January 23, 2026 • 34 mins
In the final part of this week’s four-part Better Offline Enshittifinancial Crisis special, Ed Zitron walks you through how ignoring the horrifying margins and destructive practices of the AI industry is setting up venture capital and the global markets for a catastrophe, and how all of this could’ve been avoided.
This series was a ton of work, so please support me by subscribing to my premium newsletter - here’s $10 off your first year of annual https://edzitronswheresyouredatghostio.outpost.pub/public/promo-subscription/84rt762qen
Read along with the newsletter version for links! https://www.wheresyoured.at/the-enshittifinancial-crisis/
YOU CAN NOW BUY BETTER OFFLINE MERCH! Go to https://cottonbureau.com/people/better-offline and use code FREE99 for free shipping on orders of $99 or more.
---
LINKS: https://www.tinyurl.com/betterofflinelinks
Newsletter: https://www.wheresyoured.at/
Reddit: https://www.reddit.com/r/BetterOffline/
Discord: chat.wheresyoured.at
Ed's Socials:
https://www.instagram.com/edzitron
https://bsky.app/profile/edzitron.com
https://www.threads.net/@edzitron
Email Me: ez@betteroffline.com
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:03):
Media. Hello and welcome back to Better Offline. I'm your
host ed ZiT Try. This is the last part of
our four part series about how the financial system and
(00:23):
the financial markets are the last victims of the inshitification process.
Over the past four episodes, in a twenty thousand or
so word newsletter which this series is based on, I've
explained how and in this final episode, I'm going to
explain how all this was going to fall apart and
make it clear where I'm so fucking alarmed. Now let's
(00:43):
get back to those data centers. We've already got some
signs of concern within the banking world around them, and
I'm kind of worried, and so are they about their exposure.
In November, the ft reported the Deutsche Bank, which back
cor we've multiple times and several data centers, was the
exploring ways the hedge it's exposure to the AI data
centers after extending billions of dollars in debt, including shortening
(01:06):
a basket of AI related stocks or buying default protection
on some of its debt using synthetic risk transfers, which
are when a bank sells the full or partial credit
risk of a loan or loans to another bank or
keeping the loans on their book paying a monthly fee
to investors. This is a simplification. Please don't be mad
at me. In December, Fortune reported the Morgan Stanley three
(01:28):
times with core Weave IPI partners, Hyperions, Soft Banks Bridge
Loan was also considering synthetic risk transfers on loans to
businesses involved in AI infrastructure. One of what those were
for now. Back in April, SMBC sold synthetic risk transfers.
Back in April, SMBC sold synthetic risk transfers tied to
(01:48):
private debt BDCs. And while this predates the larger data
center deals done by Blue Al, who is a BDC themselves,
smbc's oversee multiple blue ol deals in the past. In December,
s MBC closed another SRT selling off risk from Australian
and Asian project finance loans, though we can't confirm if
any of those were data center related now. In December,
(02:10):
Goldman Sachs pause the planned mortgage bond sale for data
Center Operate A Cyrus one, with the intent to revive
it in the first quarter of twenty twenty six. I
have heard nothing from that since Oracle's credit risk reaches
sixteen year high in the middle of December, with credit
default swaps basically betting the Oracle will default on its debt,
which is unlikely yet no longer impossible, climbing to their
(02:31):
highest price since the Great Financial Crisis. While Morgan Stanley
and Deutsche Banks SRTs are yet to close, it's still
notable that two of the largest players in data center
financing feel the need to hedge their bets. So what
exactly are they hedging against? Simple that tenants won't arrive
and debts won't get paid. I also believe that they're
(02:51):
going to need bigger hedges because I don't think there's
enough actual AI demand to meet the data center's being built,
and I think most data center loans up being underwater
within the next two years. To be clear, this isn't
a speculative chance, but rather one wrong on the chain
of pain that's been forged and looks like it's ready
to snap. Let's start by quoting my Premium newsletter from
(03:12):
December fifth. While many people talk about how circular the
AI bubble may or may not be, the reality is
that it's far more like a chain, a deeply vulnerable
one held together by debt and venture capital. A company
buys a GPU from Nvidia, at which point nobody is
making any profit anymore. These GPUs are purchased for the
most part using debt provided by banks or financial institutions,
(03:33):
while hyperscalers like Microsoft can and do fund GPUs using
cash flow even they have started to turn to debt.
At that point, the company that bought the GPUs six
hundreds of millions of dollars or even billions of dollars
to build out a data center, and once it turns on,
provides compute to a model provider such as open ai Oranthropic,
or even a smaller company, which then begins losing money
(03:54):
selling access to those GPUs, or of course training models
on them, which only loses money eat sample. Both open
ai and Anthropic loose billions of dollars, and both rely
on venture capital to fund their ability to continue paying
for accessing those GPUs. At that point, open ai and
Anthropic offer either subscriptions, which cost far more to offer
than the revenue they provide, or API access to their
(04:16):
models on a per million token basis. These rates are
often subsidized too. AI startups pay to access those models
to run their services, which end up costing more than
the revenue that they make from offering them, which means
they have to raise venture capital to continue paying to
access those models. Outside of hyperscalers paying in video for
GPUs out of cash flow, None of the AI industry
(04:38):
is fueled by revenue. Every single part of the industry
is fueled by a kind of subsidy, debt, or equity.
As a result, the AI bubble is really a stress
test of the global venture capital, private equity, private credit,
institutional and banking system and its willingness to fund all
of this forever. Because there isn't a single generative AI
(04:58):
company that's got a path to profit stability. You see,
every little link in the chain of pain is necessary
to understand things. The intitified stock market pumped not by
actual cash flow or productivity, by which I mean products
that are selling Google, Microsoft, Amazon matter they don't actually
generate real revenue and definitely not profit from AI, but
(05:19):
by signals read by analysts and investors trained over decades
the push consumers to invest in the Magnificent seven stocks.
These stocks represent as much as forty percent of the
value of the S and P five hundred and their values,
as I said, were pumped by analysts and the media,
misleading investors into believing that revenue growth for companies like Microsoft, Meta, Amazon,
and Google have anything to do with AI versus the
(05:39):
growth of their regular services. The other part in the
chain is venture capital's liquidity crisis, one peaking at a
time when startups have become more capital intensive than any
other point in history, and the underlying value of their
investments is only valuable as long as these companies stay alive.
And then, of course, the ballooning centralized data center debt
bubble funded based on customer contracts, will built for demand
(06:00):
that does not exist, finding massive data centers of GPUs
that immediately become commoditized as a result of this hysteria.
In really simple terms, I believe that almost every investment
in a data center AI startup may go to zero.
Let me explain if we assume that half of the
one hundred and seventy one point five billion dollars of
data center debt is in GPUs, so eighty five billion
(06:21):
or so, that's about three point two gigawats of data
set capacity. Based on my model of in Videa's approximate
split of sales between different aigpus from a premium priest
from a few weeks ago. As a brief explainer, I
made the model by taking in videos data center revenues,
estimating how many of each GPU they sell, and dividing
those revenues by those numbers, and markeplying the result by
the power output of the GPUs. In really simple terms,
(06:43):
I worked out how many GPUs they sold them what
the power to run them would be. There'd be about
three point two gigawts. The likelihood of the majority of
these products being a completed within the next year and
b completed on budget is very very small. Every delay
increases the likelihood off, as each of these projects is
heavily debt based. I also want to add open Ai claims,
(07:05):
and I actually think they're bullshitting here that they had
two gigawatts of compute at the end of twenty twenty five,
two gigawatts. They are the only company with any real
consumer exposure. They're the only ones with any number of
real users other than Google, who has literally changed Google
a system for Gemini. I don't think there's even like
(07:26):
five hundred megawatts of real demand outside of people training
models or Elon Musk's s SAM Generator, and that's a
whole different story now. In general, the customers of these
projects are either hyperscalers who are only doing AI because
they have no other hyper growth ideas and because Wall
Street currently approves, or AI startups, all of whom are unprofitable.
