August 14, 2024 • 40 mins
After a year of opulent spending, the markets have begun to sour on big tech's $200 billion bet on generative AI, unfortunately timing with the delay of Nvidia's new AI-specialized "blackwell" chips. In this episode, Ed Zitron walks you through the pale horses of the AI apocalypse - and what this means for the tech industry at large.
LINKS: http://www.tinyurl.com/betterofflinelinks
Newsletter: wheresyoured.at
Reddit: http://www.reddit.com/r/betteroffline
Discord: chat.wheresyoured.at
Ed's Socials -
http://www.twitter.com/edzitron
https://bsky.app/profile/zitron.bsky.social
https://www.threads.net/@edzitron
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 (00:02):
All Zone Media. Hello, and welcome to Better Offline Live
and direct, coming at you from the middle of nowhere
the center of everywhere, and I'm your host ed zetron.
What else it feels like forever since I talked directly
(00:26):
to you, my dear listeners. So I'll get right to it.
And as a reminder, I'll be including links to everything
I'm talking about in the episode notes. And if you're
just catching up, all you really need to know here
is that open ai makes large language models like GPT
and of course the product chat GPT and throp it
makes claude a similar product, and that generative AI products
(00:47):
have yet to really prove a use case that justifies
them losing money on every single transaction. Let's begin. August second,
twenty twenty four was Black Friday for the AI boom,
as a week of rough earnings from big tech led
to what felt like the entire media industry asking one question,
is the AI bubble popping? And that's the question I'm
(01:08):
going to try and answer for you today and in
the next episode two, the Guardians sought to answer why
the big seven tech companies had been hit with AI
boomed outs. CNN asked has the AI bubble burst, and
The Atlantic suggested several months too late that the generative
AI revolution may indeed be a bubble. The Financial Times
reported hedge fund Elliott Management told investors that Nvidia was
(01:32):
a bubble, and Bloomberg reported that big tech had failed
to convince Wall Street that AI was paying off. It's
all part of a growing trend where people are suddenly
realizing what I've been saying for months, that this unprofitable,
energy hungry technology that creates mediocre outputs is not actually
the future. It's just another cloud do hickey. I had
(01:53):
not a great way of putting it, I realized, but
this is my nice way of saying. Generative AI isn't
completely useless, but the large language models they kind of
are based on the costs. These articles were for the
most part, talking about the sudden and violent declines and
the share prices of companies like Microsoft, Amazon, and Google,
which for the most part haven't really recovered and as
(02:15):
of about August tenth, are down significantly over the past month.
Although these articles didn't say anything particularly new or reveal
any individual missteps or scandals that might have prompted the slide,
they hinted to a broader awareness among Wall Street that
the AI ambitions of these companies and by which I
really mean generative AI, will require these massive upfront investments
(02:36):
and the payoff might not n actually be there, and
at least if it is, it's not going to be
there for a while. We're talking years, if not a decade,
and once the narrative gets settled, is very, very very
hard to move in. While a stock market isn't always rational,
check out Tesla for more. The one day sell off
of these stocks and the comments from analysts and industry
(02:57):
figures that followed, suggest that Wall Street is growing in
increasingly uncomfortable with the vast amounts of money required to
build and grow generative AI into whatever the fuck that's
meant to be. These companies, while still making over ten
billion dollars in profit, of course, referring to Microsoft, Google, Amazon,
and the like in the last quarter alone, have also
spent an absolute shit ton of money on infrastructure to
(03:19):
capture the so called demand for cloud services from generative AI. However,
one little problem, none of them seem to actually be
making that much money from the thing they're investing in.
