The $28.3 Billion Warning

What happens when you mix:

Investors start looking for the exit before everyone else is forced to.

That may already be happening.

As I covered in Sunday’s Weekly Plan, leveraged money and asset managers have been selling Nasdaq exposure at the same time.

That’s unusual.

These two groups normally sit on opposite ends of the trade.

So when they begin moving together, I pay attention.

And I think the reason goes much deeper than valuations.

The AI trade has a supply problem.

I’m not just talking about physical supply, where memory and neoclouds will be the worst hit.

Memory, data centers, neoclouds… It all matters.

But there’s another kind of supply beginning to worry me even more:

Capital supply.

Every major investment boom eventually reaches a point where money becomes easier to access, whether it is by issuing debt or by Wall Street banks making it conveniently simple to structure complicated financing deals among companies.

I’m not married to an AI bear case by the way.

The Offside Portfolio owns AI longs, it also owns AI shorts.

My job is to let the economics tell me which side deserves more capital.

So far, the short side has won that vote.

This week, SK Hynix handed me another vote.

$28.3 billion worth of it.

SK Hynix chose to buy $0 of its own stock in 2022 when the memory cycle bottomed.

It chose to buy $28.3 billion worth of stock at the peak of the memory cycle and a P/B of 5.0x.

You understand from my previous memory analysis that:

Are all pointing to a peak and turn of the industry in the coming months, with record-high capital commitments coming after it to worsen the effects.

Anyway,

Buying stock at today’s valuation high means SK Hynix must now generate (1 / 5) 20% on that $28.3 billion just to break even on the economics of buybacks.

Which tells you two things:

Either way, it tells you a whole lot about where the cycle is today.

Both scenarios point to operators running their company at peak confidence, somewhat the same narrative that is being pushed to retail of a memory “super cycle.”

The day after the SK Hynix announcement, Micron had to figure out how to deliver a comeback.

So they reiterated a $250 commitment to manufacturing and R&D capacity, along with a new $10 R&D facility in Boise.

Just this week, commitments have gone up by nearly $300 billion, with $700 billion last week between OpenAI and NVIDIA.

We’ll talk about one more in a minute…

For now, here’s what I want you to never forget:

Next time someone tells you the memory cycle is never going to turn back, that this is no longer a commodity…

Show them this chart.

The chip industry has two main cycles by its very nature:

Given the intensity of developing new chips, and how competitive the space has become, this back and forth is more pronounced than ever before!

As of the last twelve months (LTM), the memory industry has seen a record low R&D representation of revenue.

Though not lower than previous normal ranges of 5%-10%.

What should follow next, by industry law, is a return toward the ~15% plus range, which is exactly what you see with Micron beginning to expand R&D capacity and capabilities.

Which is also why, as a reminder from our last post, markets are discounting memory stocks aggressively now:

For the first time since the 2021 capital cycle peak,

First quarter global M&A activity has passed $800 billion, with second quarter activity already racking up to well over historical cycles.

The mechanics here are relatively simple.

As an area of the market gets hot (like SPACs in 2021), Wall Street banks will do whatever they can to get these mandates assigned for IPOs and other transactions like mergers or debt/equity issuance.

Today,

Hyperscalers are issuing stock and debt at record speeds, IPOs are coming in hot, and equity buybacks or stakes (like NVIDIA and Google) are also at record highs.

Here’s why I believe this is dangerous for the cycle:

Were these transactions spread across a wider set of the economy, and this should all be fine.

But,

That’s not the case today, as over 3x the amount of M&A is happening only in tech/AI compared to the next area being Utilities.

Even then,

Utility M&A can also be tied to the data center and energy aspect of the broader AI buildout.

There’s a reason why Google’s deal with Marvell has been broken out into 240 tranches rather than a more reasonable number.

The reason is that Google will have to become more than twice its current size in order to occupy the required revenue gap for the vested shares to be delivered.

Considering that Marvell’s revenue from Google currently is set to grow to $18.5 billion, ten times below the currently required vested rate.

Simply put,

Wall Street has made a relatively straightforward deal more complex so nobody suspects anything.

Google won’t commit much unless Marvell grows, because the entire deal is founded on unrealistic demand numbers.

As we continue to cover this thematic research, you can already see a handful of evidence points leading you to an undeniable peak in the capital cycle now.

More to come on this later,

Stay tuned.

After warning you of SMH exhaustion, the ETF has now rejected the key retracement level and made its way lower.

As we continue to close in on the September volatility breakout after a dispersion unwind, I suspect that the revisit of a 20% drawdown from highs (~$540) will result in a more accentuated liquidation event.

I have already pointed out the two weakest areas of the entire trade should this path continue.

Memory and neoclouds, my highest-conviction shorts since calling the capitulation of South Korean indexes.

In fact,

A specific trade structure has been put on and pitched for the Offside Portfolio, where a long/short options trade will both finance itself the longer it takes, and deliver on over 6:1 R:R.

Here’s the structure I alerted Premium members of.

As I dig further into this capital cycle peak, I am finding several more names exposed to the tail risks of defaults and rising bond yields.

Even some banks are on the hook for this…

But,

That’s a story for another post.

Until then, be safe out there and consider our $2/day premium membership where a review of your portfolio can be delivered upon request.

Because your best defense is having someone like me on your corner.

The Treasury buyback program can be a little confusing at first, and I am currently planning on how to cover it in the simplest of terms.

Meanwhile, here’s a video that attempts to simplify everything that’s happening right now and how it affects the equity market:

Until next time,

OFFSIDE RESEARCH

Against the Tape, Ahead of the Curve.