(07:47):
While there are potentially hedge funds or other small companies
looking for private AI integrations. I think this is a
very very small market. On top of that, AI compute
itself may not be profitable. And because by my estima,
everybody is spent about eighty five billion dollars on filling
data centers with the same GPUs, the aggregate price of
renting out GPUs will decline. Already, the average price of
(08:08):
renting out Blackwell GPUs is declined to an average of
four dollars and forty five cents an hour according to
Silicon Data, and that's before the majority of black Well
powered GPUs come online. Yeah, the customer base shrinks from
there because the majority of AI startups aren't actually renting GPUs.
They build products on top of models built by Open
Ai or Anthropic, who have made it clear they're buying
(08:29):
capacity from either hyperscalers, or, in open AI's case, getting
Oracle or core Weave to build it for them. Why
Because building your own model is incredibly capital intensive, both
on the compute and the data you have to get,
and it's hard to tell if the results will be
worth it. Now, let's assume I don't actually believe it will,
but let's try anyway that all of that three point
(08:49):
two gigawatsop capacity comes online? How much of it? How much? Actually?
How much compute does an AI company use? Like I said,
open Ai ended twenty twenty five to two gig what's
of capacity? And they apparently have nine hundred million weekly
active users. I don't think there are any AI companies
with even ten percent of that user base. But even
if they were, open ai spent eight point six seven
(09:10):
billion dollars on inference through the end of September. Who
can afford to pay even ten percent of that a
year or five percent? Yeah? In reality, open Ai is
likely more indicative of the overall compute spend of the
entire AI industry. As I've said, most companies are powered
not by their own GPU driven models, but by renting
them from model providers like open Ai. Themselves. Open Ai
(09:31):
and and Thropic Spenter combined eleven point three to three
billion dollars in compute on Zoro AWS, respectively through the
first nine months of last year. And that's my own reporting,
by the way, and as the two largest consumers of
AI compute, Well, I'm a little worried. These things suggest
two things. The market for AI compute is actually very
very small. If you assume that Anthropics spent the same
(09:54):
amount on Google Cloud as it did in AWS, that'd
be about a total of five point three two billion dollars,
and that Core Weaves revenue called it five billion dollars,
most of which was open Ai via Microsoft or in video.
That doesn't appear to be an AI compute market outside
of servicing two companies. The other problem is the market
for AI compute is not actually growing. In the last
(10:15):
two years, no new major customers of AI compute have emerged.
Every company that assigned a large compute deal has either
been open Ai, Anthropic or a hyperscaler. Even if Cursor
and AI Coding Company were to dump its entire two
point three billion dollars into funding AI compute, that would
still not be very much. In fact, it would take
sinking every single dollar of venture capital over two hundred
(10:39):
billion dollars every single year, and then some funneled into
AI compute just to produce the revenue to justify these
deals existence. In the space of a year, Microsoft Zure
made seventy five billion dollars, Google Cloud forty three billion dollars,
and Amazon Web Services one hundred billion dollars. Now that
is just add those numbers together, it's a little over
(10:59):
two hundred billion dollars, and that's the three largest providers
of any kind of compute, not just AI compute. Most
of that is not AI compute. In fact, they barely
make anything on that. They won't even talk about. The
AI revenue is probably because they're so stinky. Well, if
(11:24):
you need more proof, If you still don't believe me,
skip to page eighteen of Nvidia's most recent earnings and
I quote multi year cloud service agreement commitments as of
October twenty six, twenty twenty five, were twenty six billion dollars,
for which one billion dollars, six billion dollars, six billion dollars,
five billion dollars, four billion dollars, and four billion dollars
will we paid in fiscal years twenty twenty six, fourth quarter,
(11:44):
twenty twenty seven, twenty twenty eight, twenty twenty nine, twenty
thirty and twenty thirty one and thereafter respectively. Now I
know that was confusing, but just to line that up,
that means in fiscal year twenty twenty seven, which begins Oh,
I would have to look that up. Yeah, I'm literally
doing this on the air, and I'm gonna shit. So
their fiscal year begins in February, So basically twenty twenty
(12:07):
seven begins this February twenty twenty six. So they're going
to be spending six billion dollars on AI compute in
twenty twenty seven. This is a company that does not
rent out compute. If there's such incredible surging demand, why
is it? Why exactly is in Vidia spending six billion
fucking dollars a year in twenty twenty six and twenty
twenty seven on it. In Vidia doesn't need compute, it
(12:28):
just shut down. It's Amazon Web Services. Compared to those
it's called DJX cloud. It looks far more like in
video is propping up an industry with non existent demand.
I'm afraid there is no secret Amazon Web Services size
spend waiting in the wings for the right moment to pounce.
There is no secret demand wave, nor is there any
capacity crunch that is holding back incredible swarths of revenue oracles.
(12:50):
Five hundred and twenty three billion dollars in remaining performance
obligations are made up of OpenAI meta and you'll never
guess who in video fucking in VideA. The AI bubble
is just in Vidia being handed money and then handing
it to people. They're very profitable as well. They do
grame everyone else not so much. For AI data center
(13:13):
deals to make sense, most startups would have to start
becoming direct users of AI compute while also spending more
on cloud compute services than they've ever spent. The largest
consumers of AI compute are both unprofitable unsustainable monstrosities dependent
on venture capital. Eventually, reality will dawn on one or
more of these banks. Projects will get delayed thanks to
(13:33):
weather or budgetary issues, or when customers walk away, as
just happened to a data center ariit called FERMI, and
loan payments will start going unpaid elsewhere. AI startups will
start asking for money again and again, and for a
while they'll keep raising until the evaluations get too high,
or VC coffers get too low. You're probably going to
say at this point that Open AI or Anthropic might
(13:54):
go public, which will infuse capital into the system, and
I want to give you a preview of what that
might look like. Courtesy of AI Labs, Minimax, and Zipu,
as reported by the Information which just filed to go Bubblica.
I'm a little scared of These are the first real
numbers about the AI BUBBLA. Hope, I'm not going to
get embarrassed. Everybody's saying that behind the scenes these companies
(14:14):
are secretly great. Let's take a look. Oh my god,
these numbers aren't great at all. These numbers are terrible.
In the first half of this of twenty twenty five,
Zipu had a net loss of three hundred and thirty
four million dollars on twenty seven million dollars of revenue. Meanwhile,
Minimax made fifty three point four million dollars in revenue
in the first nine months of twenty twenty five and
(14:36):
bun two hundred and eleven million dollars to earn it.
It's time to wake up. These are the real life
costs of running an AI company, Open Ai and Anthropic
are going to be even worse than this. This is
why nobody wants to take AI company's public This is
why nobody wants to talk about the actual costs of
running AI. This is why nobody wants you to know
(14:58):
the hourly cost of running a GPU, And this is
why open Ai and Anthropic burn billions and billions of dollars.
The margins fucking stink. Every product is unprofitable and none
of these companies can afford their bills based on their
actual cash flow. Generative AI is not a functional industry,
and once the money works that out, everything burns. Though
(15:19):
many AI data centers boast of having tendency agreements, remember
these agreements are either with AI startups that will run
out of money who really shouldn't be renting GPUs by
the way, or hyperscalers with legal teams numbering in the thousands.