In the last fiscal year, Microsoft's capex the capital expenditures
was about fifty five point seven billion dollars, which is
up seventy five percent year over year, with more than
(03:42):
one third nineteen billion dollars spent in the last quarter
and in June thirtieth, twenty twenty four. This is reportedly
split fifty to fifty between infrastructure and technology, which suggests
an aggressive data center build out, with Chief financial officer
for Microsoft Amyhood saying that the company expects capital expenditures
to increase on a sequential basis, giving cloud and the
(04:02):
and I quote AI demand that, as I've repeatedly said,
is not really there. And what exactly does increasing on
a sequential basis mean in dollar terms? I have absolutely
no idea. It's vague and perhaps vague enough to rattle
the markets, particularly when Microsoft will be starting from an
already quite high valuation. We're still, Hood added, and I
(04:24):
quote Microsoft's earnings calles that AI related spend represented nearly
all of our total capital expenditures, with roughly half for
infrastructure needs that will support monetization over the next fifteen
years and beyond ugh. In essence, Microsoft spent nineteen billion
dollars in the last quarter on cloud and AI expenses
and has made it clear that it's not done spending
(04:44):
more money than it's ever spent before on a technology
that neither makes Microsoft nor the people paying them that
much money. It's very stupid and for context, Microsoft made
twenty two point zero four billion dollars in profits in
Q two twenty twenty five. Is this really worth sinking
an entire quarters worth of profits into? Let me give
(05:05):
it another way to look at it. Microsoft's net profit
margin has dropped from thirty nine point four four percent
in Q three twenty twenty three to thirty four point
four percent in Q two twenty twenty four, meaning that
it's taking home less money than it usually would because
they're really investing in this thing that only loses money.
It's very good, it's very good that this is happening.
This is what you love to see. The other cloud
(05:27):
providers aren't really doing much better. Google's capital expenditures are
forecast to be fifty billion dollars in twenty twenty four,
and it's spent eleven billion dollars in Q four twenty
twenty three, driven by and I quote mostly by technical
infrastructure meaning servers and data centers, and twelve billion dollars
was what they spent in Q one twenty twenty four.
(05:48):
The reason I'm not breaking things out much with Google
is because it hasn't been extremely guarded about its AI expenses,
probably because they're really high and not making any money
from it. Amazon's similarly guarded, with its capex last year
hanging somewhere around forty eight point four billion dollars. It's
spent thirty point five billion dollars so far in twenty
twenty four, an absolutely ridiculous amount considering its profit for
(06:11):
Q two twenty twenty four was thirteen point four to
eight billion. They're just sinking their profits into these things,
and there's really no sign that anything changes. Every hyperscaler
has said that they intend to keep spending all of
this money on AI, and I haven't even mentioned companies
like Oracle, which expects to spend ten billion dollars on
cloud infrastructure this year, with much of that new capacity
(06:32):
going to support Microsoft and Google Cloud. I don't know
what happens if they don't need it anymore. Anyone think
about anyway anyway, But no Black Friday. It was a
collective realization of the scale and the cost of AI
and the first signs that the markets are starting to
ask those annoying little questions about whether it's actually worth it.
(06:53):
Yet the real chaos, and one that comparatively speaking slipped
under the radar, came in the form of one of
my pale horses. Last month, I put out an episode
called top Culture where I suggested that the first signs
of the AI bubbles collapse would be in the failure
of a major AI company, though one not operating at
the scale of say Open Ai, and I specifically picked
(07:14):
one out character dot Ai, which raised one hundred and
fifty million dollars in funding, and the information had already
hinted might sell itself to one of the big tech companies. Now.
The reason I picked them is that their app which
allows you to talk to AI chatbots of Elon Musk
and Staragojo from jiu Jitsu Kaizen, the anime manga which
we all know I'm tired of waiting for. That's not
(07:37):
a business, by the way, and they never really had
a sustainable business model or really a meaningful product. And
there one hundred and fifty million dollar raise was the
first big stupid capital raise in the generative AI boom.
Early twenty twenty three, when nobody else could raise at all.
In my opinion, their continued existence, other than being a
disgusting insult to company building and startups everywhere, was proof
(07:59):
that the bubble existed, and their death or whatever form
this really is, is a sign that the tolerance for
bullshit is leaving the market, as my blunt force for
shadowing suggests. On Friday August second, and yes, most of
this stuff happened. On that Friday, Google said that it
would license Character AI's technology and hire the company's leaders,
Noam Shazia and Daniel Defritas, along with their research team
(08:22):
of thirty people, to work at Google's DeepMind AI division.
The fates of the other one hundred and forty employees
remain uncertain. I assume they're alive, but I imagine some
will remain on as staff as the character dot ai
app remains operational. It's a strange, sad end for a
company that never really had any business existing, and in
many ways feels like a con happening in broad daylight.
(08:45):
Shazir and Defritas are both former Google employees, having left
the company in twenty twenty one to create Character dot ai.
