This is When AI Winners Become Losers

Everyone is obsessed with Micron’s “cheap” P/E of ~6x.

What most fail to realize is that cyclical companies with a cheap valuation multiple spell more trouble than opportunity.

Simply put, a cheap P/E tells you the market expects earnings to get worse.

Here’s a great way to put it:

In the end, Intel went on an insane run while NVIDIA underperformed the entire chip and tech sector.

Where it gets tricky is figuring out whether the market is right or wrong in assigning these valuation multiples to such stocks.

Considering the amount of analysts and independent research services covering a name like Micron, it’s a lot easier to pinpoint the shortfalls versus a more obscure name.

And all these opinions seem to be focused on one thing and one thing only:

I’ve introduced you to the concept of expectations investing, where we begin by asking “what must be true” instead of making our own assumptions about the future.

In the same contrarian way, I would like to expose you to the Offside view in this topic.

Rather than focusing on demand (which is infinitely harder to forecast.)

I want you to focus on the supply side instead, which more often than not is laid out in front of us.

As it turns out, supply has begun hitting every single corner of the AI trade this year.

From the ~$725 billion in hyperscaler capex flooding the capital supply…

To the introduction of Chinese, Japanese, South Korean, and European players in the semiconductor and memory manufacturing supply chain.

We all know what the demand figures look like, unable to be filled for at least five years according to Micron’s management team.

What these analysts seem to ignore is how many chips are actually beginning to come online now.

As record profits are centered in a handful of names, competitors begin to overinvest in order to catch up and overshoot the supply they bring to market, and in the end, profit margins and returns on capital collapse (along with stock prices.)

There are plenty of examples.

Simplified by focusing on supply rather than demand.

In fact, I already showed you why demand will in fact be a lot weaker than Wall Street thinks.

Today, we answer the other side of the equation.

Will supply turn AI winners into losers?

Bigger balance sheets should lead to higher stock prices, right?

That’s usually true in the short term, but today’s evidence will show you that - in the long run - that is not the case.

In fact, quite the opposite is true!

When companies spin off divisions, buy back stock, pay dividends, and restructure their balance sheets… Stock prices tend to go up.

This is what many industry professionals call the “Asset growth anomaly.”

It states that businesses that invest and grow their balance sheets aggressively tend to see their share prices fall after a short-term upswing.

There’s a fine balance between underinvesting and overinvesting.

Right in the middle is where the most returns on capital come as a result, but the one that destroys capital the fastest is the overinvestment.

So, as hyperscalers are set to invest a combined $725 billion on AI infrastructure projects in 2026…

I strongly believe they will fall into the right side of this curve and see their returns fall along with their stock prices.

Some are already suffering from this anomaly:

The reason why this anomaly exists is because of the market’s short-sighted leniency.

When demand looks strong for the foreseeable future, everyone rushes to get a piece of that demand, and usually the first wave of implementation sees the bulk of benefits in profits and rising margins.

However,

After this initial wave of success, competitors begin to realize that they too must bet even larger amounts in order to remain relevant and not lose market share.

What you have in the end is today’s capex monstrosity, where the anxiety of getting left behind has led the “Big 4” to overinvest and overshoot their expectations.

We can begin to time this capital cycle by gauging return measures in the AI space.

How high and how long this goes, nobody knows.

But,

What I can tell you is that the peak is already in for the semiconductors from a capital cycle perspective.

They’ve overbuilt, and are starting to pay for it.

Now it’s down to the logic and memory names to see this return cycle peak, where I suspect stock prices will follow.

Given we are now in a new earnings season, that’s a must-watch development for these companies.

In fact, I think this is why some multiples have remained so low, the market expects this cycle to have peaked, and for expectations to be missed.

This is where the herd mentality is spotted.

I mentioned that, as record profits center around a handful of names, more competitors choose to enter and take a piece of those records for themselves.

When this happens, the entire market mechanics begin to change as well.

Take a look at the hyperscalers:

Throughout every capital cycle, it is the investment bankers that drive it all.

Today, the above has turned into:

In other words,

It is not only the capital supply that’s beginning to come in hot, it is the materials, chips, and computing availability that has also come in hotter than expected.

Markets have seen this before, which is why they’re so quick to discount the hottest names in the AI trade right now.

I put together a report for you to compare today’s capital cycle to that of the internet bubble as well.

Now here’s where everything starts to tie up together:

The statistical extremes between value and growth stocks was reached in October 2025.

Consequently, this is when the “HALO” trade was born, and when real economy names started to outperform even the hottest AI names.

Then 1Q’26 changed everything…

Capex came into the mix, and a sharp reversal back into growth was made.

Growth investors love growth, and momentum investors love momentum!

But,

What happens when the capital cycle that’s already killed off semiconductors starts to hit memory and compute?

Growth slows alongside momentum.

And value wins the day.

The only thing you need to ask is:

Which AI winners will become losers?

I think the capital cycle peaking in memory/compute has influenced some options traders already:

The expectation is for Micron to head lower into the $600-$800 range by the end of summer.

I suspect that this earnings season will shed some light into the state of returns in the sector, and whether constituents are affected by the overinvestment wave that’s currently going on.

That said, I am now prompted to look into the space for some long/short equity trade ideas to bring forth to my Premium Members.

Over the past month, these calls have brought amazing results, and I expect this hotter area to be no different.

Consider joining today, and lock in lower rates.

Prices will go up starting July 17th (Friday.)

In the meantime, here’s a podcast episode from one of my former mentors after I left Goldman Sachs, where they touch on the rotation theme I have outlined several times in morning digests:

Until next time,

OFFSIDE RESEARCH

Against the Tape, Ahead of the Curve.