Every single deal that Microsoft, Amazon, Meta, Googler, and Video
signs is riddled withouts specifically hedging against this scenario, and
(15:39):
there won't be a damn thing that anybody can do
about it. If Hyperscalers decide to walk away before then,
in Video's bubble is likely to burst. As I discussed
a few weeks ago, and Vidia claims to have shipped
six million Blackwell GPUs, and while it may be employing
very dodgy maths claiming that each GPU is actually to
chips Jensen Jensen, I've been saying. My modeling of its
(16:02):
last three quarters suggests that in videos shipped around five
point three to three gigawatts worth of GPUs, and based
on reading about every single data center deal I can find,
it doesn't appear that many have been built and powered on,
or even started building in many cases. Worse still, in
Video's diversified revenue is collapsing. In the first quarter of
its fiscal year twenty twenty six, I'm sorry, you're just
(16:25):
going to have to ride with me on this one,
two customers represented sixteen percent and fourteen percent in revenue.
In the second quarter of fiscal year twenty six, two
customers represented twenty three percent and sixteen percent, and in
the third quarter, four customers represented twenty two percent, fifteen percent,
thirteen percent, and eleven percent of total revenue, with all
of that money going towards GPUs or networking gear. In
(16:46):
simpler terms, in Vidia's revenue is no longer coming from
a diverse swarth of customers. In its first quarter fiscal
year twenty twenty six, it had thirty point eight four
billion dollars of diversified revenue. That's from a customer that
from customers that do not number more than ten percent
of its revenue, and in the second quarter of FY
twenty six that was twenty eight point five billion Q three.
It's twenty two point twenty three billion. In simpler terms,
(17:10):
there are only a few companies that can buy GPUs,
and other companies are no longer buying GPUs at the
same rate. I doubt this number is going to increase.
You see in video, GPUs are astronomically expensive four point
five million dollars for a GB three hundred radc of
seventy two B three hundred GPUs, for example, and feeling
data centers full of them requires debt unless you're a hyperscaler. Well,
(17:31):
I can't say for sure. I believe in videos, diversified
revenue collapses a sign that smaller data center projects are
starting to have issues getting funded and or hyperscalers are
pulling back on the GPU purchases from those Taiwanese companies.
I was mentioning, Now, let me look through the eyes
of an AI boosted for a second. Okay, okay, everything
is blue and yellow as usual. Okay. One might say
(17:52):
that these big customers are covering the loss of revenue,
but the reality is that these big projects are run
on debt issued by banks are becoming increasingly worried about
nobody paying them back. The mistake that every investor, commentator, analyst,
and member of the media makes about Nvidia is believing
that its sales are an expression of demand for AI compume,
when in reality it's more of a statement about the
availability of debt from banks in private credit. Similarly, the
(18:16):
continued existence of AI startups is an expression of the
desperation of venture capital and the continuing flow of massive
funding rounds as assigned that they see no other revenues
for growth and that their startups can't wipe their own assholes. Eventually,
data centers are going to go unbuilt and data center
debt packages will begin to fall apart. Remember, Oracle's thirty
eight billion dollar data center deal is actually yet to close,
(18:37):
much like Stargate New Mexico is yet to close, and
much like Stargate Michigan is yet to close. These deals
while seeming like they're trending positively, are all incredibly important
to the future of the AI bubble, and any failure
will spooken already notice mark them only one link in
this chain of pain needs to break every single part
(18:57):
of the AI bubble. This fucking charade is unprofitable save
for in Video and the construction firms erecting future laser
tag arenas full of negative margin GPUs tell me what
happens if the debt stops flowing data centers? How will
in Vidia sell their so called twenty million Blackwell and
veried Ruben GPUs their claiming they'll ship by the end
(19:17):
of twenty twenty six. What happens if venture capitalists start
running low on funds and can't keep feeding hundreds of
millions of dollars to AI startups so that they can
feed them directly to open Ai or Anthropic. What happens
to open Ai and Anthropic and they're already negative margin
businesses when their customers can't pay them. What happens to
Oracle or core weaves, work in progress data centers if
(19:40):
open ai can't pay their bills? What happens to Anthropics?
Twenty one billion dollars of Broadcom orders or tens of
billions of dollars of Google cloud spend. And what if
I'm right about everything I've said, not just in this series,
but in the episodes and newsletters before him. In the
last year, I estimate I've been asked the question what
(20:00):
if You're wrong? Over twenty five times. Every single time
the question comes with this undercurrent venom, the suggestion that
I'm being an asshole for daring to question the wondrous
AI bubble. Every single person who's asked me this has
been poorly read, both in terms of my work and
the surrounding economics and technological possibilities of large language models
and believes they're defending technology, when in reality they're defending
(20:21):
growth in the rock economy's growth at all cost mindset.
In many cases, they are not excited about technology at all,
but the prospects of being the first in line to
lick an already sparkling boot. This has never ever been
about progress or productivity. If it was, we'd seen actual
progress or productivity boosts. So anything other than the frothiest debt,
(20:42):
the frothiest venture capital markets, and the most annoying and
incorrect headlines I've ever read in the news. Large language
models do not create novel concepts. They are inconsistent and unreliable,
and even the things that people like them for, like
coding very wildly thanks to the dramatic very tents of
a giant probability machine l alams, are not good enough
(21:05):
for people to pay regular software prices at any scale,
and the consequences of this will be that every single
dollar spent on GPUs has been for exactly one point
manipulating the value of stocks by making hyperscalers look good
so that they can keep pretending that they create anything.
AI does not have the business returns and may indeed
(21:27):
have negative gross margins across the board. It is inconsistent, ugly, unreliable,
expensive and environmentally ruinous, pissing off a large chunk of consumers,
and underwhelming most of the rest, other than those convinced
they're smart for using it, or those who have resigned
to giving up at the site of a confidence game
sold by a tech industry that stopped making products primarily
(21:48):
focused on solving the problems of consumers or businesses a
long time ago. You may say I'm wrong because Google, Microsoft,
Meta and Amazon continue to have healthy net revenues and
revenue growth. And as I've previously said, these companies are
not sharing AI revenues, and their existing businesses are still
growing due to the massive monopolies they've built. And I
(22:08):
want to plead to the AI boosters and bullish analysts alike,
you are being had. Sacha Adela, Sam Altman, Dario Amade,
Jensen Huang, Mark Zuckerberg, Larry Ellison, saffracats, Elon Musk, Klay mcguk,
Mike Cecilia, Mike truell Aravincerivenus, all of them are laughing
at you behind your back because they know that you're
(22:30):
never going to ask the obvious questions that would defeat
my arguments, and know that you will never ever push
back on them. They know the truth, they just don't
want to tell you because you don't care to argue.
(22:58):
The inceritification of the share has the downstream effect of
an intitification of the media and the Wall Street analysts
writ large analysts media. These companies own you, they own
your asses. They treat you with disdain and condescension because
they know you'll let them. They know that no Cell
side analysts will ever ask them when will you be profitable?
(23:19):
Or how much you spending? Or if you do ask,
they know you will experience temporary amnesia and forget whatever
answer they give, because these are the incentives of an
inshitified stock market, where stocks are not extrapolations of shareholder
value but chips in a fucking casino where the house
always wins and changes the rules every three months. These
companies have changed the meaning of the word stock to
(23:41):
mean whatever the market will reward, and when you allow
companies to start dictating the terms of what will be rewarded.
As neoliberalism, Freedman, Reagan, Nixon, NAFTA, and every other fucked
up growth focused policy has orienting everything exclusively around growth,
companies eventually cut off any powers that may curtail any
(24:02):
reevaluation of the fundamental terms of capitalism and the incentives
with him focusing on growth at all costs thinking naturally encourages, enables,
and empowers grifters, because all they ever have to promise
is more more users, more debt, more venture, more features,
more everything forever. From Adam Becker by it Today, the
(24:22):
very institutions that are meant to hold companies accountable. Analysts
and the media are far more desperate to trade scoops
for interviews, to pull punches, to find ways to explain
why a company is right, rather than understand what the
company is doing. And this is something pushed not by
writers but by editors that want to make sure they
stay on the right side of the largest companies. And goddamn,
do I know a few stories there that one day
(24:43):
I might even tell. And if I'm right, open AI's
death will kill off most, if not all, AI startups
Anthropic included. Every investor that invested in AI will take
massive losses. Every startup that builds on the back of
their models will see their companies fold if it hasn't
already done so to the massive costs an upcoming price
increase we've already seen signs of with priority processing with
(25:05):
both Anthropic and open Ai in the middle of last year.