It's unclear whether given their previous employment, they'll be required
to wear the new GLO propeller hat that you get
if you're new to Google. I actually don't know if
they still do that. Email me is easy at better
offline dot com. That's the letter E than the letter
(09:05):
Z or Z for my British and Canadian fans. If
you know about this propeller or indeed can get me one,
I would very much like one anyway. While this is
being framed as a typical licensing employee poaching deal where
nothing acquisition adjacent has happened, it is actually an acquisition.
Google's paying two point five billion dollars to investors. Employee
(09:26):
stock options will vest, and that means when you get
stock in a company, it usually takes time for you
to actually earn it, so that you can't just take
it and leave it immediately, as I would, And that
will keep vesting until July twenty twenty six at the
acquisition rate of eighty eight dollars a share paid for
by the money from the licensing deal. That is not
an acquisition from Google. But after that point your shit
(09:47):
out of luck. It's no longer guaranteed. This is a
terrible situation. All around for everyone other than the earliest investors.
It's another great situation where the bad guys win. In
essence anyone who has options that aren't fully there. By
July twenty twenty six, maybe shit out of luck and
character AI's original funding valued them a billion dollars, making
this a situation where and Resent Horowitz aw a one
(10:09):
point five x return on their investment in a company
that never really did anything. Employees kind of got screwed.
Maybe they didn't, It's unclear, But what is clear is
that as of now, character ai is effectively dead. The
company will shift from using their own models their literal
only thing that they did, to publicly available open source
large language models like metas Lama, and the original engineering
(10:32):
team is effectively gone. The company will likely shamble a
long life lessly until it curls up in the corner
like an old cat, Except an old cat has more
purpose and meaning in the universe than the bullshit chatbot company.
So why does this matter? On one level, the death
of character dot ai is an indicator of the unsustainability
of many of these generative AI applications and companies. Even
(10:55):
with one hundred and fifty million dollars in funding, which
is a decent amount of money for a normal company.
Character likely couldn't keep the lights on for very long,
which kind of made it necessary for them to be
absorbed or acquired before they shut down and embarrassed everyone involved.
It's also just like a Charlatan's Olympics. It's this company sucked.
It never made money, it didn't really do anything that different,
(11:19):
but they just get bailed out. These fucking start up people.
They always go on about the meritocracy. They're always going
on about, Oh, it's about working hard, It's not about
your circumstances. These jack holes, these fucking idiots. They got
to sell their bullshit company to Google, a company they
left the found it. It's just so annoying. And this
agreement also strongly resembles an earlier one between Inflection AI
(11:42):
and Microsoft, where Microsoft bought Inflections technology and team without
actually acquiring its equity, so the actual stock in the company.
Founded by DeepMind co founder mister Fassuliman and LinkedIn co
founder Reid Hoffman in twenty twenty two, Inflection was, in
Layman's terms, intended to be a more emotion aware version
of CHATGBT just one year after its creation, they'd raised
(12:04):
one point three billion dollars at evaluation of four billion.
As I noted previously, I believe that Microsoft's acquisition of
Inflection was an attempt to avoid regulatory scrutiny by structuring
an acquisition as a transfer of tech and talent, rather
than buying the company where the tech and the talent live.
And yet I'm also convinced that Inflection would struggle to
(12:24):
last out on its own, even with its smaller user
base and with a one point three billion dollar war chest,
and eventually it would have had to get acquired or
just die. Regardless, I'm beginning to see it patten AI's
point of failure is shifting, or more accurately centralizing. At
one point, the burden of this entire generative AI fast
(12:46):
was shouldered by a large and disparate group of startups
and investors, and it's now moving to a few shoulders,
those of giants like Microsoft, Google, Amazon, Meta and to
a much lesser extent, Apple. The two most prevalent large
language models outside of course of Metaslama open AI's GPT
and Anthropics claud are effectively big tech welfare recipients receiving
(13:06):
billions of dollars in cloud credits to run their extremely
expensive models without having to build their own infrastructure. It's
also kind of a con because that's not real money,
that's chuckie geese tokens, and how are those valued. Ten
billion dollars goes to open Ai, mostly in cloud credits,
and they get equity in return. That doesn't make it.
I can't invest in companies and air miles. Why can
(13:27):
Microsoft do it? Anyway? Microsoft invested billions of dollars in
open Ai, and Google and Amazon are propping Upanthropic too,
with the latter completing its four billion dollar investment in
the ai company earlier this year. I need to be
clear that open Ai and Anthropic are wholly reliant on
the mercies of a handful of trillion dollar companies who
are themselves at the mercy of the rot economy in
(13:49):
the public markets, and especially in the case of open
AI's relationship with Microsoft, they're not just dependent on big
tech firms financial support, but also on their ability to
procure and build the actual infratre structure of these products.
