the part that matters to shareholders.

Anthropic's seven co-founders will collectively control 50.1% of ordinary shareholder voting power through a special Founder LLC.

That isn't Dario's personal voting stake.

But it means outside shareholders have limited influence over ordinary shareholder decisions.

Meanwhile, an independent Long-Term Benefit Trust generally has the authority to elect and remove a majority of Anthropic's board.

The company is also organized as a public benefit corporation.

Meaning its directors must balance shareholders' financial interests with its stated public benefit purpose and the interests of people affected by its actions.

Think about that structure.

A company asking the public to invest billions...

While deliberately limiting the public's ability to influence its direction.

And empowering a board structure designed to protect its mission.

Of course, this could prove to be an extraordinary competitive advantage.

Careful development could protect Anthropic from catastrophic failures, regulatory action, and reputational disasters.

It could help build the trust required to put Claude inside banks, hospitals, factories, and other critical institutions.

But there's another possibility.

A competitor releases a more powerful model while Anthropic delays its own. Customers leave. Revenue suffers.

And ordinary shareholders have limited ability to change the company's direction.

To me, that's a genuine investment risk.

Not because Dario is necessarily wrong.

In fact, he may be exactly the kind of CEO humanity needs.

Remember the Ford Pinto?

Ford executives knew their cars had dangerous fuel-tank vulnerabilities.

Yet the company sold them anyway.

Or Boeing's 737 MAX?

Competitive and financial pressures helped create a situation where critical safety failures went unaddressed.

346 people died.

In each case, the people responsible for developing or deploying a powerful technology failed to adequately protect the people depending on it.

Now imagine a technology capable of discovering thousands of previously unknown security vulnerabilities...

Writing its own software...

And eventually outperforming humans at increasingly consequential tasks.

That's the responsibility Dario carries.

And I believe he takes it seriously.

But here's the problem for investors.

His responsibilities to humanity and his responsibilities to shareholders may occasionally pull him in opposite directions.

And with Anthropic's unusual governance structure, shareholders may have very little influence when those decisions are made.

Again...

Technology wants to be everywhere.

Yet Anthropic's leadership has deliberately preserved the ability to say...

Not yet.

That's something investors should understand before buying ANT.

My Investment Decision

So my position is simple.

I would watch Anthropic's IPO rather than chase it.

To be clear, I already own Anthropic shares I bought pre-IPO (as I mentioned in October), and I may add more.

I want more evidence that its expanding product portfolio can produce durable profits.

I want revenue growth by business line.

I want to understand how much revenue comes from enterprise integrations versus standalone applications.

And I want to see whether computing costs continue declining relative to revenue as Anthropic expands.

I also want to understand how its unusual governance affects commercial decisions.

Because the company's long-term success will depend on more than having the smartest AI models.

It will depend on getting those models into applications that customers use every day.

Profitably. Reliably. And with enough competitive advantage to keep those customers from switching.

If those answers improve, and the stock trades at a reasonable valuation, Anthropic could become an extraordinary long-term investment.

I'd become particularly interested if its valuation fell below $1 trillion, provided the underlying business remained strong.

That isn't an automatic buy price.

But it would give investors more room for error than the proposed IPO valuation.

I won't confuse enthusiasm for a technology with enthusiasm for any price attached to it.

That's a distinction Warren Buffett understood long before the first AI chip existed.

Technology Wants to Be Everywhere

Tuesday, we examined Nvidia's extraordinary rise...

Jensen Huang's fall...

And the political, technological, and manufacturing obstacles confronting its future.

Today, we've looked at a company trying to make some of those obstacles disappear.

Anthropic wants to make intelligence available across multiple chip architectures.

Across software products. Across industries. And eventually, across much of the global economy.

In fact, its own IPO prospectus describes a strategy to make Claude available everywhere.

Sound familiar?

I've been arguing that technology wants to be everywhere.

Anthropic is attempting to make that idea a business model.

But there's a fascinating difference between Tuesday's story and today's.

Jensen Huang spent decades building a technological empire he ultimately lost control of.

Dario Amodei is building another technological empire...

One whose governance was deliberately designed to limit the power of ordinary shareholders.

And one whose leadership may choose to slow the development or deployment of its own products.

Not because the technology is failing.

But because it's becoming extraordinarily powerful.

That tension could become one of the most important stories in the AI industry.

Will AI continue spreading into every industry?

I believe so.

Will Anthropic capture enough of the profits to justify a $1.4 trillion valuation?

That remains to be demonstrated.

