coreweave's crash scares the sh*t out of me

Good old Cororeweave, the company that

Nvidia put a floor price on at $40,

saying they would buy up to $250 million

of shares of Cororeweave anytime the

stock trades under $40. Is having a bad

day. At the time of this recording, it's

down 17% on the day after earnings. Uh,

and it's losing support after support

after support. Uh and in this video,

we've got to talk about why we would see

this sort of euphoric rally post

liberation in April uh to now what's

turning out to be a bleed out confirmed

by earnings that are likely to continue

this bleed out. We need to understand

why this is happening. First, it's

important to remember the cycle here.

Why would Nvidia promote investing in

Corewave? Well, it's pretty simple. See,

Nvidia makes chips. We know that. Well,

they don't actually manufacture them.

They just design them. Uh, better hope

that China doesn't take over Taiwan

because 90% of our chip manufacturing

would evaporate if that happened. But

anyway, yesterday I got myself one of

these, the Nvidia 5090, and I realized

that the streaming computer I had been

using, and actually still am using cuz

I'm still setting up the 5090, uh, it

was using a 4090. And I'll have you

know, that's 2-year-old technology now.

So, you know, when it came to using AI

processing for a green screen, uh, which

I rarely use, but when I do, it just

isn't keeping up anymore. So, like the

software is exceeding the technological

capabilities of the chips we had

essentially 2 years ago. So I find that

very interesting uh that we're kind of

already on this upgrade cycle and that

becomes really important because see as

a customer of Nvidia uh owning I don't

know maybe 5 4090 chips uh which we now

use for uh AI um uh in-house sort of ML

training that we do at night over at

house hack which which is great because

we don't have a rush for it very you

know they call it latency insensitive uh

training that we can do but anyway What

we find is that the technology gets old

and when it gets old, I'm not mad at

Nvidia. I bought the 4090 knowing it's

going to get old at some point and I'll

put it to use on something that doesn't

need instantaneous uh you know, pressure

uh of of workflow. Uh and I'll go buy a

new one. And so now here I am with the

5090 that I know is basically going to

be worthless in 2 years. Maybe not worth

less, but it'll be worth a lot less in

two years. and I'll probably be

worthless in four years and I'll still

love Nvidia. I'll still come back and go

buy the 6090 or the, you know, 7090 67,

you know, something like that. I'll

still do that and I'll still love Nvidia

even though their stuff becomes

worthless after 2 years or 3 years,

right? Uh, and and you know, don't mind

me complaining about the fact that the

chip, which is basically the thickness

of this, is pretty much like tiny and

the rest is all a heat sink and fans.

Kind of crazy.

That said, Cororeweave buys this junk

and they hold it. They hodddle these

chips and that becomes really important

because they support Nvidia by buying

the Nvidia stuff, the SMCI server rack

uh buildouts, the AMD processors. They

buy this stuff. All those downstreams

win. Verive with the water cooling for

Blackwell, they all win immediately. But

Coree holds the bag. Okay, that's the

old news. We know that longer term. So

what's happening now that's making us a

little bit more curious about coreweave

now that we've got sort of some of the

basics out of the way about how this

functions with coreweave and chips.

Okay, this is going to get into some of

the nitty-gritty but there's a lot that

we have to understand. So the first

thing that I'd like to do is I want you

to know that they borrow money at

expensive rates. This is their earnings

call from the earnings that just came

out and they talk about three financing

deals that they just participated in. Of

the three financing deals, two of the

deals were in the quote high yield

space, which is a fancy way of saying

junk debt. And then their third deal,

which they were cheering about, was

sofur plus 400, which means they're

probably paying sofur plus,300

on the other deals. Okay, that's really,

really high. Now, what is sofur? Well,

you could Google what sofur is today,

the secured overnight financing rate.

You don't really have to know exactly

what all of that means, but basically

it's just a floating rate that changes

every single day. Today, sofur is 4.36.

So, if you're paying sofur plus 400,

you're paying 8.36% interest. If you're

paying sofur plus,300,

uh you are actually paying uh well, in

this case, well, uh sorry, plus 1300.

Yeah, you're paying like 17% interest,

which is insane. Crazy junk bond level

debt. Uh so why would lenders be

charging such exorbitant levels of

interest on coreweave debt? Well,

because the lenders are actually

realizing what I just told you about the

5090 and the 4090 chips.

