Last week, the Federal Reserve raised

interest rates for the first time in

three years. The decision was unanimous.

The tally was 12 to nothing. But not

even seven weeks earlier, that exact

same room couldn't agree on anything. At

the previous July meeting, the Federal

Open Market Committee's vote came back 9

to3. And not in the direction you'd

think. Nine officials voted to leave

rates alone. Three voted to raise them,

which means 3/4 of the committee was

against the rate hikes back in July. But

by September, not a single person was.

So what changed? The Fed chair, Kevin

Worsh, explained it at the press

conference. The inflation numbers that

came in over the summer showed that

inflation, quote, was not abating on its

own, and monetary policy was not doing

enough to restrain excess demand. In

plain English, inflation won't go away,

and rates aren't high enough to slow

spending, which also gave the bond

market the validation it's been looking

for. The 10-year Treasury, the rate

almost every other interest rate on

Earth gets priced off of, closed above

5%, hitting the highest level since

2007. Since 1965, the Federal Reserve

has run 11 tightening cycles. The

smallest raised rates by 1 and 3/4

percentage points, the biggest 13. The

typical one ran about a year and a half

and added close to 4%. Only one of these

tightening cycles stopped at just one

hike, as in they raised rates once, had

a look around, and said, "That's

enough." It was in March 1997. Alan

Greenspan raised rates a single time,

decided that was plenty, and was back to

cutting them 18 months later, which is

basically what the Fed is telling us

right now. In its own projections, 16 of

18 officials have at least one more hike

pencled in before the end of the year.

The day before the meeting, the

conference board's senior US economist

put out a brief and the title of it was

more inflation coming. The Fed rarely

stops after one move. She's forecasting

three in a row, September, October,

December. Researchers at the New York

Fed went back and counted 14 tightening

cycles between 1955 and 2009. Of those

14 cycles, 10 were followed by a

recession within 18 months of the final

hike. Then of the remaining four, one of

them led to a jump in unemployment and a

credit crunch that plenty of economists,

including Nobel Prize winner Milton

Freriedman, still count as a recession,

which leaves three times where we

achieved the highly coveted soft

landing. Three times in 50 years where

America raised interest rates without

up the economy. So, it's not

necessarily the number that matters

here. It's what one of these cycles does

once it gets going. It's about where

these cycles land in the economy once

they get started. For most of the last

70 years, a rate hike got to you through

a clear door. Mortgage rates went up.

Buying a house cost more, so fewer

people bought one. Refinancing stopped

making sense. Builders slowed down.

People spent less. And the economy

cooled off. Since the end of the Second

World War, housing was the single most

direct lever the Federal Reserve had.

But that lever is breaking. Only about

40% of American households have a

mortgage at all. And of the ones that

do, 78% are locked in below 6%.

Twothirds are below five, half are

sitting under 4%. Which means if you've

got a 4% mortgage and today's rates sit

near seven, you're not even going to

consider selling. The Fed's own report

to Congress in July has a name for this,

a rate lock, which is a fancy way of

saying people looked at today's mortgage

rates, then at the mortgage rate they

already have, and decided they're never

moving again. And you can see this in

the sales data. Existing homes are

selling at a pace of about 4 million a

year. The 15-year average is closer to

5. Last year was the slowest year since

1995, which means the housing market is

already frozen. There isn't much left

that a rate hike can slow down. So, if

the Fed's favorite break lever isn't

working anymore, where do these rate

hikes land? Well, Fed Chair Kevin Walsh

answered that himself 19 days before he

raised rates. He said, "Everexpanding

pools of capital are pouring into AI

related infrastructure of all sorts."

That's the chairman of the Federal

Reserve pointing straight at the data

centers. So, how much money are we

actually talking about? Well, JP Morgan

estimates that the global data center

buildout could cost roughly $5 trillion

through 2030. A finance professor at

Colombia ran the numbers on just the

American side of the AI buildout. If it

gets built the way it's planned, it runs

at roughly 2.8% of everything America

produces every year through 2032. For

context, in the 1800s, building every

railroad in America ran at about 2 and

a.5% of the economy, wiring the entire

country for electricity, about 1%. The

construction of the interstate highway

system about 1 and a half. So this is

the biggest thing America has ever

built. And the tech companies at the

center of this are the most profitable

companies that have ever existed. So the

reasonable assumption is that they're

paying for it themselves. But they're

not. The operating cash flows, selling

stock, and everything they can generate

on their own, all of it together, covers

roughly a quarter of the $5.5 trillion.

