To The Moon

Something interesting happened across the AI trade this week.

AI demand didn’t suddenly disappear.

Data-centre spending didn’t collapse.

The long-term AI story didn’t magically break overnight.

Yet several AI hardware names were punished despite reporting numbers that, on the surface, looked strong.

That tells me the market may be entering a different phase of the AI cycle.

The question is no longer simply:

“Is AI growing?”

We already know the answer to that.

The more important question might now be:

“Which companies can grow fast enough to justify what investors are already paying for that growth?”

And that distinction could become extremely important heading into September.

💥 MARVELL SHOWED US THE PROBLEM

$MRVL is probably one of the clearest examples.

Marvell delivered strong quarterly growth and raised its outlook, with AI and data-centre demand continuing to drive the business.

Normally that sounds like exactly what investors want.

Beat expectations.

Raise guidance.

Show strong AI demand.

Stock goes up.

Except that isn’t what happened.

The shares were hit hard.

Why?

Because markets don’t reward companies for producing good numbers in isolation.

They reward companies for producing numbers relative to the expectations already embedded in the share price.

When expectations become extreme, “great” can suddenly become disappointing.

That is a very different investing environment from the early stages of the AI boom, when almost any evidence of accelerating AI demand could send semiconductor stocks flying.

🟢 NVIDIA PROVED AI DEMAND IS STILL VERY REAL

Then there is $NVDA.

NVIDIA remains the centre of the AI infrastructure buildout and continues producing extraordinary numbers at enormous scale.

That matters.

If AI spending were genuinely collapsing, NVIDIA would probably be one of the first places we’d see serious evidence of it.

Instead, the bigger debate around NVIDIA increasingly seems to be about expectations, sustainability, capital intensity and how much future growth is already reflected in the valuation.

That is an important distinction.

AI demand can remain incredibly strong while AI stocks still fall.

Those two things are not contradictory.

A company can grow revenue rapidly, increase profits and remain strategically dominant while its share price struggles because investors had already priced in something even better.

That is why I think we’re seeing the AI trade mature.

⚡ THE MARKET MAY BE SEPARATING AI DEMAND FROM AI VALUATION

For the last few years, investors could almost treat “AI exposure” as a single investment thesis.

AI spending rises.

Chip demand rises.

Infrastructure spending rises.

Semiconductor stocks rise.

But eventually markets become more selective.

Now investors appear increasingly willing to ask:

How sustainable are these growth rates?

How much capex will be required?

Where are margins heading?

How much competition is coming?

How concentrated are customers?

How much of the future has already been priced in?

And perhaps most importantly:

Where does the economic value created by AI ultimately accumulate?

That brings me to the other side of this trade.

💰 WHAT IF THE NEXT AI WINNERS ARE THE COMPANIES BUYING THE CHIPS?

Think about $MSFT, $GOOGL, $AMZN and $META.

These companies are spending enormous amounts building AI infrastructure.

For hardware suppliers, that spending represents revenue.

But for the hyperscalers, the objective isn’t simply to own the biggest collection of GPUs on Earth.

They want to turn that infrastructure into products, cloud revenue, advertising improvements, productivity tools, subscriptions and eventually higher earnings.

That creates an interesting possibility.

The next phase of the AI trade may become less about:

“Who sells the hardware?”

and increasingly about:

“Who generates the best return from owning and deploying it?”

That doesn’t mean semiconductor companies suddenly become bad investments.

Far from it.

AI infrastructure could remain one of the biggest technology investment cycles we’ve ever seen.

But leadership inside that cycle can change.

🧠 THE AI TRADE COULD BE SPLITTING

This is the framework I’m watching heading into September.

🔴 AI HARDWARE

Companies such as $NVDA, $MRVL and other semiconductor/infrastructure names still have enormous opportunities.