The majority of GPU based data centers, which really have
no other revenue stream, will be left inert, likely powered down,
waiting for the day that someone works it all out,
which they won't because literally everybody has these things now,
and I truly believe they've tried everything. And I'm gonna
hear from someone that says us, just like the Duck combo,
they're fiber optic cable. We had so much fiber optic
(25:27):
cable and then we put internet in them. This is
not GPUs are not fiber optic cable. They're not I
just did a premium piece on this. I'm probably going
to have to do an episode on it. But they
are not the same fucking thing. And people that keep
saying this are flippant about the damage that's going to
be caused and just flat out wrong. I really do
challenge you to actually read history before you make any
(25:48):
fucking statements like that. This is not for the people
who are saying this just because they haven't looked and
they just assumed that the world wouldn't be stupid. I
don't think many people saying that or off that camp.
I think they are people that come up with comfortable
reasons to validate bad behavior. And I must be clear,
I'm watching I'm watching everyone. I'm watching how everyone handles this,
(26:12):
and I'm going to be watching extra hard on the
way down because anybody who has been a booster that
tries to change their target and tries to change their
direction without a direct acceptance and real contrition about their role.
I have taken such detailed notes, but I will add something.
(26:33):
I don't hate on AI because I'm a hater. I
hate on it because it fucking sucks. And what I'm
worried about happening seems to be happening. The tech industry
has run out of hypergrowth ideas, and in its desperation,
hitched itself to the least profitable hardware and software in history.
Then spent three straight years lying about what was possible
to the media, alice, and shareholders. And they were allowed
(26:54):
to lie because everybody lapped it the fuck up. They
didn't need to worry about convincing Anybodyncers, editors, analysts, and
investors were already drafting reasons why they were so excited
about something they didn't really understand or believe in, other
than the fact it promised more. This is what happens
when you make everything about growth. Everybody becomes stupid, ready
(27:15):
to be conned, ready to hear what the next big
growth thing is because asking nasty questions gets your fucking fired.
And what's left is a tech industry that doesn't build
technology but growth focused startups. Look at silicon value. Do
you see these fucking people ever building a new kind
of computer? Do you believe these men are fit to
even imagine a future? These men care about the status quo.
(27:36):
They want to always have more software to sell in
what or ways to increase advertising revenue so that the
stock number goes up, so they receive more money in
the form of stock compensation. They are concerned with neither
actual business value on this exchange of value or societal value.
Their existence is only in shareholder value, which is how
they are incentivized by their board of directors. And right
(27:56):
now they're in shitifying being a shareholder. But really, if
you're still defending AI, does it matter to any of
you that this software fucking sucks? If you think it's good,
you don't know much about software. Sure, your coding tool
is useful in whatever way it can fart out and
(28:16):
language simula or whatever. Oh, it can do some of
the easy coding things. I'm sorry, I'm sorry, man, I
can't get excited about that. We're hundreds of billions of
dollars in the hole. I need something more than that.
I need so much more than that. And if I'm honest,
the problem with large language models, it's not good software.
(28:37):
It doesn't respond precisely at any point to a user
or a programmers intent. That's bad software. And I don't
care that you've heard developers really like it, because that
doesn't fix the underlying economic and social poison in AI.
I don't care that it's sort of replaced search for you.
I don't care if you know a team of engineers
that use it. Every single AI app is subsidized. Its
price is fake. You are being laid to, and none
(28:58):
of this is real. I also, and this is a
side a little bit of a side quest, I think
AI psychosis is so much bigger than people think. I
feel like I might have mentioned this on the recent monologue,
but it's just it's really bothering me. I think there
is something that happens when you play with these things
and you make them work, versus just being like, oh,
(29:18):
they don't work. I think what happens is it convinces
you you're smart, when what you really are is being
a slave to the software yourself. You are being tricked.
You are being convinced that the software is doing something
because you are bonking it on the head until it
does something you need it to. Really, none of this
is real, none of it. I'm so scared. I'm so
(29:40):
scared because on one hand, I kind of look like
I'm about to be proven right in a way that's
going to be good for business, I guess. But being
right here means watching something really calamitous and when the
collapse happens. Do not let a single person that waved
off the economics have a moment's piece. Do not let
(30:01):
anybody who sat in front of Dario Amadey or Sam
Ortman such as I don't know on a New York
Times podcast and squealed with delight at whatever vacuous talking
points they burped out. Forget that they didn't push them,
They didn't ask the hard questions. They didn't worry or
wonder or feel any concerns for investors of the general public.
They only cared about getting rock hard listening to whatever
bullshit the latest large Language model asshole has to say.
(30:25):
Do not let a single analyst that called AI skeptics, loodites,
Daniel Newman futuream group, or equate them to flat earthers
hear the end of it. Do not let anybody who
claimed that we lost control of AI or that it
blackmailed developers go without their complimentary fell for it again, badge,
because no, these systems didn't blackmail anyone, they were literally
(30:46):
being prompted to do it. You fell for it, and
you will You'll No one is ever going to hear
the end of this from me, And when it happens,
I promise I won't be too insufferable, but I will
be calling for accountability for anybody who boosted AI twenty
twenty seven, who sat in front of Sam Ortman of
Dariyama Day and once again refused to ask real questions
(31:07):
or allowed bullshit answers, And for anyone anyone who collected
anything resembling detailed notes about me or any other AI
skeptic threatening independent reporters because I didn't clap for Daddy's
venture capital. Bullshit is insufferable, poisonous, and will never ever
be forgotten. And if you think I'm talking about you,
(31:28):
I probably am. And I have a question. Why didn't
you approach the AI companies with as much skepticism as
you did the skeptics. Why were you so concerned about
minor details when you could have cared about the major
details that will very likely fuck our markets up lead
to an apocalyptic state of venture capital and startups, and genuinely,
(31:51):
I don't even know if a recession will be it.
I also genuinely promise you if I'm wrong, I'll happily
explain how and why, and I'll do so at length
to I'll have links, I'll have citations, I'll do podcast episodes.
I will make a good faith effort to explain every
single failing because my concern is the truth and I
(32:12):
would love everybody else to follow suit. Do you think
any AI booster will have the same courtesy? Do you
think they care about the truth or do they want
their fish biscuit from Sam Ormon or Jensen Joan. It's
absolutely pathetic, and as we enter into twenty twenty six,
I'm going to fucking floor it with this stuff. I've
taught myself a great deal about numbers in the last year.
(32:33):
I've really got a good head for reading these ten cues,
and there's already stuff I'm seeing that is spelling the
end that I'm yet to talk about. This has been
a long episode, and it's been a long four parter.
I really hope you've enjoyed it, because I've loved reading it.
More to come next week I'll be joined by mister
Matt Rosoff, wonderful editor in chief over at the Register.
(32:55):
We're going to talk about the dot com bubble. Just folks,
hit me up on the round app, slub me on Goop.
Thank you for listening to Better Offline. The editor and
composer of the Better Offline theme song is Matasowski. You
can check out more of his music and audio projects
(33:17):
at Matasowski dot com, M A T T O S
O W s ki dot com. You can email me
at easy at Better Offline dot com or visit Better
Offline dot com to find more podcast links and of
course my newsletter. I also really recommend you go to
chat dot Where's Youreed dot at to visit the discord,
and go to our slash Better Offline to check out
(33:37):
our reddit. Thank you so much for listening. Better Offline
is a production of cool Zone Media. For more from
cool Zone Media, visit our website cool Zonemedia dot com,
or check us out on the iHeartRadio app, Apple Podcasts,
or wherever you get your podcasts.