Open Ai does not have their own service, neither does Anthropic.
Perhaps they have some a few, but the majority of
their actual processing is done by big tech firms that
(14:13):
paid them in Chuck e GE's tokens in cloud credits
to run their services. Who knows how you even amortize that.
It's completely insane. It drives me insane that they're able
to do this every time they find a horrible little
thing to do. But me, I'm at the mercy of
one of the fine products that you'll hear after this
ad break, one that will no doubt echo my exact thoughts, feelings,
(14:36):
and desires. Buy their things or don't. That's up to you,
and we're back. As I was previously saying, Friday, August second,
twenty twenty four is a rough day for AI, and
(14:57):
it somehow got worse. That's SAE. The Information reported that
Nvidia has reportedly told Microsoft and another unnamed cloud customer
that its next generation Blackwell chips, which are designed primarily
for accelerating AI compute tasks, will be delayed by three
months due to an unspecified design floor identified by its
contract foundry TSMC. It's entirely possible that this floor takes
(15:21):
a little more than three months to resolve. Semiconductor manufacturing
is really, really, really really hard and issues that initially
see minor can rapidly spiral out of control. On top
of that, you can't rush them in the same way
you can with software. You can't just throw a bunch
more money. TSMC has plenty. But on top of that,
(15:42):
this isn't a money problem. This is one of those
scientific ones that you can't just skip over. And with
black Whill representing a major leap forward in capabilities both
in the terms of power efficiency and sheer compute power,
this is a huge blow to open ai and it's
many many competitors, by which I mean it's other competitor,
which is also bankrolled by a big tech sugar daddy.
(16:05):
It's also likely to delay some of open AI's grander
ambitions for an indeterminate amount of time, which in turn
kind of dampens its appeal to potential investors. And to
be clear, open ai basically has to raise in the
next six months, as the Information reported that open ai
could lose as much as five billion dollars this year alone.
I will get to that later, but just to be clear,
(16:27):
that's not five billion in losses plus revenue. That is,
after you take the money out that they made. It's
really bad it's really really bad. Though. It's all so
very bad. I'm not sure why more people aren't freaking out.
I'm mostly chilling just because they don't really care if
they live or die. I feel bad for the people
(16:47):
with the jobs, but I'm rambling. We'll get back to it,
and this is a problem that expands to Google and
Microsoft too. In Nvidious Blackwell chips was supposed to be
a major technological leap for generative AI, designed with both
influence the way that large language models generate answers in
training in mind, and thus providing vastly more compute parerent
energy efficiency necessary to make this shit move. But as
(17:10):
of now, these chips have been delayed, with their availability
not expected into the first quarter of twenty twenty five
at the earliest, and the Information reports that open Ai
probably won't get access to them until March twenty twenty four,
and Nvidia might be forced to do new test runs
with its foundry partner before scaling up to mass production.
One particularly worrying quote, and I kind of just hinted
(17:32):
in it there, was that Microsoft managers had planned to
make Blackwell powered servers available to open Ai by January,
but like I said, may need to plan for March
or early spring, and that was said by a person
with knowledge of the situation. Beautiful name the reason that's bad, Well,
it's not really just one. First, open ai desperately needs
(17:52):
something new. They need something that will show both investors
and the media that open ai is building something meaningful.
And while it's possible it will be able to deliver
GPT five, which is their next model, in the near
term future, even potential customers don't believe the jump will
be significant enough from GPT four. Second, for reasons I've
alluded to at the start of this episode, investors are
(18:13):
getting a touch itchy with the AI hype boom, and
more specifically with the giant tech companies that are bankrolling it.
The street needs to see results, or at the very
least ap plausible case for the future profitability and marketability
of generative AI, a thing that it's genuinely bad at.
