The $1.8 Trillion Hiding In AI

Unsplash

The S&P 500 gapped up to a brand new high today, and everybody chasing it is buying the same story.

The story is artificial intelligence.

I spent the weekend hunting and pecking through the research. I found a number sitting underneath this AI run that almost nobody mentions.

It’s $1.8 trillion.

Give me four minutes. I’ll show you what that number is, and why it once ended three companies cold.

$1.8 Trillion Sitting Off The Books

That number lives at the heart of the AI super cycle. It’s off-balance-sheet risk, and that phrase alone should raise the hair on your neck.

The hyperscalers are driving this entire AI move. They’re funding it with something called an SPV.

SPV stands for Special Purpose Vehicle. Keep that term in your head.

Where You’ve Seen This Before

Reach back for the last time SPVs ran wild on Wall Street. Three names leaned on them: Tyco International, Enron, and HealthSouth.

All three turned out to be scams. All three went to zero.

That’s the company these AI hyperscalers are keeping right now.

An SPV hides a spending commitment before a company proves it can monetize. That’s the entire purpose of the structure. It buries the obligation where shareholders can’t see it.

Why They’re Reaching For It

A lot of these AI hyperscalers don’t make a dime, and there’s no real cash flow underneath them. They’re spending enormous money to build out, and they simply aren’t monetizing yet.

So they use off-balance-sheet financing to grab capital breaks, tax breaks, and write-offs. The headline CapEx figures you read understate the real commitment by a mile.

They’re hiding the true spend so they don’t scare their own shareholders.

This is creative accounting, and it tends to attract the Justice Department eventually. When that happens, some of these names get whacked. I don’t yet know which ones.

Your Chart Won’t Save You

I love a clean chart as much as anyone. On these names, a chart is a waste of time.

Accounting games like this never show up on a candle. You have to do the work by hand instead.

Here’s the homework I’d hand you if you sat on my desk:

• Pull up every hyperscaler you own and study the hard numbers behind each name.

• Visit each company’s site and trace exactly how it turns AI spend into profit.

• Flag any name leaning on off-balance-sheet structures to fund the build-out.

That work is tedious. It’s also the only thing that keeps you out of the next Enron.

Once one of these starts to break, you won’t get out clean. That’s precisely how Enron and Tyco went. The exit door slams shut fast.

Not Everybody Is Cooked

I want to be fair about this. Not every hyperscaler is playing this game.

I don’t think Nvidia (NVDA) sits in this bucket, and I don’t think Oracle (ORCL) gets tagged either. The rest of them you vet one by one.

Here’s the rule I’ve leaned on for 39 years. I never buy a stock that doesn’t make money.

That single rule is how I’ve avoided the risk of ruin. Revenue alone doesn’t interest me, because the name of the game is making money.

A company can post a trillion in revenue and still be completely hollow. You go look at the real numbers, or you’re just buying air.

The Takeaway

The crowd is buying the AI headline today. That $1.8 trillion underneath it is the part they’re ignoring.

Do your homework before you own a single one of these names. When the accounting finally catches up, the door gets very small.