Media. Hello and welcome back to Better Offline. I'm your
host ed ZiT Try. This is the last part of
our four part series about how the financial system and
(00:23):
the financial markets are the last victims of the inshitification process.
Over the past four episodes, in a twenty thousand or
so word newsletter which this series is based on, I've
explained how and in this final episode, I'm going to
explain how all this was going to fall apart and
make it clear where I'm so fucking alarmed. Now let's
(00:43):
get back to those data centers. We've already got some
signs of concern within the banking world around them, and
I'm kind of worried, and so are they about their exposure.
In November, the ft reported the Deutsche Bank, which back
cor we've multiple times and several data centers, was the
exploring ways the hedge it's exposure to the AI data
centers after extending billions of dollars in debt, including shortening
(01:06):
a basket of AI related stocks or buying default protection
on some of its debt using synthetic risk transfers, which
are when a bank sells the full or partial credit
risk of a loan or loans to another bank or
keeping the loans on their book paying a monthly fee
to investors. This is a simplification. Please don't be mad
at me. In December, Fortune reported the Morgan Stanley three
(01:28):
times with core Weave IPI partners, Hyperions, Soft Banks Bridge
Loan was also considering synthetic risk transfers on loans to
businesses involved in AI infrastructure. One of what those were
for now. Back in April, SMBC sold synthetic risk transfers.
Back in April, SMBC sold synthetic risk transfers tied to
(01:48):
private debt BDCs. And while this predates the larger data
center deals done by Blue Al, who is a BDC themselves,
smbc's oversee multiple blue ol deals in the past. In December,
s MBC closed another SRT selling off risk from Australian
and Asian project finance loans, though we can't confirm if
any of those were data center related now. In December,
(02:10):
Goldman Sachs pause the planned mortgage bond sale for data
Center Operate A Cyrus one, with the intent to revive
it in the first quarter of twenty twenty six. I
have heard nothing from that since Oracle's credit risk reaches
sixteen year high in the middle of December, with credit
default swaps basically betting the Oracle will default on its debt,
which is unlikely yet no longer impossible, climbing to their
(02:31):
highest price since the Great Financial Crisis. While Morgan Stanley
and Deutsche Banks SRTs are yet to close, it's still
notable that two of the largest players in data center
financing feel the need to hedge their bets. So what
exactly are they hedging against? Simple that tenants won't arrive
and debts won't get paid. I also believe that they're
(02:51):
going to need bigger hedges because I don't think there's
enough actual AI demand to meet the data center's being built,
and I think most data center loans up being underwater
within the next two years. To be clear, this isn't
a speculative chance, but rather one wrong on the chain
of pain that's been forged and looks like it's ready
to snap. Let's start by quoting my Premium newsletter from
(03:12):
December fifth. While many people talk about how circular the
AI bubble may or may not be, the reality is
that it's far more like a chain, a deeply vulnerable
one held together by debt and venture capital. A company
buys a GPU from Nvidia, at which point nobody is
making any profit anymore. These GPUs are purchased for the
most part using debt provided by banks or financial institutions,
(03:33):
while hyperscalers like Microsoft can and do fund GPUs using
cash flow even they have started to turn to debt.
At that point, the company that bought the GPUs six
hundreds of millions of dollars or even billions of dollars
to build out a data center, and once it turns on,
provides compute to a model provider such as open ai Oranthropic,
or even a smaller company, which then begins losing money
(03:54):
selling access to those GPUs, or of course training models
on them, which only loses money eat sample. Both open
ai and Anthropic loose billions of dollars, and both rely
on venture capital to fund their ability to continue paying
for accessing those GPUs. At that point, open ai and
Anthropic offer either subscriptions, which cost far more to offer
than the revenue they provide, or API access to their
(04:16):
models on a per million token basis. These rates are
often subsidized too. AI startups pay to access those models
to run their services, which end up costing more than
the revenue that they make from offering them, which means
they have to raise venture capital to continue paying to
access those models. Outside of hyperscalers paying in video for
GPUs out of cash flow, None of the AI industry
(04:38):
is fueled by revenue. Every single part of the industry
is fueled by a kind of subsidy, debt, or equity.
As a result, the AI bubble is really a stress
test of the global venture capital, private equity, private credit,
institutional and banking system and its willingness to fund all
of this forever. Because there isn't a single generative AI
(04:58):
company that's got a path to profit stability. You see,
every little link in the chain of pain is necessary
to understand things. The intitified stock market pumped not by
actual cash flow or productivity, by which I mean products
that are selling Google, Microsoft, Amazon matter they don't actually
generate real revenue and definitely not profit from AI, but
(05:19):
by signals read by analysts and investors trained over decades
the push consumers to invest in the Magnificent seven stocks.
These stocks represent as much as forty percent of the
value of the S and P five hundred and their values,
as I said, were pumped by analysts and the media,
misleading investors into believing that revenue growth for companies like Microsoft, Meta, Amazon,
and Google have anything to do with AI versus the
(05:39):
growth of their regular services. The other part in the
chain is venture capital's liquidity crisis, one peaking at a
time when startups have become more capital intensive than any
other point in history, and the underlying value of their
investments is only valuable as long as these companies stay alive.
And then, of course, the ballooning centralized data center debt
bubble funded based on customer contracts, will built for demand
(06:00):
that does not exist, finding massive data centers of GPUs
that immediately become commoditized as a result of this hysteria.
In really simple terms, I believe that almost every investment
in a data center AI startup may go to zero.
Let me explain if we assume that half of the
one hundred and seventy one point five billion dollars of
data center debt is in GPUs, so eighty five billion
(06:21):
or so, that's about three point two gigawats of data
set capacity. Based on my model of in Videa's approximate
split of sales between different aigpus from a premium priest
from a few weeks ago. As a brief explainer, I
made the model by taking in videos data center revenues,
estimating how many of each GPU they sell, and dividing
those revenues by those numbers, and markeplying the result by
the power output of the GPUs. In really simple terms,
(06:43):
I worked out how many GPUs they sold them what
the power to run them would be. There'd be about
three point two gigawts. The likelihood of the majority of
these products being a completed within the next year and
b completed on budget is very very small. Every delay
increases the likelihood off, as each of these projects is
heavily debt based. I also want to add open Ai claims,
(07:05):
and I actually think they're bullshitting here that they had
two gigawatts of compute at the end of twenty twenty five,
two gigawatts. They are the only company with any real
consumer exposure. They're the only ones with any number of
real users other than Google, who has literally changed Google
a system for Gemini. I don't think there's even like
(07:26):
five hundred megawatts of real demand outside of people training
models or Elon Musk's s SAM Generator, and that's a
whole different story now. In general, the customers of these
projects are either hyperscalers who are only doing AI because
they have no other hyper growth ideas and because Wall
Street currently approves, or AI startups, all of whom are unprofitable.
(07:47):
While there are potentially hedge funds or other small companies
looking for private AI integrations. I think this is a
very very small market. On top of that, AI compute
itself may not be profitable. And because by my estima,
everybody is spent about eighty five billion dollars on filling
data centers with the same GPUs, the aggregate price of
renting out GPUs will decline. Already, the average price of
(08:08):
renting out Blackwell GPUs is declined to an average of
four dollars and forty five cents an hour according to
Silicon Data, and that's before the majority of black Well
powered GPUs come online. Yeah, the customer base shrinks from
there because the majority of AI startups aren't actually renting GPUs.
They build products on top of models built by Open
Ai or Anthropic, who have made it clear they're buying
(08:29):
capacity from either hyperscalers, or, in open AI's case, getting
Oracle or core Weave to build it for them. Why
Because building your own model is incredibly capital intensive, both
on the compute and the data you have to get,
and it's hard to tell if the results will be
worth it. Now, let's assume I don't actually believe it will,
but let's try anyway that all of that three point
(08:49):
two gigawatsop capacity comes online? How much of it? How much? Actually?