In vidious Blackwell chips could help open ai models run
faster and ingest training data faster, as well as, even
(18:36):
if it's not possible, fooling investors into believing that open
ai had the latest tech and thus could build the
average general intelligence that sam Ortmand's been lying about and
allow him to continue his con regardless, the article from
the Information put it very plainly open AI only gets
access to the latest technology as fast as Microsoft permits,
(18:58):
or as fast as Microsoft is able to if I
did on their own, even one of these events would
be a deeply worrying sign of the bubble popping, and
together they threatened to begin a collapse that I've been
predicting since March, where I said that AI companies had
about three quarters to prove themselves and I quote savaging
the revenues of the biggest companies in tech when things
don't work out. But I've realized now that it isn't
(19:20):
really super useful to attach things to time. Though I
stand by bud prediction, and thus I think it's more
illustrative to suggest what the terms of the bubble popping
actually are, So let's define them. What do I mean
when I talk about the bubble popping? For the sake
of clarity, I'm defining it as the major cloud companies
reducing capital expenditures related to generative AI in a public
(19:43):
and significant manner, as in an actual statement from satching
Adelaer at Microsoft or Sun Duppishi at Google or one
of the major large language model companies Anthropic or open
AI collapsing in some way. No cheap thrills, no half measures.
We're not calling this fucking thing until it actually pops,
and it will not pop into one or both of
these companies go tits up. To be clear, when I
(20:06):
talk about Anthropic and Open Air collapsing, I don't actually
mean in the Enron style way. These companies are probably
not going to fall apart, shut down, close their doors suddenly.
It could be that they're absorbed into another company on
unfavorable terms, you know, like character AI and inflection, like
(20:26):
the things that just happened. And in that case, I
think investors are going to take a massive loss. Open
AI or Anthropic might be forced to radically limit their operations,
either by shutting down their free chatbots or limiting access
to enterprise customers, or by slowing down development massively, which
by the way, will be death's embrace. It will be over.
(20:47):
We could see massive layoffs, or we could see these
companies pivot to a less capital intensive business model, such
as licensing their patterns and IP to other technologies or companies,
but not really offering an actual tangible product to consumers
or business customers, as opposed to what they're doing today,
which is offering a tangible thing that doesn't really doom
that much. I'm rambling again. Yeah, I can imagine all
(21:08):
sorts of terminal endings for Generative AI is big two,
and some of them aren't even legally actionable. But I
believe that the collapse or absorption of these companies is
the one critical sign to look for. Open ai began
this hype cycle and Sam Altman is definitely the pt
barnum of the large language model circus, and open ai
has absorbed more money and attention than any other startup
(21:29):
of the last few years, and outside of Uber, I
think they might be getting the most all time. God,
I hate it. It's symbolic of the excess in the
waste of the generative AI boom and it's death or
as I mentioned, some alternative collapse is the sign that
we're done here in the same way that FTX signaled
the end of the cryptocurrency boom. And if open AI's
collapse marks the apocalypse for generative AI, I believe it
(21:52):
will be accompanied by one or several of the following
pale horses, real tangible signs that things are falling apart,
things you can send to your friends, family and laugh at.
Let's take a look, shall we. Number one, Any significant
price changes by any large language model company are a
bad sign. Desperate times require desperate measures, and any oscillation
(22:13):
in pricing is a bad sign. We could see a
race to the bottom, which is kind of already happening
with model developers releasing increasingly cheaper options to grow revenue
and build their customer bases, which is an idea that
only makes sense if the underpinning technology is profitable, which
Generative Ai is not, and it's already kind of begun
with open AI's reduced price GPT four oh mini in
(22:33):
Google Gemini's Google Flash one point five I believe it's called,
and they just reduce that pricing to compete with GPT
four oh mini's cheapest option. This is totally unsustainable, which
means it fits right in with the rest of the
hype cycle. Conversely, if prices start increasing, this is a
sign that the company is getting really, really desperate and
(22:53):
have to find a way to start recouping the massive
costs of the unsustainable shit show they've been participating in well.
This I think is probably one of the later ones.