Lenders realize that the underlying

asset will be worthless at some point.

Now it it won't necessarily be 2 years.

Okay, maybe that's extreme. Like the

4090 is still good today, right? But

let's be real. In 5 years from when

these chips come out, 5 to 10 years,

call it, they're almost certainly going

to be worthless. This is very different

from buying real estate. So, for

example, my real estate startup, House

Hack, probably ending that fund raise

within the next month or so. Uh

househack.com. Read the paperwork over

there if you want. But with house hack,

we buy houses, fixer uppers and

apartment buildings and we develop

properties and accessory dwelling units

in and and we expect in these

highquality areas these assets are going

to maintain their value for 30 to 100

years, right? Because we maintain them

every month. We maintain them. We have a

maintenance expense and we keep them up.

And because they're in locations people

want to live and there's a lack of

supply, we uh uh you know, the value of

these assets goes up over time. Chips,

on the other hand, become worthless over

5 to 10 years. And that's why you're

seeing the actual lenders

be pretty aggressive in the financing

rates they have against Cororeweave.

This is a problem. Cororeef is basically

also a commodity provider. And in

providing this commodity, the best thing

you can do if you're a provider of a

commodity is have committed reserved

data usage. So if you have 10 data

centers, you want to know like imagine

these 10 data centers are like 10 rental

properties, you want 100% occupancy,

right? You want all of your data centers

firing on all cylinders. But what

happens when you have your 10th data

center, let's say, so 10% of your data

centers like, oh, uh, you know, we don't

have tenants right now. Okay. Well, when

you don't have tenants that are going to

reserve spacing, you might consider

throwing it into the hotel market for

chips, the Airbnb for chips, we call

this the ondemand and spot market. In

other words, if I want an H100 right now

for one hour, what is it going to cost

me?

Somewhere around $2.40 an hour is what

it's going to cost you. So for $2.40 an

hour, you too can have an H100

processing your crap for an hour. All

right? And you pay $2.40.

Now, why does that matter? Well, it

matters for a few reasons.

spot pricing or this pricing for

ondemand

uh servers uh specifically H100s has

been plummeting.

Take a look at this. This is the spot

market for an H100 based on

hyperstack.cloud/GPU

pricing. their H100, which came out

October of 22 and really broadly came

into use in the first and second quarter

of 2023 through the artificial

intelligence revolution and then later

in 2023 through uh Elon Musk and XAI

purchasing a ton of these chips. You

really saw the buildout of the H100s in

2023 and these chips are coming up on 2

years old right now. So, we're two years

old on the H100 cycle. And what are we

seeing after two years on H100s? Well,

let's go back one year. So, when the

chips were one year old, HyperStack was

selling them for $3.40 an hour, 44 an

hour. Okay. Reservations, which would be

your long-term tenants, only $2 an hour,

$26.

But on demand per hour, 344. Okay, that

was when the chips were about, call it

one year old. Then they turned into

little 18-month-old babies. This is

using the Wayback machine. So, sorry the

rendering is a little funky here. But

the way back machine shows us that the

GPU pricing on the ondemand side, which

is your excess capacity side, fell from

$344

an hour uh down to $3 an hour, which is

about a 12.8%

decline in pricing from when it was one

year old to 18 months old. Then when we

go to pricing today, we can see the

ondemand. This is very important.

Ondemand pricing per hour is down at

$2.40

per hour. $2.40 per hour after one year

represents a pricing decline of 30%.

It's about 30.2% decline in pricing per

hour. Now, this ondemand side is

important because the ondemand side is

different from your reservation side.

the reservation pricing hasn't moved

that much, but that's because these are

people willing to make long-term

commitments. The excess capacity, so the

place you're going to see the first set

of damage on pricing is always going to

be on your vacant units, your ondemand

stuff. And what's depressing here is not

only do we have this 31% decline in

pricing over one year on the excess

demand or excess supply product, I

should say, right? There's excess

supply, so lack of demand. We have

vacancy which is represents a little bit

of a lack of demand on on demand.

Um that's wordy I understand but that's

that's a pretty big decline. And look at

what Cororeweave is telling you.

Coreeave is telling you hey uh we want

to expand our footprint into ondemand.