How much compute does an AI company use? Like I said,
open Ai ended twenty twenty five to two gig what's
of capacity? And they apparently have nine hundred million weekly
active users. I don't think there are any AI companies
with even ten percent of that user base. But even
if they were, open ai spent eight point six seven
(09:10):
billion dollars on inference through the end of September. Who
can afford to pay even ten percent of that a
year or five percent? Yeah? In reality, open Ai is
likely more indicative of the overall compute spend of the
entire AI industry. As I've said, most companies are powered
not by their own GPU driven models, but by renting
them from model providers like open Ai. Themselves. Open Ai
(09:31):
and and Thropic Spenter combined eleven point three to three
billion dollars in compute on Zoro AWS, respectively through the
first nine months of last year. And that's my own reporting,
by the way, and as the two largest consumers of
AI compute, Well, I'm a little worried. These things suggest
two things. The market for AI compute is actually very
very small. If you assume that Anthropics spent the same
(09:54):
amount on Google Cloud as it did in AWS, that'd
be about a total of five point three two billion dollars,
and that Core Weaves revenue called it five billion dollars,
most of which was open Ai via Microsoft or in video.
That doesn't appear to be an AI compute market outside
of servicing two companies. The other problem is the market
for AI compute is not actually growing. In the last
(10:15):
two years, no new major customers of AI compute have emerged.
Every company that assigned a large compute deal has either
been open Ai, Anthropic or a hyperscaler. Even if Cursor
and AI Coding Company were to dump its entire two
point three billion dollars into funding AI compute, that would
still not be very much. In fact, it would take
sinking every single dollar of venture capital over two hundred
(10:39):
billion dollars every single year, and then some funneled into
AI compute just to produce the revenue to justify these
deals existence. In the space of a year, Microsoft Zure
made seventy five billion dollars, Google Cloud forty three billion dollars,
and Amazon Web Services one hundred billion dollars. Now that
is just add those numbers together, it's a little over
(10:59):
two hundred billion dollars, and that's the three largest providers
of any kind of compute, not just AI compute. Most
of that is not AI compute. In fact, they barely
make anything on that. They won't even talk about. The
AI revenue is probably because they're so stinky. Well, if
(11:24):
you need more proof, If you still don't believe me,
skip to page eighteen of Nvidia's most recent earnings and
I quote multi year cloud service agreement commitments as of
October twenty six, twenty twenty five, were twenty six billion dollars,
for which one billion dollars, six billion dollars, six billion dollars,
five billion dollars, four billion dollars, and four billion dollars
will we paid in fiscal years twenty twenty six, fourth quarter,
(11:44):
twenty twenty seven, twenty twenty eight, twenty twenty nine, twenty
thirty and twenty thirty one and thereafter respectively. Now I
know that was confusing, but just to line that up,
that means in fiscal year twenty twenty seven, which begins Oh,
I would have to look that up. Yeah, I'm literally
doing this on the air, and I'm gonna shit. So
their fiscal year begins in February, So basically twenty twenty
(12:07):
seven begins this February twenty twenty six. So they're going
to be spending six billion dollars on AI compute in
twenty twenty seven. This is a company that does not
rent out compute. If there's such incredible surging demand, why
is it? Why exactly is in Vidia spending six billion
fucking dollars a year in twenty twenty six and twenty
twenty seven on it. In Vidia doesn't need compute, it
(12:28):
just shut down. It's Amazon Web Services. Compared to those
it's called DJX cloud. It looks far more like in
video is propping up an industry with non existent demand.
I'm afraid there is no secret Amazon Web Services size
spend waiting in the wings for the right moment to pounce.
There is no secret demand wave, nor is there any
capacity crunch that is holding back incredible swarths of revenue oracles.
(12:50):
Five hundred and twenty three billion dollars in remaining performance
obligations are made up of OpenAI meta and you'll never
guess who in video fucking in VideA. The AI bubble
is just in Vidia being handed money and then handing
it to people. They're very profitable as well. They do
grame everyone else not so much. For AI data center
(13:13):
deals to make sense, most startups would have to start
becoming direct users of AI compute while also spending more
on cloud compute services than they've ever spent. The largest
consumers of AI compute are both unprofitable unsustainable monstrosities dependent
on venture capital. Eventually, reality will dawn on one or
more of these banks. Projects will get delayed thanks to
(13:33):
weather or budgetary issues, or when customers walk away, as
just happened to a data center ariit called FERMI, and
loan payments will start going unpaid elsewhere. AI startups will
start asking for money again and again, and for a
while they'll keep raising until the evaluations get too high,
or VC coffers get too low. You're probably going to
say at this point that Open AI or Anthropic might
(13:54):
go public, which will infuse capital into the system, and
I want to give you a preview of what that
might look like. Courtesy of AI Labs, Minimax, and Zipu,
as reported by the Information which just filed to go Bubblica.
I'm a little scared of These are the first real
numbers about the AI BUBBLA. Hope, I'm not going to
get embarrassed. Everybody's saying that behind the scenes these companies
(14:14):
are secretly great. Let's take a look. Oh my god,
these numbers aren't great at all. These numbers are terrible.
In the first half of this of twenty twenty five,
Zipu had a net loss of three hundred and thirty
four million dollars on twenty seven million dollars of revenue. Meanwhile,
Minimax made fifty three point four million dollars in revenue
in the first nine months of twenty twenty five and
(14:36):
bun two hundred and eleven million dollars to earn it.
It's time to wake up. These are the real life
costs of running an AI company, Open Ai and Anthropic
are going to be even worse than this. This is
why nobody wants to take AI company's public This is
why nobody wants to talk about the actual costs of
running AI. This is why nobody wants you to know
(14:58):
the hourly cost of running a GPU, And this is
why open Ai and Anthropic burn billions and billions of dollars.
The margins fucking stink. Every product is unprofitable and none
of these companies can afford their bills based on their
actual cash flow. Generative AI is not a functional industry,
and once the money works that out, everything burns. Though
(15:19):
many AI data centers boast of having tendency agreements, remember
these agreements are either with AI startups that will run
out of money who really shouldn't be renting GPUs by
the way, or hyperscalers with legal teams numbering in the thousands.
Every single deal that Microsoft, Amazon, Meta, Googler, and Video
signs is riddled withouts specifically hedging against this scenario, and
(15:39):
there won't be a damn thing that anybody can do
about it. If Hyperscalers decide to walk away before then,
in Video's bubble is likely to burst. As I discussed
a few weeks ago, and Vidia claims to have shipped
six million Blackwell GPUs, and while it may be employing
very dodgy maths claiming that each GPU is actually to
chips Jensen Jensen, I've been saying. My modeling of its
(16:02):
last three quarters suggests that in videos shipped around five
point three to three gigawatts worth of GPUs, and based
on reading about every single data center deal I can find,
it doesn't appear that many have been built and powered on,
or even started building in many cases. Worse still, in
Video's diversified revenue is collapsing. In the first quarter of
its fiscal year twenty twenty six, I'm sorry, you're just
(16:25):
going to have to ride with me on this one,
two customers represented sixteen percent and fourteen percent in revenue.
In the second quarter of fiscal year twenty six, two
customers represented twenty three percent and sixteen percent, and in
the third quarter, four customers represented twenty two percent, fifteen percent,
thirteen percent, and eleven percent of total revenue, with all
of that money going towards GPUs or networking gear. In
(16:46):
simpler terms, in Vidia's revenue is no longer coming from
a diverse swarth of customers. In its first quarter fiscal
year twenty twenty six, it had thirty point eight four
billion dollars of diversified revenue. That's from a customer that
from customers that do not number more than ten percent
of its revenue, and in the second quarter of FY
twenty six that was twenty eight point five billion Q three.
