To The Moon

Pretty interesting couple of days for Anthropic. $Anthropic(ANT)$

On November 9, the company introduced Claude Fable 5.2 and Mythos 5.2, showing another major leap in coding, AI agents and complex problem-solving.

Then, just a day later, Reuters reported that Anthropic had withdrawn preliminary IPO pricing information circulated to banks because some of the details were inaccurate.

The withdrawn materials reportedly included a $70–$80 share price range and a potential offering size of up to $69.3 billion.

To be clear, Anthropic hasn't withdrawn its IPO application. The pricing information was retracted, and those numbers shouldn't be treated as confirmed offering terms.

But the pricing issue isn't really what interests me most.

What I find more interesting is the timing.

On one side, AI models keep getting dramatically better. On the other, investors are being asked to justify valuations that were almost unimaginable just a few years ago.

And at some point, those two stories have to meet.

Claude 5.2 is impressive. But what actually matters?

Looking at Anthropic's published benchmarks, the improvements are meaningful.

Agentic coding performance rose from 32.4% to 49.1%.

Software engineering performance increased from 56.3% to 69.9%.

Business workflow automation jumped from 31.4% to 51.4%.

Anthropic also says output generation is up to 30% faster than the previous generation.

Those are strong numbers, although benchmarks don't always translate directly into real-world productivity.

What caught my attention more were the actual tasks Anthropic says the new models can handle.

In one experiment, Fable 5.2 coordinated more than 1,000 AI agents over eight days to build a prototype operating system that could boot on a laptop.

In another, a group of AI agents helped formalize an existing proof of Fermat's Last Theorem in roughly 30 hours.

These are company-reported examples, and independent validation still matters.

But the broader direction is becoming harder to ignore.

AI is moving from answering questions to actually getting work done.

And I think that changes the economics of the industry.

A company might hesitate to spend thousands of dollars on an AI chatbot. But if that same system can reliably complete engineering projects, automate business processes or handle work that previously required several employees, the value proposition becomes very different.

Instead of paying for tokens, businesses start thinking about paying for results.

That's a much bigger opportunity.

The question is whether model companies can capture enough of that value themselves.

Revenue is exploding. So are the costs.

This is where Anthropic's financial disclosures get really interesting.

According to its updated filing, the company generated approximately $16.96 billion in Q3 2026 revenue, up 47% quarter over quarter.

That's an extraordinary amount of revenue for an AI company.

But there's another side to the story.

Revenue growth was 144% quarter over quarter in the previous period, meaning the pace of growth has slowed considerably, even though the business is now much larger.

Anthropic also reported an operating loss of approximately $567 million for Q3.

And then there's the number that really stands out: roughly $895.4 billion in long-term compute-related purchase and lease commitments.

That doesn't mean Anthropic has already spent nearly $900 billion. These commitments cover multiple years.

Still, the scale is enormous.

And it highlights something I've been thinking about for a while.

AI companies don't necessarily have the same economics as traditional software businesses.

Once conventional software is developed, serving additional customers can be relatively inexpensive.

With frontier AI models, every complex request consumes computing resources. Training is expensive, inference is expensive, and increasingly sophisticated agent workflows can require multiple models working together over extended periods.

More customers can mean more revenue, but also significantly more computing costs.

That's why I don't think revenue growth alone tells us enough about the long-term value of an AI model company.

We need to understand how much each task costs to complete, how efficiently infrastructure is being used, and how much of that revenue eventually becomes profit.

There's also the pricing pressure.

New models keep getting faster and cheaper. Great for customers, obviously.

But if one company releases a significantly better model at a lower price, competitors may have to respond.

That makes it harder to maintain pricing power, even when demand is growing.

Of course, lower prices could also unlock entirely new applications and dramatically expand the market.

Both things can be true.

The real question is whether growing usage and improving efficiency can more than offset falling prices.

So what is Anthropic actually worth?

Before the latest IPO developments, Reuters reported valuation discussions reaching as high as roughly $2 trillion, with some investors viewing $1.5 trillion as a high-end reference point.

These are reported valuation expectations, not confirmed IPO pricing.

But let's put those numbers into perspective.

If we simply annualize Anthropic's latest quarterly revenue, we get a revenue run rate of approximately $68 billion.