Oh why do we want to do that? Because we

think it's important.

What?

That's it. Like hey why do you want to

do that? because I want to. Wait, that's

the CEO's answer. Are we serious? Hold

on, Kevin. Are you reading this right?

Because the CEO, chief executive officer

says, "And we really want to be able to

continue to expand our footprint there

around on demand because we think it's

important."

Oh, okay. So, you want to expand your

footprint into the Airbnb of chips where

pricing just declined 31%

and you know when the age of a chip went

from 1 to two. Uh that's where you want

to expand your footprint. Well, how much

utilization are we going to have for

these chips when these chips are 5 years

old? Right? What you want is you want

more 5-year lease commitments. See,

OpenAI in 2023, there was such a

shortage for these chips. They gave

Coreweave a 5-year commitment and this

led to a lot of excitement. People are

like, "Oh my gosh, OpenAI is going to

promise to to lease these these chips

for 5 years." But the thing is, they've

already walked back this talk about

5-year lease commitments. And now

they're only willing to say, "We're

signing multi-year

contracts for Blackwell." They're not

telling you they're doing five-year

contracts for Blackwell. They're telling

you they're doing multi-year, which

multi-year is anything more than one.

So, could be two-year contracts, could

be three-year contracts. And the reason

you're probably seeing this is because

if you're Microsoft, why would you go or

open AAI, why would you go to Coreweave

today as there's more supply of these

data centers? Why would you go to them

and say, "Hey, we want to pay a top tier

price for five years when on the margin

pricing is plummeting for these per hour

servers

because we're starting to see supply

catch up.

On top of that, and uh this this was a

very interesting observation that

somebody came up with here. This user,

James Woodman, said, "There's something

very interesting going on about accounts

receivable growing 1.5 billion in the

first half of 2025. This revenue growth,

uh, uh, this revenue backlog is growing

materially with an increase in the

amount of short-term buckets. Uh,

Cororeweave is basically making a bet

that the companies are going to be able

to monetize this compute profitably,

otherwise they can't pay their bills."

Now, this was very interesting. And so

then what I did is I went into the

financials for coreweave and I wanted to

corroborate what this user was saying.

And so

really as I change my zoom my

highlighting goes away. That's a scam.

Uh okay.

No problem. I remember.

Uh shouldn't have done that. Anyway,

look at the accounts receivable right

here. So accounts receivable for

Coreweave just Oh, there we go. My zoom

came back. Accounts receivable for

Coreweave exploded to 1.9 billion.

That's 4x. It's actually 4.5x

explosion in accounts receivable. Well,

how much did their plant, property, and

equipment go up? Uh they went from 11.9

to 16.6,

which is an increase of about 40%. 39ish

and some change. So, their plant

property and equipment, their actual

infrastructure grew by 40%. But their

billings grew by 4.5%. Now, there are

two ways you could look at that. You

could look at that and go, "Hey, like

that's great. That means they're making

more money, right? Or it means they're

just not getting all their payments as

quickly as they previously had. So,

let's try to figure that out. Their

revenue in the first 3 months uh ending

or the three months ending June 30th was

about 400 million in 24." Okay, 400

million in 24 about 100% of that was uh

receivable. Oh, wait, hold on. That's

not a comparison. I can't make that

comparison. Darn it. This shows me the 3

months ending 2024 on June 30th, but

this over here shows me the end of the

year. Uh, bummer. But anyway,

all right. So, to the end of the year,

uh, that's a little less clear. But

anyway, the point is

the receivables right now exceed our

revenue substantially. Our revenue right

now is about 1.2 billion

and receivables are almost twice that at

1.9 billion. At the end of 2024, we had

about 400 mil of receivables. Uh and we

don't know revenue was was some some

level of this. Uh but the point is they

were today a lot of their money is

hopefully coming through. Now we don't

actually think that they're going to

default uh or like these people that owe

Cororee a lot of money are on the edge

of bankruptcy and they're not going to

pay their bills. So I'm not that

skeptical that these this cash won't

come through, but I know some people are

and it's just sort of something to keep

in the back of the mind and pay

attention to. I don't see it as a

massive red flag or issue right now, but

it is interesting to see a lot more of

that cash that should be showing up,

maybe should be discounted a little bit.

And the reason this is important is

because Coref has a lot of current debt.