If they do this, if they even get to it,
it's just the sign that things are falling apart. I
also think we're going to see stories about general discord
in AI investment. Any articles you see about investors fleeing
(23:13):
private aideals as they did with the metaverse, or a
sign that things are falling apart. And the things I'm hearing,
by the way, is that this has already begun, but
that's anecdotal. Don't take my word for it. Just watch
the media. We might also see stories about open ai
having trouble raising money. Up until now, there haven't been
any rumors about them having trouble, despite the fact that
they're expected to lose like I mentioned, five billion dollars
(23:36):
in twenty twenty four. They need to raise billions of
dollars and they have to do it soon. And if
they can't raise, nobody can. But moving on, there's another
harbinger to look out for, and it would be evidence
that Google or Microsoft is reducing the capex as discussed,
because it's really important to realize that venture capital and
Silicon Valley really are not They're not the ones propping
(23:58):
this up. They might be the year leaders, they might
be the ones talking about how important is But Google
and Microsoft are the ones holding this up, and Amazon
as well. They're the ones who are bankrolling Anthropic and
open Ai. Microsoft as I will get to basically owns
open ai. If those companies decide they don't want this
to continue, it will stop. Venture capital cannot afford to
(24:20):
keep open ai alive. But moving on, you should keep
out and beady little I for any discord within any
of the major players in generative AI. But I'm not
talking about open ai and Anthropic, talking about companies like
Scale Ai, who are training data company that's raised over
one on a half billion dollars and cohere, another large
language model company that recently raised another four hundred and
(24:41):
fifty million dollars from Nvidia and Salesforce and Salesforce up.
They've been talking about doing AI for like a decade.
I don't know what product they sell. I'd watch those
two very carefully. They're seen as stable, revenue generating yet
as discussed, unprofitable and sillery firms that will likely need
to raise again in the next twelve months. If we
(25:02):
hear about problems such as employees being unhappy, or trouble's
fundraising or making money. This means that people on the
inside are making nasty little noises because they know things
are falling apart. And the same, by the way, goes
for any discord in open Ai or Anthropic. This one's obvious.
If we hear there are problems they're worried, or if
they're getting worried, if they can't raise money, if they're
(25:24):
losing people, which is kind of already happening, if they
can't seem to make a deal work, if they're having
trouble progressing their software. Anything about these companies is probably
a bad sign. If I were you, I'd keep an
eye on the Washington Post. Natasha Tiku's done great work there,
as the people at Shera A. V. Day and Jeffrey Fowler.
But Natasha's done some great work. Or mister Altman, the
(25:47):
Information has got some great people doing work on this
Reuter's Tech Crunch for or film media. They're the ones
that keep an eye out for. We've already seen a
little bit of this happening already. A few days ago,
open Ai lost two of its co founders, John Shulman,
left for competitor anthropic and chief operating officer Greg Brockman
has gone on something he's calling extended leave. As on aside,
(26:10):
this is objectively one of the funniest things it's anyone
has ever done in tech. Yeah, man, nothing big going on.
Enjoy a vacation, mate, Not like the entire industry is
falling apart, not like the bubble might be popping. Enjoy
carbo Anyway. As things get especially bad, I think we're
going to see discussions of layoffs based on people I've
talked to. Generative AI companies are paying these massive salaries
(26:33):
in stock and cash, and they're not run particularly efficiently
because they are really driven by dreamers, by which I
mean conn artists. If AI jobs are no longer these
cushy little Heidi holes for the most expensive engineers in
the valley, things are going to sour fast because they
need the best to do this mediocre crab that you
see every day. But I'd argue the biggest and most
(26:54):
desperate of these pale horses would be some sort of big,
stupid magic trick. As these companies get desperate, expect someone,
especially Open AI, to try and show something new and
crazy as a means of trying to turn this narrative around.
When or if this happens, look very carefully at what
they say about the products availability, what it can actually do,
(27:16):
or who they show it to. Is it available publicly?
Are there limitations on it? And are there any weird
terms and conditions to look out for. If open ai
gets desperate, it may move up the public launch of
soro It's generative video product. Doing those really only going
to cause more problems. There isn't a chance in hell
that this thing is profitable, and I'm one hundred percent
(27:37):
confident that it's much much, much, much more expensive than
the video and the text based generation tools they already have,
and I imagine its visual ginconsistencies and hallucinations would make
for some entertaining content for YouTubers and tech reporters, which
will mean that they just have hundreds thousands, hundreds of thousands,
perhaps the people requesting shitty videos off of this thing,
(27:58):
burning servers quite literally in this case, losing the money.
And for what how much can they charge for this
thing because it's not going to be enough. The common
thread between all of these events is that they're all
expressions of desperation fear and a total lack of confidence
in the underlying company. So far, and this is Napkin maths,
(28:19):
I'd estimate that a total of two hundred billion dollars
has been spent to get Generative AI to this point
in infrastructure and funding and energy in so many different
meaningless ways, or to get us to the point that
we have a tool that's really good at generating things
that aren't as good as what a humor could make,
and only sometimes are they good enough to actually use.
To be clear, we're not quite at the bubble popping yet.