For now, I'd rather own a proven economic moat at a reasonable price than chase an IPO whose most ambitious financial outcomes may already be reflected in its valuation.

But Anthropic is absolutely a company I'll be watching.

Because technology wants to be everywhere.

And Dario Amodei may be one of the few people willing to tell it to slow down.

Always be prospering,

Harry Seldon
Moonshot Minute

P.S. There's an interesting irony here. Nvidia is among the companies investing in Anthropic alongside its IPO. So even as Anthropic develops technology that could eventually weaken Nvidia's competitive moat, Nvidia is investing in the company developing it. In technology, today's competitors can also be tomorrow's customers, suppliers, and shareholders.

P.P.S. Here's a screenshot of the current Moonshot Minute Portfolio. I've blurred out the tickers since that information is only for Premium Members, but you can see how we've done so far:

🔓 Premium Content Begins Here 🔒

In today's Premium Section: Friday's buy alert. A name, a price, a watch price and the five things that would make us sell, plus a straight answer on our first rare-earth pick after its quarter. If you're not a Premium Member yet, you'll want to join today.

I hope you’ve been paying attention because many of our picks are currently beating the S&P by up to 4-to-1 over the last 12 months.

Most financial newsletters charge $500, $1,000, even $5,000 per year. Why? Because they know they can.

I don’t.

I built my wealth the old-fashioned way, not by selling subscriptions.

That’s why I priced this at $35/month, or $300/year.

Not because it’s low quality, but because I don’t need to charge the typical prices other newsletters charge.

One good trade, idea, or concept could pay for your next decade of subscriptions.

The question isn’t ‘Why is this so cheap?’ The question is, ‘Why would I charge more?’

P.S. If this newsletter were $1,000 per year, you’d have to think about it.

You’d weigh your options. You’d analyze the risk.

But it’s $35 a month.

That’s the price of a bad lunch decision.

And remember, just one good idea could pay for your subscription for a decade.

Recent comments from Premium Members:

Amazing! Moonshot is hands down the best $150 decision I have ever made. Up 64% on TICKER REDACTED (so far). Can't thank you enough for your service, advice, recommendations, insight, and every other positive accolade in the dictionary.

Very respectfully and gratefully,

CK

I’m up 71% in 6 weeks would you recommend adding to this bucket if capital allows?

MS

Hello Double D,

As it happened, I already owned some TICKER REDACTED shares when you recommended the stock. Upon your recommendation, I bought more. All told, I'm up over 70% in a month or two.

I greatly appreciate the detailed discussion you and your team provide for your recommendations.

Thank you.

A happy subscriber,

PK

I finally got some liquidity I could use and bought the stock as well as March 2026 calls yesterday morning (October 2nd) when it was at $17.80.

That is easily the best-timed investment/trade I've ever made, and I have your team's perpetual hard work and research to thank for it.

Thanks again for all the hard work. You and your team push out a lot of solid research, and the effort doesn't go unnoticed.

It is greatly appreciated,

MD

Closing at 24% gain, and enough profits to pay for 2yrs of your newsletter. Thank you for this! I especially appreciate you detailing the rationale behind your picks. As a newer investor it’s important for me to know why just as much as what.

MS

Up 68.87% on TICKER REDACTED to date, great recommendation!

BW

Thanks for the great tip on TICKER REDACTED! I bought, and just eight trading days later, it's up 52% as of this very minute. I'm new to your newsletter, less than two months, but I have found it to be quite soundly researched and a truly invaluable source.

I've been actively investing for many years and have, at one time or another, subscribed to various investment advisors. None have been as useful (nor as affordable!) as the Moonshot Minute.

You, sir, do excellent work and we individual investors much appreciate it.

SD

Up 66%! Thanks.

SB

Kudos and thank you for the TICKER REDACTED recommendation. TICKER REDACTED has been awesome and I do understand/believe this to possibly be only the beginning. I bought 200 shares at $16 and another 100 at $18, the day before the surge started. Again, I am very grateful.

RH

I've only been with you a few weeks now, and overall, my portfolio is up 41%. Couldn't have done it without you, DD. Thanks again.

HJP

I joined your plan about 2 months ago.

TICKER REDACTED was a real hit - and I did fully realize it yesterday for a rise - 141%. Great deal!!

IS

Just wanted to drop you a quick THANK YOU! Been a member for about a week (I wanted to see your picks for the electrical asymmetry) and I picked up some TICKER REDACTED & TICKER REDACTED. I’m already up $1,300.00 so my membership is covered for 5 years in about a week!

Keep up the great work! Again, THANKS! Glad to be a subscriber!

RH