Look right here, current liabilities.

And I'm going to remove deferred revenue

because deferred revenues we'll talk

about in just a moment.

If you look at deferred revenues, you've

got $951 million of deferred revenues.

Take that off current liabilities.

You've got about $650 or sorry, $6.5

billion of debt due in the next year.

So, think about this for a moment. Let's

say you had $650

worth of bills due within the next 12

months. Okay? So, you owe $650, which is

actually 650

uh 6. It's actually $6.5 billion for

Core Weave, but for you, it's $650. You

owe $650 over the next year. Okay, cool.

What's your money coming in? Well,

you've got $115 in cash and you've got

$200 of money supposedly coming in.

So, if I add that together and I see I

assume all of my receivables are going

to show up, 200 plus 115, I've got $315

and I need to pay $650 worth of bills

over the next year.

That's problematic. I don't have enough

money to pay for my bills over the next

year. And that's one of the problems

over here. And that's why I started by

talking out about some of the debt and

the high interest that they're

financing, high interest rate debt that

they're financing. One of the reasons

they have to finance debt at such high

interest rates is because they don't

have the money to pay their bills.

So they have an asset, this plant

property and equipment side that they're

only depreciating at, you know,

somewhere around uh $500 million a

quarter. So $2 billion a year. So

they're we'll talk about depreciation in

just a moment as well. But they're

telling us it's not really becoming

worthless that quickly. Don't worry,

everything's fine. We're going to keep

borrowing against this $16.6 billion,

but we don't have enough money to pay

for our bills over the next year. So,

we're going to go out and we're going to

keep borrowing money at really high

interest rates because the lenders

realize that we're borrowing money on

assets that are probably going to be

worthless by year five, which is not

good.

So, that's problematic. Now, if we look

at the cash flow statement for the

company, this is where we could see how

much they're depreciating. They're

depreciating about $500 million every 3

months. That'll be important in just a

moment. And they blew about $3.9 billion

in capital expenditures in the last 6

months. So they're burning about $2

billion in capex every quarter. And so

it makes sense that their revenue is

going up because what they're doing is

they're basically buying more rental

property, right? So revenue growth

totally reasonable if you're borrowing a

bunch of money and you're deploying more

rental property essentially, you know,

server stacks. uh, of course your

revenue is going to go up, but we don't

really care about that. What we care

about is, hey, what's the underlying

profitability of this business? Like,

how much money is this business actually

going to make and what's the return on

investment that they're getting? All

right, so this is where I made some

return on investment calculations. And I

find this very interesting because what

I wanted to do was find out how much of

a return is this company actually making

on their assets. Well, here we go. Right

now, we're making about $1.2 billion of

revenue on 6

uh uh 16.6 billion of assets, right? So

1.2 billion per quarter on 16.6 billion

in assets. Uh we have operating net

income of about uh $500 million

uh per quarter. And that's because we

have $700 million of costs per quarter

on 1.2 billion of revenue. Now, where

can we see that? Right here. We look

right here. Knowing from their cash flow

statement that expenses are about $500

million for depreciation,

we can say that let's take out of the

opex off of the 1.2500

and we end up with $700 million of cost

of goods sold, which means we really

have an operating income of about $500

million equal roughly to their

depreciation. So their net operating

return on $16.6 billion of assets is

$500 million per quarter. That works out

to $2 billion per year. So imagine you

had $16,600

and you were getting paid $2,000 of a

return per year on that. Okay? So you

get a $2,000 return on 16.6. That works

out to about a 12% return. Sounds pretty

good, right? That's before interest. So

we're not even considering interest on

the debt right here. But the problem is

of that $16.6 6 billion

we are losing about 15th to depreciation

per year. So 16.6 divided by uh or or

just times2 is 3.3. So, we have $2

billion of operating income, not

including interest

less a loss of $3.3 billion in chip

value, which means we are actually

losing $1.3 billion per year on this

business on $16.6 billion of assets.

That works out to a 7.8% return on

assets. So, I want you to think about

that bottom line for a moment.

Considering a 5-year worthless time

frame for chips and $700 million

operating profit per quarter with 500

mil of uh depreciation

already factored uh in, you know, so 700

uh does not include the 500 mil to be

clear, right? you are losing this

company is returning negative 7.8%

on their chip investments at this point

uh of $16.6 billion per year.