It's twenty two point twenty three billion. In simpler terms,
(17:10):
there are only a few companies that can buy GPUs,
and other companies are no longer buying GPUs at the
same rate. I doubt this number is going to increase.
You see in video, GPUs are astronomically expensive four point
five million dollars for a GB three hundred radc of
seventy two B three hundred GPUs, for example, and feeling
data centers full of them requires debt unless you're a hyperscaler. Well,
(17:31):
I can't say for sure. I believe in videos, diversified
revenue collapses a sign that smaller data center projects are
starting to have issues getting funded and or hyperscalers are
pulling back on the GPU purchases from those Taiwanese companies.
I was mentioning, Now, let me look through the eyes
of an AI boosted for a second. Okay, okay, everything
is blue and yellow as usual. Okay. One might say
(17:52):
that these big customers are covering the loss of revenue,
but the reality is that these big projects are run
on debt issued by banks are becoming increasingly worried about
nobody paying them back. The mistake that every investor, commentator, analyst,
and member of the media makes about Nvidia is believing
that its sales are an expression of demand for AI compume,
when in reality it's more of a statement about the
availability of debt from banks in private credit. Similarly, the
(18:16):
continued existence of AI startups is an expression of the
desperation of venture capital and the continuing flow of massive
funding rounds as assigned that they see no other revenues
for growth and that their startups can't wipe their own assholes. Eventually,
data centers are going to go unbuilt and data center
debt packages will begin to fall apart. Remember, Oracle's thirty
eight billion dollar data center deal is actually yet to close,
(18:37):
much like Stargate New Mexico is yet to close, and
much like Stargate Michigan is yet to close. These deals
while seeming like they're trending positively, are all incredibly important
to the future of the AI bubble, and any failure
will spooken already notice mark them only one link in
this chain of pain needs to break every single part
(18:57):
of the AI bubble. This fucking charade is unprofitable save
for in Video and the construction firms erecting future laser
tag arenas full of negative margin GPUs tell me what
happens if the debt stops flowing data centers? How will
in Vidia sell their so called twenty million Blackwell and
veried Ruben GPUs their claiming they'll ship by the end
(19:17):
of twenty twenty six. What happens if venture capitalists start
running low on funds and can't keep feeding hundreds of
millions of dollars to AI startups so that they can
feed them directly to open Ai or Anthropic. What happens
to open Ai and Anthropic and they're already negative margin
businesses when their customers can't pay them. What happens to
Oracle or core weaves, work in progress data centers if
(19:40):
open ai can't pay their bills? What happens to Anthropics?
Twenty one billion dollars of Broadcom orders or tens of
billions of dollars of Google cloud spend. And what if
I'm right about everything I've said, not just in this series,
but in the episodes and newsletters before him. In the
last year, I estimate I've been asked the question what
(20:00):
if You're wrong? Over twenty five times. Every single time
the question comes with this undercurrent venom, the suggestion that
I'm being an asshole for daring to question the wondrous
AI bubble. Every single person who's asked me this has
been poorly read, both in terms of my work and
the surrounding economics and technological possibilities of large language models
and believes they're defending technology, when in reality they're defending
(20:21):
growth in the rock economy's growth at all cost mindset.
In many cases, they are not excited about technology at all,
but the prospects of being the first in line to
lick an already sparkling boot. This has never ever been
about progress or productivity. If it was, we'd seen actual
progress or productivity boosts. So anything other than the frothiest debt,
(20:42):
the frothiest venture capital markets, and the most annoying and
incorrect headlines I've ever read in the news. Large language
models do not create novel concepts. They are inconsistent and unreliable,
and even the things that people like them for, like
coding very wildly thanks to the dramatic very tents of
a giant probability machine l alams, are not good enough
(21:05):
for people to pay regular software prices at any scale,
and the consequences of this will be that every single
dollar spent on GPUs has been for exactly one point
manipulating the value of stocks by making hyperscalers look good
so that they can keep pretending that they create anything.
AI does not have the business returns and may indeed
(21:27):
have negative gross margins across the board. It is inconsistent, ugly, unreliable,
expensive and environmentally ruinous, pissing off a large chunk of consumers,
and underwhelming most of the rest, other than those convinced
they're smart for using it, or those who have resigned
to giving up at the site of a confidence game
sold by a tech industry that stopped making products primarily
(21:48):
focused on solving the problems of consumers or businesses a
long time ago. You may say I'm wrong because Google, Microsoft,
Meta and Amazon continue to have healthy net revenues and
revenue growth. And as I've previously said, these companies are
not sharing AI revenues, and their existing businesses are still
growing due to the massive monopolies they've built. And I
(22:08):
want to plead to the AI boosters and bullish analysts alike,
you are being had. Sacha Adela, Sam Altman, Dario Amade,
Jensen Huang, Mark Zuckerberg, Larry Ellison, saffracats, Elon Musk, Klay mcguk,
Mike Cecilia, Mike truell Aravincerivenus, all of them are laughing
at you behind your back because they know that you're
(22:30):
never going to ask the obvious questions that would defeat
my arguments, and know that you will never ever push
back on them. They know the truth, they just don't
want to tell you because you don't care to argue.
(22:58):
The inceritification of the share has the downstream effect of
an intitification of the media and the Wall Street analysts
writ large analysts media. These companies own you, they own
your asses. They treat you with disdain and condescension because
they know you'll let them. They know that no Cell
side analysts will ever ask them when will you be profitable?
(23:19):
Or how much you spending? Or if you do ask,
they know you will experience temporary amnesia and forget whatever
answer they give, because these are the incentives of an
inshitified stock market, where stocks are not extrapolations of shareholder
value but chips in a fucking casino where the house
always wins and changes the rules every three months. These
companies have changed the meaning of the word stock to
(23:41):
mean whatever the market will reward, and when you allow
companies to start dictating the terms of what will be rewarded.
As neoliberalism, Freedman, Reagan, Nixon, NAFTA, and every other fucked
up growth focused policy has orienting everything exclusively around growth,
companies eventually cut off any powers that may curtail any
(24:02):
reevaluation of the fundamental terms of capitalism and the incentives
with him focusing on growth at all costs thinking naturally encourages, enables,
and empowers grifters, because all they ever have to promise
is more more users, more debt, more venture, more features,
more everything forever. From Adam Becker by it Today, the
(24:22):
very institutions that are meant to hold companies accountable. Analysts
and the media are far more desperate to trade scoops
for interviews, to pull punches, to find ways to explain
why a company is right, rather than understand what the
company is doing. And this is something pushed not by
writers but by editors that want to make sure they
stay on the right side of the largest companies. And goddamn,
do I know a few stories there that one day
(24:43):
I might even tell. And if I'm right, open AI's
death will kill off most, if not all, AI startups
Anthropic included. Every investor that invested in AI will take
massive losses. Every startup that builds on the back of
their models will see their companies fold if it hasn't
already done so to the massive costs an upcoming price
increase we've already seen signs of with priority processing with
(25:05):
both Anthropic and open Ai in the middle of last year.
The majority of GPU based data centers, which really have
no other revenue stream, will be left inert, likely powered down,
waiting for the day that someone works it all out,
which they won't because literally everybody has these things now,
and I truly believe they've tried everything. And I'm gonna
hear from someone that says us, just like the Duck combo,
they're fiber optic cable. We had so much fiber optic
(25:27):
cable and then we put internet in them. This is
not GPUs are not fiber optic cable. They're not I
just did a premium piece on this. I'm probably going
to have to do an episode on it. But they
are not the same fucking thing. And people that keep
saying this are flippant about the damage that's going to
be caused and just flat out wrong. I really do
challenge you to actually read history before you make any
(25:48):
fucking statements like that. This is not for the people
who are saying this just because they haven't looked and
they just assumed that the world wouldn't be stupid. I
don't think many people saying that or off that camp.