JP Morgan expects the other 4 trillion

to be borrowed. So where do you go when

you need to find $4 trillion? If you

were thinking Nancy Pelosy's trading

account, you'd be right. But these

companies decided to look somewhere

else, the bond market. From 2020 through

2024, the five tech companies in the

middle of the AI bubble issued about $35

billion of bonds a year. In 2025, 93

billion. And by the end of July this

year, 7 months into the year, they're

already at 132 billion. A risk advisory

firm called Chadam Financial puts this

year's AI infrastructure spend at more

than $830 billion. It comes out to about

half the volume of the entire American

investment grade bond market. 2/3 of

every leveraged loan and significantly

more than the entire high yield bond

market put together. It's a handful of

companies in one industry trying to fit

through every door in the credit market

at the same time. And the numbers are

only getting bigger. Vanguard has

capital spending at these companies

above a trillion dollars a year over the

next 3 years. So the buildout isn't a

check somebody writes. It's a standing

appointment with the bond market every

year for the rest of the decade. Which

means the price of money isn't a detail

in this story. It's the entire story.

But before I get into that, let's take a

quick pause. This video was made

possible by today's sponsor, Odo. ODU is

an all-in-one management software that

brings your business operations onto one

platform. They offer a range of services

including invoicing, sales, accounting,

marketing, website building, and so much

more. It's all your applications

centralized into one database. And ODO's

applications communicate seamlessly with

each other. So your sales are linked to

your accounting, your inventory, and

your projects. Which means there's no

more double manual entry of data, no

more jumping back and forth between

tools, and no more searching through

different sources of information. ODU is

the source of truth for your whole

business. ODU allows you to focus on the

stuff that actually matters in your

business rather than wasting time on

manual repetitive tasks to keep things

in order. You can use any one of ODO's

applications for free or get the entire

suite starting at under $25 per month.

With more than 28 million users

worldwide, ODO's track record speaks for

itself. So whether you're already

running a business or looking to start

one, you can book a free meeting with an

ODU expert or try the full platform free

for 15 days with no credit card required

using the link below. Thank you to ODU

for sponsoring this video. Now, back to

how one small change in the price of

money could wreck havoc on the biggest

buildout in history. When a company

sells bonds, it pays whatever the

government pays to borrow, plus a little

extra on top. That extra is called the

spread. It's the market's way of saying

how nervous it is about you paying the

money back. This year, the spread on

those five companies widened by about 30

basis points. The rest of the investment

grade market widened by just two, so 15

times the move on some of the most

profitable companies in history. That's

the best of the best in the market. Go

down the stack and it's not as polite.

The debt that funds the actual data

center projects prices about 100 basis

points wider than those companies own

bonds. About 200 wider once you're down

in junk territory. And near the bottom

of the stack, it gets ugly. A few months

ago, Cororeweave, one of the companies

that rents out GPUs to everyone else,

had a $2.6 billion loan out for

syndication. Halfway through, the lender

sent it back. The interest rate

increased by a full point. And then came

the conditions. The loan has to be paid

down in full over its life. No big lump

sum at the end. A coverage test which

says the cash coming in has to beat the

debt payments going out by a set margin.

And a minimum cash balance, which is all

just the long way of saying the lenders

want to make sure there's still money

when it's their turn to get paid. It

ended up clearing at 10.4%.

Over a year, Cororeweave's interest bill

went from $267 million a quarter to 640.

In the second quarter, interest alone

ate 42 cents of every dollar of adjusted

earnings. That's the bottom of the AI

market repricing in real time while the

thing is still being built. Which brings

us to the question nobody in this

industry wants to say out loud. At what

price does this stop being worth doing?