But many also face:

🔥 Extremely high expectations

🔥 Greater valuation sensitivity

🔥 Increasing competition

🔥 Customer concentration risk

🔥 Huge pressure to continuously beat forecasts

🔥 A market where simply reporting “good” results may no longer be enough

🟢 AI MEGA-CAPS

Meanwhile, the hyperscalers have something different.

💰 Massive cash flows

🏦 Powerful balance sheets

☁️ Existing cloud ecosystems

📊 Huge customer bases

🤖 Multiple paths to AI monetisation

💵 Businesses capable of funding enormous AI investment internally

That doesn’t automatically make them better investments.

But it changes the risk/reward equation.

A company selling AI infrastructure needs customers to continue spending.

A hyperscaler deploying AI infrastructure needs that spending to eventually generate an acceptable return.

Different businesses. Different expectations. Different risks.

And increasingly, potentially different stock-market outcomes.

📊 THIS IS WHY “BEAT OR MISS” ISN’T ENOUGH ANYMORE

One thing I’ve been trying to improve in my own investing research is looking beyond whether a company simply beats earnings estimates.

Because a beat tells us what happened compared with analyst forecasts.

It doesn’t necessarily tell us what the market expected.

Those are different things.

A stock can beat Wall Street estimates and fall 10%.

Another can technically miss a metric and rally.

The reaction often tells us something about positioning and expectations that the headline numbers don’t.

That’s exactly why Marvell caught my attention.

Strong company.

Strong AI exposure.

Strong growth.

Yet the market effectively said:

“We wanted more.”

That is incredibly useful information.

🎯 MY SEPTEMBER AI WATCHLIST IS CHANGING

I’m still bullish on the long-term infrastructure required to support AI.

But I’m becoming more selective about where I’d want exposure.

Instead of simply asking:

“Which company benefits from AI?”

I’m asking:

1. What growth is already priced into this stock?

2. What would the company need to deliver to surprise investors from here?

3. Is the company selling the AI infrastructure or monetising what that infrastructure produces?

4. What happens if hyperscaler capex growth eventually slows?

5. Which businesses still have room for expectations to move higher?

Those questions matter more to me now than simply finding another stock with “AI” somewhere in the investment thesis.

🔭 WHAT I’M WATCHING NEXT

September could tell us a lot about where this cycle is heading.

I’ll be watching hyperscaler capex closely.

I’ll be watching semiconductor guidance.

I’ll be watching whether hardware weakness spreads or stabilises.

I’ll be watching valuations.

And I’ll especially be watching whether money starts rotating toward companies that can demonstrate actual AI monetisation, rather than simply exposure to AI spending.

Because perhaps the market isn’t abandoning the AI trade at all.

Perhaps it’s simply becoming more demanding.

🧩 MY TAKE

I don’t think the AI boom is over.

I think the easy version of the AI trade might be.

There was a period where identifying the companies supplying the AI buildout was enough.

The next stage could require something harder.

Understanding expectations.

Understanding valuation.

Understanding where margins ultimately settle.

Understanding who has pricing power.

And understanding which companies can turn billions of dollars of AI investment into sustainable free cash flow.

That’s why the recent weakness across AI hardware interests me more than it scares me.

It could be noise.

It could be a temporary rotation.

Or it could be the beginning of a much more selective phase of the AI market.

And if that’s what’s happening, the biggest winners of the next stage may not necessarily be the same companies that dominated the last one.

🐯 TIGER COMMUNITY

What do you think?

Is this just a temporary AI hardware reset, or is the AI trade genuinely beginning to split?

If you had to choose one group for the next stage:

🅰️ AI hardware and semiconductors

🅱️ Hyperscalers and AI monetisation

🅲️ Both, but only at the right valuation

🅳️ Neither, expectations are still too high

I’m especially interested in where everyone thinks the next dollar of AI profit ultimately lands.

$NVDA $MRVL $MSFT $GOOGL $AMZN $META

Personal market analysis and opinion only. Not financial advice. Always do your own research.