I think they are people that come up with comfortable
reasons to validate bad behavior. And I must be clear,
I'm watching I'm watching everyone. I'm watching how everyone handles this,
(26:12):
and I'm going to be watching extra hard on the
way down because anybody who has been a booster that
tries to change their target and tries to change their
direction without a direct acceptance and real contrition about their role.
I have taken such detailed notes, but I will add something.
(26:33):
I don't hate on AI because I'm a hater. I
hate on it because it fucking sucks. And what I'm
worried about happening seems to be happening. The tech industry
has run out of hypergrowth ideas, and in its desperation,
hitched itself to the least profitable hardware and software in history.
Then spent three straight years lying about what was possible
to the media, alice, and shareholders. And they were allowed
(26:54):
to lie because everybody lapped it the fuck up. They
didn't need to worry about convincing Anybodyncers, editors, analysts, and
investors were already drafting reasons why they were so excited
about something they didn't really understand or believe in, other
than the fact it promised more. This is what happens
when you make everything about growth. Everybody becomes stupid, ready
(27:15):
to be conned, ready to hear what the next big
growth thing is because asking nasty questions gets your fucking fired.
And what's left is a tech industry that doesn't build
technology but growth focused startups. Look at silicon value. Do
you see these fucking people ever building a new kind
of computer? Do you believe these men are fit to
even imagine a future? These men care about the status quo.
(27:36):
They want to always have more software to sell in
what or ways to increase advertising revenue so that the
stock number goes up, so they receive more money in
the form of stock compensation. They are concerned with neither
actual business value on this exchange of value or societal value.
Their existence is only in shareholder value, which is how
they are incentivized by their board of directors. And right
(27:56):
now they're in shitifying being a shareholder. But really, if
you're still defending AI, does it matter to any of
you that this software fucking sucks? If you think it's good,
you don't know much about software. Sure, your coding tool
is useful in whatever way it can fart out and
(28:16):
language simula or whatever. Oh, it can do some of
the easy coding things. I'm sorry, I'm sorry, man, I
can't get excited about that. We're hundreds of billions of
dollars in the hole. I need something more than that.
I need so much more than that. And if I'm honest,
the problem with large language models, it's not good software.
(28:37):
It doesn't respond precisely at any point to a user
or a programmers intent. That's bad software. And I don't
care that you've heard developers really like it, because that
doesn't fix the underlying economic and social poison in AI.
I don't care that it's sort of replaced search for you.
I don't care if you know a team of engineers
that use it. Every single AI app is subsidized. Its
price is fake. You are being laid to, and none
(28:58):
of this is real. I also, and this is a
side a little bit of a side quest, I think
AI psychosis is so much bigger than people think. I
feel like I might have mentioned this on the recent monologue,
but it's just it's really bothering me. I think there
is something that happens when you play with these things
and you make them work, versus just being like, oh,
(29:18):
they don't work. I think what happens is it convinces
you you're smart, when what you really are is being
a slave to the software yourself. You are being tricked.
You are being convinced that the software is doing something
because you are bonking it on the head until it
does something you need it to. Really, none of this
is real, none of it. I'm so scared. I'm so
(29:40):
scared because on one hand, I kind of look like
I'm about to be proven right in a way that's
going to be good for business, I guess. But being
right here means watching something really calamitous and when the
collapse happens. Do not let a single person that waved
off the economics have a moment's piece. Do not let
(30:01):
anybody who sat in front of Dario Amadey or Sam
Ortman such as I don't know on a New York
Times podcast and squealed with delight at whatever vacuous talking
points they burped out. Forget that they didn't push them,
They didn't ask the hard questions. They didn't worry or
wonder or feel any concerns for investors of the general public.
They only cared about getting rock hard listening to whatever
bullshit the latest large Language model asshole has to say.
(30:25):
Do not let a single analyst that called AI skeptics, loodites,
Daniel Newman futuream group, or equate them to flat earthers
hear the end of it. Do not let anybody who
claimed that we lost control of AI or that it
blackmailed developers go without their complimentary fell for it again, badge,
because no, these systems didn't blackmail anyone, they were literally
(30:46):
being prompted to do it. You fell for it, and
you will You'll No one is ever going to hear
the end of this from me, And when it happens,
I promise I won't be too insufferable, but I will
be calling for accountability for anybody who boosted AI twenty
twenty seven, who sat in front of Sam Ortman of
Dariyama Day and once again refused to ask real questions
(31:07):
or allowed bullshit answers, And for anyone anyone who collected
anything resembling detailed notes about me or any other AI
skeptic threatening independent reporters because I didn't clap for Daddy's
venture capital. Bullshit is insufferable, poisonous, and will never ever
be forgotten. And if you think I'm talking about you,
(31:28):
I probably am. And I have a question. Why didn't
you approach the AI companies with as much skepticism as
you did the skeptics. Why were you so concerned about
minor details when you could have cared about the major
details that will very likely fuck our markets up lead
to an apocalyptic state of venture capital and startups, and genuinely,
(31:51):
I don't even know if a recession will be it.
I also genuinely promise you if I'm wrong, I'll happily
explain how and why, and I'll do so at length
to I'll have links, I'll have citations, I'll do podcast episodes.
I will make a good faith effort to explain every
single failing because my concern is the truth and I
(32:12):
would love everybody else to follow suit. Do you think
any AI booster will have the same courtesy? Do you
think they care about the truth or do they want
their fish biscuit from Sam Ormon or Jensen Joan. It's
absolutely pathetic, and as we enter into twenty twenty six,
I'm going to fucking floor it with this stuff. I've
taught myself a great deal about numbers in the last year.
(32:33):
I've really got a good head for reading these ten cues,
and there's already stuff I'm seeing that is spelling the
end that I'm yet to talk about. This has been
a long episode, and it's been a long four parter.
I really hope you've enjoyed it, because I've loved reading it.
More to come next week I'll be joined by mister
Matt Rosoff, wonderful editor in chief over at the Register.
(32:55):
We're going to talk about the dot com bubble. Just folks,
hit me up on the round app, slub me on Goop.
Thank you for listening to Better Offline. The editor and
composer of the Better Offline theme song is Matasowski. You
can check out more of his music and audio projects
(33:17):
at Matasowski dot com, M A T T O S
O W s ki dot com. You can email me
at easy at Better Offline dot com or visit Better
Offline dot com to find more podcast links and of
course my newsletter. I also really recommend you go to
chat dot Where's Youreed dot at to visit the discord,
and go to our slash Better Offline to check out
(33:37):
our reddit. Thank you so much for listening. Better Offline
is a production of cool Zone Media. For more from
cool Zone Media, visit our website cool Zonemedia dot com,
or check us out on the iHeartRadio app, Apple Podcasts,
or wherever you get your podcasts.
Host
Ed Zitron
Popular Podcasts
Two Guys, Five Rings: Matt, Bowen & The Olympics
Two Guys (Bowen Yang and Matt Rogers). Five Rings (you know, from the Olympics logo). One essential podcast for the 2026 Milan-Cortina Winter Olympics. Bowen Yang (SNL, Wicked) and Matt Rogers (Palm Royale, No Good Deed) of Las Culturistas are back for a second season of Two Guys, Five Rings, a collaboration with NBC Sports and iHeartRadio. In this 15-episode event, Bowen and Matt discuss the top storylines, obsess over Italian culture, and find out what really goes on in the Olympic Village.
iHeartOlympics: The Latest
Listen to the latest news from the 2026 Winter Olympics.
Milan Cortina Winter Olympics
The 2026 Winter Olympics in Milan Cortina are here and have everyone talking. iHeartPodcasts is buzzing with content in honor of the XXV Winter Olympics We’re bringing you episodes from a variety of iHeartPodcast shows to help you keep up with the action. Follow Milan Cortina Winter Olympics so you don’t miss any coverage of the 2026 Winter Olympics, and if you like what you hear, be sure to follow each Podcast in the feed for more great content from iHeartPodcasts.