At a $1.5 trillion valuation, that's roughly 22 times annualized revenue.

At $2 trillion, it's approaching 30 times.

Now, to be fair, annualizing a single quarter isn't the same as forecasting full-year revenue. Anthropic could grow significantly over the next year.

But it gives us a sense of how much future growth investors would already be paying for.

And that's where things become tricky.

A great company can still be an expensive investment.

Anthropic might continue building some of the world's best models, gain enterprise customers and grow revenue rapidly.

But if expectations are already extremely high, even strong results may not always be enough to justify the valuation.

I think its eventual IPO could become an important test for the entire AI sector.

Private-market valuations are one thing. Public-market investors reassessing growth, margins and capital requirements every quarter are another.

What does this mean for AI stocks?

Since Anthropic hasn't completed its IPO, I'm more interested in what its growth could mean for the broader AI ecosystem.

First, the cloud companies.

$AMZN, $GOOGL and $MSFT are all worth watching.

AWS and Google Cloud have important infrastructure relationships with Anthropic, while Microsoft Azure also provides access to Claude models.

If enterprise AI adoption continues accelerating, these platforms could benefit from increased workloads.

But there's a catch.

Cloud companies are also committing enormous amounts of capital to data centers and AI infrastructure.

So higher AI demand doesn't automatically translate into better returns.

We still need to see utilization rates, revenue growth and whether those investments are producing attractive margins.

Then there's the semiconductor side.

$NVDA, $AMD, $AVGO and $TSM all have exposure to different parts of the AI infrastructure buildout.

One development I'm particularly interested in is the shift from training-heavy workloads toward increasingly complex inference.

If AI agents start handling much longer and more demanding tasks, inference demand could become an even bigger driver of future computing investment.

That could create opportunities across GPUs, custom accelerators, networking and advanced chip manufacturing.

But it could also change the competitive balance between hardware suppliers.

And then there's memory.

$MU remains on my watchlist because advanced AI systems require high-performance memory, particularly HBM.

As models and agent workloads become more demanding, memory capacity and bandwidth remain important parts of the infrastructure story.

Still, memory is a cyclical industry. Strong AI demand doesn't eliminate supply risks or pricing cycles.

That's why I wouldn't automatically treat every model upgrade as a buy signal for every AI-related stock.

The companies might all be exposed to the same trend, but their economics are very different.

And there's a big difference between a company benefiting from AI growth and its stock being attractively priced.

My Take

I don't think the AI story is anywhere close to being over.

If anything, the pace of technological progress suggests we're still in the early stages of figuring out what these systems can actually do.

But I do think the way investors evaluate AI companies is starting to change.

For the past few years, the market has focused heavily on capabilities.

Whose model is the smartest?

Who has the most GPUs?

Who is spending the most on AI infrastructure?

Those questions still matter, but they're no longer enough.

At some point, investors will want clearer answers about profits and returns.

Who has real pricing power?

Who can retain enterprise customers when competing models keep improving?

Who can deliver increasingly complex tasks at a lower cost?

And perhaps most importantly, who actually captures the economic value created by AI?

That's the part I find most interesting.

AI could generate enormous productivity gains across the economy, but the companies developing the models might not necessarily capture the largest share of those gains.

Some of the value could flow to cloud providers, semiconductor companies, enterprise software businesses or simply the companies using AI to reduce operating costs.

We don't know yet.

That's why I'm paying more attention to margins, cash flow and infrastructure efficiency rather than simply chasing the latest benchmark scores.

Anthropic's IPO pricing issue doesn't prove that the AI trade is over, or that investors have suddenly lost confidence.

But its eventual public listing could give us a much clearer picture of how the market values frontier AI businesses.

The first phase of the AI boom was about building smarter models.

The next phase might be about figuring out who can actually make money from them.

And for investors, those are two very different things.

Poll: If you could only choose one AI investment theme right now, which would it be?

A. AI model developers — Anthropic, OpenAI and the long-term upside of frontier models

B. Cloud giants — $AMZN, $GOOGL, $MSFT and their enterprise distribution advantages

C. Chips & memory — $NVDA, $AMD, $MU and the infrastructure behind AI

D. Wait and see — Great technology, but valuations and spending are getting hard to justify

I'm curious: **Would