In partnership with
The Last Time Stocks Were This Expensive Was December 1999.
"Right now, it's good. But it was in '72, '86, 2000, and 2007." - Jamie Dimon, May 2026.
The Shiller CAPE ratio just hit 42.3. The only time in 140 years it's been higher? December 1999.
Stocks can stay expensive for a long time...
It’s one metric to consider, but when your portfolio is built around the most expensive equities in modern history, what else you diversify with could really matter.
Blue-chip contemporary and post war art has shown near-zero correlation with the S&P since 1995.* Prices are largely driven by private collectors competing for a fixed supply of artwork by artists like Banksy, Basquiat, and Picasso.
Masterworks lets you invest in shares of that market.
$1.3B deployed across 500+ artworks
29 exits to date
Net annualized returns like 16.5%, 17.6%, and 17.8%, not including those unsold
*According to Masterworks data. Investing involves risk. Past performance is not indicative of future returns. See important Reg A disclosures at masterworks.com/cd.
THE MOST IMPORTANT NUMBER
Every AI bull has the same argument:
“$7 trillion of capex is coming.”
They’re right.
But,
They’re focused on the wrong number…
Capex only continues if someone actually turns enough revenue to justify new investment.
This flow chart boils down the entire trade into one bottleneck, where enterprise software companies become the only reasonable players who can fix the biggest issue nobody’s willing to talk about:
AI models have committed over $3 trillion in spending
These models only generate $200 billion in ARR
Individual users like you and me won’t get them to close the gap, neither will small to medium-sized companies.
Only large enterprises and corporates can get this done.
Which is why I began covering the entire software value chain, starting with three of the moats that are absolutely necessary to making this work:
PCE Inflation, Offside Capital
When does it all begin to matter?
I have no idea, but I think inflation could be one of the possible triggers.
PCE came in at 4.1% this week, and the PPI makes it clear that AI-related commodities are the ones responsible for driving inflation indexes higher right now.
Which is why Apple and Microsoft were the first to raise their hardware prices, citing a historical spike in memory and component prices.
More importantly, here’s how that all affects the S&P 500:
At current prices, history suggests we are set up for a 2% to -2% return for the next decade
With AI driving inflation past 4% again, investors buying here are set up to lose 2% to 6% each year through 2036
I’ve given you a strong timing tool in the Commitment of Traders report.
Today, I’ll show you an even stronger signal.
For almost a decade, semis and hyperscalers have traded as if they were the same business.
That relationship just broke.
One side of the AI trade is beginning to tell a very different story.
The question is:
Which one is wrong?
CHART OF THE DAY
South Korea’s PMI and stock market is highly correlated to the United States.
Both as a manufacturing PMI leading indicator.
And,
As a leading indicator to where technology and AI-related stocks are headed.
Across the world, this market is trading in full leverage and speculation mode, with people cashing in life insurance policies to buy the rally in tech.
ARE YOU COVERED? —>
This is eerily similar to the setup that led to a lost decade in Japan, and even the great financial crisis in the US.
The thought that tech or AI “only goes up” is leading many to bet 100% of their net worth and then some.
As strong as earnings have been, a forensic analysis shows they are being aggressively inflated:
CORRELATION OR CAUSATION?
A snapping jaw has just opened the same way software opened up against semis.
Hyperscalers, historically 90% correlated to semiconductor stocks, have lagged behind to create the widest spread since 2022.
This matters because these have been co-dependent ever since hyperscalers started going big into cloud computing services.
Then AI came into the picture, and that relationship only got stronger to create two separate segments:
Check Writers (hyperscalers)
Check Receivers (semis and memory)
Each side of the capex money flow tells a different story, and their price action is the market’s way to communicate confidence on each side.
I’ve been wrestling with the idea as to why the relationship is breaking, and what that could eventually mean for the S&P 500 and the rest of the AI trade moving forward.
To clarify my confusion, I started to study the history of this trade:
Leaning on the best parameter for these situations, I have found correlations explain most of this break, as well as previous ones.
Rolling annual correlations remain unchanged so far, as the divergence between the two baskets only started in the beginning of June.