Uh and it assumes the chips will create

uh create val will still create value

in year five. But again, when we factor

in what's happening in the on demand

pricing and uh what's you know the life

cycle of chips, this is a pretty fat

assumption.

These chips might be totally worthless

by year five. So this is a problem. It's

a pretty big problem. So you have a

negative return. Mind you, this negative

return is occurring without considering

interest.

So subtract uh interest of you know what

are they paying in interest right now?

They are paying

oh my lord a quarter of a billion dollar

in interest in 3 months. Subtract

interest of $1 billion per year and

uh this is a money burning

uh bag holder.

Now, in the short term, it doesn't look

like a money losing bag holder, right?

In the short term, like when this IPOed

and was $40, I'm like, "Hey, in the

short term, this will actually probably

do very well because people aren't going

to think about how in the long term the

stock's going to burn a lot of money."

So, in the short term, it made sense why

it was doing so well. Uh, and people

just look at, oh, the revenue is

exploding. This is great. But when we

actually look at some of these these

deeper fundamentals, it's actually quite

concerning. On top of that, take a look

at this.

Somebody here writes, "In their last

earnings presentation, Cororeweave

mistakenly revealed it's running a Ponzi

scheme. You could see it in their slide.

Cororee collects prepayments for a

contract." So now this really

complicates

the analysis because a lot of the

analysis that we just did assumes that

the revenue that they're recognizing is

revenue that they're able to keep

getting. But what if some of the revenue

that we actually just factored in is

really just a prepayment,

you know? So in other words, if I have a

three-year contract and I got paid for a

year up front, that means the revenue is

actually way higher that they're showing

me right now than it actually is. If

revenue

includes prepayments

uh substantially

uh then uh this business uh will be even

less profitable or I should say we'll

lose even more money lose even more

money uh in the future as prepayment uh

revenue does not equal recurring revenue

right it's revenue that looks good today

but it's it's here today gone tomorrow

kind of

which is not good. Uh on top of that,

you know, you've uh you've got Jim

Kramer who says Cororeweave is good. You

I probably should have just started the

video with with that, but anyway, this

is something to think about like where

do we how do we feel about these 5-year

depreciation time frames? How do we feel

about the value of these chips in 5

years when on demand pricing is already

plummeting to the tune of 31% in the

span of you know the chip being 1 year

old to 2 years old? How do we feel about

burning a billion dollars a year in

interest and basically being treated as

a junk bond company by lenders

or or you know a junk debt company I

should call it? uh how do we feel about

their cost of goods sold sitting at $700

million once we take out depreciation

meaning they've got uh you know profits

of about $2 billion a year quote unquote

profits. Uh but you know when we extract

depreciation

or you know the money we're losing on

these chips on $16.6 $6 billion, we're

actually returning a negative. Uh

because these are not assets that are

going to be worth much after, you know,

five certainly to 10 years. I don't

know. I think that's probably why you've

got a CEO that's starting to argue, hey,

we're repurposing our old chips with new

contracts,

uh because they have to at, you know, or

we're moving into on demand. It's kind

of an early red flag that they're having

to reposition

these assets because people don't want

the H100 anymore. The the people with

the capacity to pay the most money today

don't want to pay for an H100. They want

to buy the Blackwell chip, but the

Blackwell chip will also be dated, you

know, in two or three years.

Kind of crazy. So, uh, that's my take on

what just happened at Coreweave. It's a

lot to digest and, uh, some of my

assumptions could be wrong. So, uh, you

know, this is this is I showed you

everything where I got my information

from and my opinions from. So, you know,

Cororeweave, don't sue me, bro. But, um,

if I've made a mistake anywhere, I'd

love to hear from you in the comments.

Uh, but, uh, someone says, "I wish I

could repurpose my chips." Yeah, I'm

kidding. Uh, but yeah, I'd love to uh

love to hear what you all have to say.

So, leave me a comment on that. Uh,

Cororey analysis.

>> Why not advertise these things that you

told us here? I feel like nobody else

knows about this.

>> We'll we'll try a little advertising and

see how it goes.

>> Congratulations, man. You have done so

much. People love you. People look up to

you.

>> Kevin Praath there, financial analyst

and YouTuber. Meet Kevin. Always great

to get your take.