Imagine you're getting off a crowded

subway car in New York City in the

summer. It's hot. Everyone's crammed

together. And all of a sudden, a woman

falls right on you. Catch her. She

apologizes. And you get off the subway.

And just as you reach in your pocket to

find your fair card, you find out your

wallet's missing. You got pickpocketed.

This is classic misdirection. It's look

over here while I steal your wallet.

That's just what Jensen did during the

Nvidia earnings call today. Nvidia is

building the minivan.

First off, the numbers. Yes, they

crushed it by a wide margin. Q4 revenue

68.1 billion. That's up 73%

year-over-year. Insane. Fiscal 2026.

Total revenue for the year 215.9 billion

with data center segment bringing in

62.3 billion alone in Q4. Those are

dream numbers for Nvidia. Then the stuff

that they kind of minimized that you're

going to find in the 10K that they

didn't talk too much about on the call

until questioners asked them. Gross

margin down from 75% to 71.1%.

Operating expenses way up. They're up

41%. Inventories doubled and accounts

receivable is ballooning. I look at this

10K and I see that revenue is

accelerating. The cost of doing business

is way, way, way up. Let's get through

some predictions of what that stock is

going to do tomorrow and then we'll get

into how this fits into our thesis.

Based on options, we're looking at a

5.8% move. That's about 1135 a share. We

got 24.3 billion shares outstanding at

195.56.

That's a swing of 285 billion, which is

almost exactly what we talked about in

yesterday evening's video. Even if we

see some moderation there, we're still

looking at about 140 $150 billion swing.

They had a very strong guide on the call

and I think that this stock is going to

rip tomorrow. That being said, nothing

I'm about to show you from the 10K

changes at all. All of this is still in

play. Here is the dead tell that we are

in the late stage of a bubble. When

growth is effortless, you're getting new

business every day. You do not pre-by

the future because that limits your

options. It locks you into a couple of

eventualities as opposed to having free

autonomy every year. But Nvidia is

starting to price in the future and

they're starting to price it in for a

long time. We're talking massive forward

engineering commitments, multi-year

cloud service obligations, data center

leases stretching till 2030. This is the

one of the reasons that their expenses

are up. They are taking a risk there to

preserve their future. You'll find this

in all latestage booms from the railroad

to the.com boom. All of a sudden you

need prepayments, guarantees, contracts

moving forward well into the future

because you can't be sure that that

business is going to stick with you. So

you get you start to get a little

nervous and you want to lock these

things down for the next 5 10 years to

make sure that you can still grow that

other companies are contractually

obligated to work with you that you have

stuff that you can sell into the future

by contract. And that's exactly what

this 10K shows. On top of that, a lot of

their customers can't finance the

buildout. So, Nvidia is funding them,

which is circular financing. I don't

care what Jensen says about this. This

is circular financing, plain and simple.

We've gone over it in all of the videos

on this channel going all the way back

to last fall when they said I was a nut

job for calling AI a bubble. They have

billions invested in private AI

companies. These are companies that have

not proved that they are profitable,

that they even have a profit model.

Nvidia has invested billions in them.

They have strategic deep partnerships

everywhere at this point. Land, power

and infrastructure guarantees, open AI

investment discussions and then of

course there was the acquisition of

Grock for $13 billion last year. So you

have the platform leader becoming the

capital allocator for the entire sector.

That's not some kind of early cycle

dominance or come up story. This is what

it looks like when we start to stabilize

in that late stage of a boom. It's kind

of like you're out there surfing. You're

riding the wave and that's how the

company gets started, but then the waves

die down. But at that point, you have

enough money, you just pay somebody to

go out there with a power boat to drive

real fast to make fake waves, but you're

paying for it. You're making the waves.

They have been kneecapped so bad by the

China deal that it is hilarious and

painful at the same time to watch them

try to downplay it. In the call and in

the documents, they say that they're

expecting zero revenue from China next

quarter at least. So their outlook

explicitly excludes China compute, which

I think is a wise decision, but it goes

to show how bad that situation is

legally for them. They cannot do it. If

they thought there was any world where

they could get something going with

China, they would have written it down

to make growth look better. And this

creates really a double-edged sword for

Nvidia where if they're selling sort of

gimped chips to China like they were

allowed because of the export controls

they were allowed to sell these lower

powered versions of their GPUs to China

in limited quantities and with a bunch

of restrictions. At least when they were

doing that they could guarantee that

China is running off of their hardware.

even if they are working with companies

like DeepSeek to unlock things that they

don't unlock for American companies and

use this to uh train defense robots

which we talked about a couple of weeks

ago on the channel. Barring any of that,

it's dangerous for Nvidia because now

China will have to really double down on

their efforts to make competitive GPUs

to make competitive compute to Nvidia.

So, it's created a massive massive need

for them. They were already working on

this stuff and making good headway, but

now with the Nvidia faucet turned off,

they're going to need to find water on

their own. So, one growth lane is

permanently impaired. Maybe will be un

unimpaired, but they're planning for it

to be permanently impaired. So, the

money is going to have to come from a

new vertical, and that new vertical is

the Vera Rubin compute. In the intro, I

was talking about being distracted,

getting mugged. This is the woman

falling on you when you come out of the

subway. This is the Vera Rubin compute.

They wanted you to be misdirected to

catch the woman falling while somebody

grabbed the wallet out of your back

pocket. And here's how I know this.