If the divergence continues to widen, then here’s what I’d look into:
Whenever Semis lead in the divergence, the outcome tends to be bullish for the S&P 500
Whenever Hypers lead in the divergence, the outcome tends to be bearish for the S&P 500
Because the semis basket is leading in this recent case, I have to lean on the conclusion that the outcome will be bullish for the S&P.
But,
This is also where I want to introduce nuance to the 2026 divergence.
Prior to 2023, capex had very little to do with Semiconductor sales.
Post 2023, AI came into the equation to change this relationship completely.
In the study between NVIDIA sales and Amazon capex, you can see that every 10% increase in capex coming from Amazon results in a ~28% increase in NVIDIA sales.
That is exactly why I say there’s nuance to the most recent divergence between hyperscalers and semiconductor stocks.
Because this time it’s not about the economy…
It’s about weighing the odds between AI succeeding or failing.
The way I see it:
Bidding Chip names is a way to bet on sales and cash flow happening now, as AI demand raises chip prices and commands more revenue
Selling down Hyperscalers is a way to protect capital from an uncertain future return.
In Day Two of our AI trade deep dive, we explain why markets may be looking to cut down this uncertainty risk.
In a nutshell:
Price wars initiated by OpenAI and Anthropic will blur future revenue assumptions
Spending cuts by major corporates significantly lowers these assumptions as well
The end result is the trillion-dollar capex plans made by hyperscalers will not see an ROI as soon as initially expected.
At the same time, chip names have already received that capital and booked revenues.
What happens if the divergence is erased?
Well, this is where it starts to get even more confusing…
OpenAI and Anthropic have delayed their IPOs to at least next year.
Now this is a strict game-theory exercise, not my actual call.
But, I hope it serves you well.
Here are the potential cases:
Hyperscalers rise up to close the gap to chip names
In my opinion, this only happens if the hyperscaler basket chooses to cut down capex spending, increasing their free cash flow and EPS volatility.
Bidding these on that reason would also mean markets are valuing defensive and certain earnings over high-growth and uncertain outcomes.
That’s bearish for the S&P considering how much of it is made by AI-related names.
Semis & Memory come off to close the gap to hyperscalers
That’s a different story, as hyperscalers staying down can mean capex will continue to provide cash flow and earnings uncertainty.
Chip names coming off would mean either semiconductor and memory prices are coming down due to oversupply, or the same AI revenue uncertainty will begin to hit them as well.
Hyperscalers rise, Semis & Memory sell off
Simply put, I think this marks the end of the AI trade, and a major risk-off rotation across the rest of the financial markets.
Why?
It means hyperscaler capex has ended the EPS uncertainty, and then semis & memory trades receive no further funding and new orders due to the missed revenue and productivity targets.
Keep an eye on this spread and its future performance, it’ll save you tons of money.
WHAT’S THE TRADE?
I stand by the analysis I made previously about the best way to play the AI trade.
In a way that makes money whether it succeeds or not.
That’s software, as it is now representing the following setup:
If AI succeeds, software will become the toll booth for revenues to keep justifying the expected $7 trillion capex
If AI fails, then the entire “AI will kill software” narrative dies
So please refer to my Adobe Deep Dive sent to paid members for more information on the software leg of this trade.
Then,
Hyperscalers (select few) also represent a great way to play this out.
Among which:
Microsoft
Meta
Any other that falls into a bear market
For the same reasons in whether AI succeeds or not.
Because successful AI will deliver the ROI necessary to boost profits, margins, and market share.
If AI fails, then free cash flows go back to the pre-AI capex period, and so do valuations.
Easy enough isn’t it?
A Final Note
COMING UP NEXT
As PMI reports come out next week, I will give you a full primer on the industries that are heating up / cooling down respectively.
From it, you will gain a further insight into where the AI trade is affecting the market, and how you can pick some short-term swing plays accordingly.
If space and information allows, members will walk away with a long/short equity trade idea to take on along with the Offside Portfolio.
In the meantime, here’s a podcast I rather enjoyed with a former hedge fund manager. The guy used to run $165 billion in AUM, and just said to avoid the S&P 500 and to sell “all” of your tech stocks, might be worth a listen:
Until next time,
OFFSIDE RESEARCH
Against the Tape, Ahead of the Curve.