Dear Reader,
Nvidia announced something this week that should have worried people.
Instead, the market cheered. On Thursday the semiconductor index jumped better than 3%.
Let me tell you what they actually announced. Then let me tell you why thirty-five years in this business has me reading it differently than the tape did.
What They Actually Said
Nvidia signed memorandums of understanding with six financial firms.
The stated goal is to mobilize more than $500 billion of outside capital to build AI data centers and buy Nvidia hardware. They're calling them “independent compute financing platforms.”
Now read the fine print.
No deals have actually been signed. There is no timetable. And Bloomberg reported that the $500 billion figure is a round number — part transactions already under discussion, part forecast of demand that hasn't shown up yet.
So what was announced was not $500 billion.
What was announced was an intention to go find $500 billion.
Why A Company With 75% Margins Needs A Lending Arm
Here's the part that stopped me.
Nvidia does not need money.
Last quarter it reported record revenue of $81.6 billion, up 85% from a year earlier, at gross margins around 75%. It has authorized another $80 billion of buybacks and raised its dividend twenty-five fold. The market values it around $5.5 trillion.
Companies like that do not arrange financing for their customers.
Unless the customers can't pay.
And that is where the arithmetic gets uncomfortable.
Morgan Stanley puts global data center capital spending through 2028 at around $2.9 trillion. Hyperscaler operating cash flow — all the money Microsoft and Amazon and Google and Meta actually generate — can cover perhaps $1.4 trillion of it.
Which leaves $1.5 trillion.
Stop and think about how strange that is. The largest buildout in the history of American business is a little over half funded by the people doing the building.
In Morgan Stanley's own bridge, the largest single piece of the remainder — about $800 billion — is expected to come from private credit.
Not banks. Not shareholders. Private credit.
I've Seen This Movie, And I Want To Be Careful About It
In 1999 and 2000, the telecom equipment makers did something that looked very smart at the time.
They lent their customers the money to buy their equipment.
It made the revenue look wonderful. Right up until the customers couldn't pay, and the receivable and the revenue went to zero in the same quarter.
Let me be careful here, because this is exactly the kind of comparison people make lazily.
Nvidia is not Lucent. It sells a product that works, to customers who are using it, at margins Lucent never dreamed of.
But analysts have already put a name to the pattern. They call it circular financing — the vendor funding its own demand. Nvidia has invested in CoreWeave, contributed billions to an OpenAI round, and joined a consortium backing xAI.
The company's answer to that criticism was to move the lending off its own books and onto somebody else's.
That solves Nvidia's problem.
It does not solve the $1.5 trillion.
The Fed Chairman Said It Out Loud
Here's the beat almost nobody picked up.
At Wednesday's press conference, after the Fed raised rates for the first time in three years, Chairman Kevin Warsh was asked why long-term interest rates have climbed all year even while the Fed's own policy rate sat still.
He gave three reasons. Economic strength. Geopolitics. And “competition for capital” from hyperscaler capital expenditure.
Read that again.
The Chairman of the Federal Reserve, asked about the ten-year Treasury — which he called “the most important asset anywhere in the world” — named data center construction as one of the three things moving it.
That is not a technology story anymore.
When a buildout gets big enough to show up in the price of money itself, it has stopped being a sector and started being the macro.
Where The Paper Is
I wrote to you on Thursday that in 2000 the real tell wasn't the valuations. It was the paper. Garbage companies coming public. Pets.com.
And I told you that right now the paper in this bull market isn't IPOs. It's debt.
I listed that as a warning sign to watch for. This week it stopped being something to watch for.
It's why I watch private credit spreads more closely than I watch the Nasdaq.
But it's worth noticing what that actually means.
Almost nothing is going public. The money is coming in privately, from institutions, in rooms you and I are not invited into.
Which makes the exception interesting.
There is one company in artificial intelligence — arguably the most important one that isn't already public — that filed confidentially with the SEC on June 1st and is targeting a debut as soon as October.
And while you were locked out of it, the institutions were not. Its last private round targeted $10 billion and drew roughly $30 billion of demand. BlackRock-affiliated funds. Fidelity. Goldman Sachs. JPMorgan. Morgan Stanley. T. Rowe Price. Even the pension fund of Ontario's schoolteachers got a seat at that table.
Federal law says you cannot buy a private company like that unless you are already a millionaire.
So my team spent months on a narrower question: what can a regular investor actually do about it before that October window closes?
I'd get through this one before October, for reasons that will be obvious about two minutes in.
The Bottom Line
I don't know whether that $1.5 trillion gets funded.
I've been doing this thirty-five years and I'll tell you honestly — it might. Private credit has more money right now than it has good places to put it, and this is the most obvious place on earth to put it.
But be clear about what changed this week.
A company earning 75% gross margins volunteered to help its customers borrow.
That is not the behavior of a business that believes its customers are fine.
Watch the paper. Not the price.
And if you want to understand the one AI company the institutions quietly loaded up on while the rest of us were locked out, read this before the October window closes.
Have a wonderful Saturday.
"The Buck Stops Here,"
Found this helpful? Share it with others.
Written by Dylan Jovine