There was a CNBC exclusive first look

time to drop right before earnings. This

is a big spec reveal of what they've

been building, how it's working, and how

that's going to save their customers a

bunch of money. The numbers on that 1.3

million components 72 Rubin GPUs plus 36

Vera CPUs per rack 100% liquid cooled

two tons per rack estimated 3.5 to 4

billion per unit. So they sort of set

this story up an hour or two before

earnings and then Jensen comes in with

the spike and the spike is a new KPI

Jensen keeps pushing uh PR department

definitely came up with this and that's

tokens per watt. tokens per watt. You're

going to hear that so much from Jensen

in the next few weeks in the media.

Tokens per watt. That is the new KPI.

And this is another sign that the bubble

is about to collapse. They apparently

have a myopic focus now on tokens per

watt. And like I've said before in these

videos, these people do not care about

the environment. They don't care about

the environment. They care that energy,

that electricity is expensive.

Electricity is expensive. And so when

they can cut the electricity spend or

get more tokens, AI tokens per watt,

then they take that extra income off the

top. So I heard Jensen pushing that

pretty heavily in the call. And with

this new Reuben setup, it claims two

times Blackwell power draw, but 10x

performance per watt and i.e. 10x lower

cost per token. So any software engineer

watching this, you're going to notice

something that's pretty familiar in our

industry, which is when you're at a

small startup, you're building, you're

experimenting. Let's go boil the ocean.

Let's go solve this problem. And then

once the company becomes bloated,

inefficient, this kind of giga corp,

what's it focused on? It's focused on

efficiency. Efficiency becomes the

topline initiative for everything

implicitly or explicitly even. And

that's exactly what's happening to

Nvidia because they've finally declared

that, hey, we are in the late stage of

the bubble pop. That huge constraint is

electricity. We're going to be doing an

episode on that. Make sure you're

subscribed. Click the bell to be

notified. Dropping next week. The other

thing that Jensen could not help himself

from saying over and over and over again

during the call was CUDA, which is

misguided at this point. I'm not sure if

he recognizes it yet. As I showed you in

the Jensen expose, he bet on CUDA over a

decade before it was successful. He had

to defend his baby from shareholders in

these or earnings calls saying like,

"No, no, no, it's worthwhile. We need to

keep putting millions of dollars into it

despite the fact we have no clear profit

model uh for a return on investment."

And finally, it's it's made Nvidia

Nvidia. That's why we're talking about

it today and they're so successful in

the AI space is because that could have

bet paid off. So, he seems to have an

emotional attachment for it. And I'm I'm

not sure if he realizes that this is

actually going to be a big arrow in the

side for the company. You see, CUDA

opens the chip up to be general purpose.

It's like having a port. Anything can

plug into it. You can do different

compute on it and establish them in the

AI game. But as you go on, anyone in

product and engineering will tell you

this. If you want something lethally

efficient at one task, it must be

specialized and purposedesigned. Okay?

If you drive a minivan, you can haul

decent amount of stuff around with it.

You can put some 2x4s back there. You

can move the kids around with it. If

you're just driving on your own, it

works fine to just drive on your own.

It'll go highway speeds and it's got air

conditioning. But if you want to get

there fast, you're going to buy a

Corvette. The Corvette, you can't fit

the kids. You can't carry any 2x4s or

anything big in it, but it's fast. It'll

get you there way faster than the

minivan. And so Nvidia is building the

minivan, and they've doubled down on the

minivan, which is CUDA. Other companies

are starting to make really good sports

cars though. You have Talas, which is

out of Canada. They're taking a

hardwired model approach. They have

insane tokens per second. It's a It's an

absolute drag racer. Nobody knows where

that's going to go, but probably

somewhere incredible and disruptive.

Then you have Cabris who's working with

OpenAI. They have wafer scale

architecture. We covered those on the

channel a few days ago. Really, really

fast inference with uh what Open AI has

deployed it with with their 5.3 Codeexc

Spark model. I gave that a try. It is

ridiculously fast. So, CUDA is indexing

on the minivan approach while people are

building not only sports cars, but like

amphibious vehicles. They're building uh

big off-road trucks, like these highly

specialized versions of what Nvidia has

initially built that do better at those

specialized use cases. You might say,

"But Dr. J, free cash flow is strong. I

looked at the 10K. Free cash flow looks

good." But what you missed and what they

tried to glaze over a little bit in the

call is you have 40 billion plus of

share repurchases. You also have 17

billion plus into private equities as

somebody astutely called out in the

question section of that call. There's

the Gro deal. There's the inventory

backup and inventory build. So revenue

is up 65% but their balance sheet

complexity is way up. More capital

intensity, more obligations,

geopolitical exposure through the China

thing and of course energy constraints.

A bubble does not pop when revenue

drops. That's a common misconception. It

pops when growth requires more

scaffolding. So demand is real.

Infrastructure buildout is massive. It

seems like they have their tech well

under control. But as the complexity of

that system grows, you just run into a

bajillion headaches. Headaches you

wouldn't even think of, like supply

triage and geopolitical implications.

What if you need some stuff from China?

What is the current administration going

to tax it at? Everyone wants their cut

of what you're doing at this point. So,

it's not collapse yet. It's definitely

not fraud. This is a regular amount of

spin to have on an earnings call, even

though I keep ribbing them over it. This

is about the regular amount of marketing

you'd have on on an earnings call to

minimize the bad things and prop up the

good numbers. But overall, this is uh

this is a sane and not fraudulent 10K as

far as I'm concerned. But the easy phase

is over. The easy phase is over. If

Nvidia is underwriting the AI economy,

the real question isn't how many tokens

per watt, it's who underwrites Nvidia.

If you want the numbers behind the AI

bubble, please sign up for the mailing

list that's linked in the description

and subscribe to be notified when our

video on the energy infrastructure comes

out that's fueling the entire AI boom.