A compute pricing firm called Mercatis

actually modeled it. They ran a single

GPU cluster at different costs of

capital. At 6% it clears comfortably. At

10, it barely clears. At 12, the whole

thing goes underwater. The break even

sits somewhere around 11%. Mercatus also

breaks the buyers down by their cost of

capital. Established companies with

investment grade balance sheets are

paying 6 to 8% on new deals. Late stage

AI companies with strong revenue 10 to

14%. Earlier stage venture-backed AI

companies 15 to 20, meaning the top of

the market clears the break even, the

middle is standing right on the break

even line and the bottom is already

underwater. And this is a price that's

also now rising. A borrowing cost isn't

something you pay once. It's more like

rent. It resets. And it resets two ways.

The first one is fast. A lot of the debt

at the bottom of this market floats.

That corewave loan is priced off an

overnight rate that moves with the Fed

plus 5 1/2 points. Allin that's nearly

10 1/2%. So when the Fed raised rates

last week, that loan got more expensive

before the next payment was even due.

The second one is slow and structural.

The bonds at the top of the market are

fixed, but every loan matures. Every

bond comes due, and when it does, almost

nobody pays it off. They roll it over

and borrow it again at whatever the

market is charging that morning. So, a

company that borrowed at a cheaper rate

a few years ago isn't going to get the

same rate when it hits the markets

looking for new debt. The average

tightening cycle since 1983 added more

than 3 percentage points. 16 of 18 Fed

officials see at least one more rate

hike this year. The Conference Board

sees three. And you can't borrow all

this money at once. This isn't an

industry that's borrowing on a one-time

basis, which means the largest capital

investment in American history is about

to have to start paying more for the

debt it once. This is what makes this

tightening cycle different from everyone

before it. In the past, a hike mostly

landed on the demand side. The

mortgages, the credit card balances, the

car loans, the boring consumer stuff.

This time the biggest impact lands on

the supply side on companies carrying

enormous debt loads trying to build the

future. That's the part people get wrong

about a rate hike. For most borrowers,

it doesn't break anything the day it

lands. It breaks things later on when

somebody has to go back for more money

and that money costs more than it did

last time. We're seeing massive demand

for debt and a rate cycle that's just

getting started running at each other

full speed like they're doing the

Oklahoma drill. And I know what a lot of

you are thinking right now. let the AI

companies burn to the ground. But the

problem is the damage doesn't stay

inside the industry. Apollo's chief

economist has been tracking data center

spending as a share of the economy.

Right now it's growing by about 0.85

percentage points of GDP a year. The

fastest stretch of the housing boom from

2002 to 2005 grew at half a point a

year. The telecom buildout of the '90s

just 1500s of a point. So the AI

buildout is being built roughly twice as

fast as the housing bubble was. And we

all remember how that one ended. AI

isn't just growing fast. It is the

growth in the economy. This year, AI

capex will be a nearly 1.5 percentage

point boost to United States economic

growth. Private data center construction

is now running at 75 billion a year, up

57% in one year, more than doubled in

two. Census Bureau data now shows

America spends more building data

centers than it spends building

conventional office buildings. The good

news is nothing's broken yet. Nobody's

pulling back spending. Nothing has

stopped moving. But the absence of

damage is also not evidence of safety.

JP Morgan went back 20 years and plotted

what the big five tech companies do with

every dollar of revenue. In all that

time, capital spending never took more

than 13. This year, it will take 41,

which is almost every cent the

businesses bring in. For the first time

on the chart, free cash flow goes

negative, which means every year going

forward, the amount that has to be

financed by debt goes up and the rate it

gets financed at goes up with it. That's

where the dots actually connect for this

recent rate hike decision. It's the kind

of detail that often gets left out of

traditional media because it doesn't fit

into an easy to package, easy to sell

headline. So, if you find this kind of

thing interesting, hit subscribe because

you're probably going to keep hearing

the same story about the Fed raising

interest rates by a quarter of a point.

But that's not the real story. The real

story is that the largest buildout in

history, the one propping up the

economy, still has to keep going back to

the market for more money for the

foreseeable future. And every one of

those trips is going to cost more than

the last one.