(28:40):
The reason I'd chose the collapse of open ai as
the event is because it will mean that Microsoft decided
to cut them off and that there wasn't enough banker
or venture capital interest to prop them up further. Open
AI's collapse would be both financial and symbolic, the sign
that the valley would let a company die and that
this idea was not good enough for everybody to stake
their fetes on it. In any case, things could change,
(29:03):
the bubble could stay inflated. But while you're thinking about how,
why don't you take a breather to hear from someone
one of our wonderful beautiful, precisely chosen advertisers that you
simply must pay money to, and we're back now. I
(29:31):
don't want to give you any homework. I no, you're
currently holding up a convenience store or maybe trying to
get the hub cap off of a car wheel. You
hear me speak enough. But a few weeks ago I
wrote a newsletter called how does open ai Survive? And
eight thousand or so word analysis of how difficult it's
going to be to keep that company going at its
current burn rate, to simmer it down as quickly as possible.
Open ai costs about eight point five billion dollars a
(29:53):
year to run and only makes about three and a
half billion to four and a half billion a year
in revenue, a number, by the way that I'm that
does not sound right, but it's what's being reported. Open
AI's costs are increasingly nearly and generative AI, as I've
said a few times, is deeply unprofitable and unsustainable. Making
open AI's models better, which doesn't necessarily mean they're more
(30:13):
valuable and more useful, also costs hundreds of millions, if
not billions, of dollars. As a result, open ai needs
to raise more money than anybody has ever raised in
the history of the valley in the next twelve to
twenty four months, or launch a products so significant that
it blows chet GPT out of the water and actually
is profitable. And these things are unlikely but possible. I
(30:36):
guess Microsoft, Google, and Amazon could somehow change the narrative,
but doing so too would require some sort of technological
breakthrough or a product, a very obvious product that would
make people money, more money than is being spent on
the thing. And the markets are also capricious, and thus
I really wouldn't count on there being a narrative turnaround,
(30:57):
especially with the shifting sands of a vestor sentiment. As
I've mentioned previously. On August second, Elliott Management, the hedge
fund with over seventy billion dollars in assets under management,
said in a letter to clients that AI is overhyped
and the Nvidia is a bubble, adding that AI had
not delivered value commensurate with the hype. While it's possible
there are others that dissent against the rapidly forming consensus
(31:19):
against generative AI to prop things up, which I do
think is possible, as especially if you deepen the whole
with Microsoft or and Video. It's worth putting Elliott Management's
note in a broader context. It's only the latest voice
in a loud or at least increasingly louder, chorus of
critics whose members include esteamed, analyst houses, and some of
(31:39):
the largest investment banks in the world. Gold and Sax
put out a report in July that you've heard in
the pop culture episode that said generative AI was too
expensive and didn't solve the complex problems that would need
to justify their expenses. At the end of July, Gartner
put out a report predicting that thirty percent of generative
AI products would be abandoned entirely after a proof of
(32:00):
concept by end of twenty twenty five, and The Washington
Post reports that Barklay's bank thinks that and I quote
Wall Street and Lists are expecting big tech companies to
spend around sixty billion dollars a year on developing AI
models by twenty twenty six, but reap only twenty billion
dollars a year in revenue from AI by that point.
The lastra is gone there for you. That does say revenue,
(32:21):
that doesn't say profim and by the way, that's only
a billion dollars more than Microsoft's entire capital expenditures in
the last quarter alone. Microsoft also has a storied history
of pumping and dumping ideas for Microsoft. Augmented reality was
and I quote, an absolute breakthrough in twenty nineteen before
it was quietly shoved in a corner with mass layoffs
(32:42):
and a business line enclosures a few years later in
twenty twenty one, satchur Innidella, their CEO, couldn't overstate how
much of a breakthrough the metaverse was yet two years later,
and big props the Preston Growler of Computer World for
saying this in February twenty twenty three, he laid off
most of the people involved in everything on artificial intelligence
and that everything thing that's a headline from Wired magazine.
(33:06):
Very good. It's also important to remember that Microsoft effectively
owns open ai as part of their twenty nineteen funding round,
which I believe was a billion dollars. Microsoft is full
access to all of open AI's intellectual property and research,
which makes thorny little question why at this point would
Microsoft give open ai any more money other than to
(33:27):
save face. That's a very real possibility that Satch in
the Dela is surrounded by yes men who don't know
what they're talking about. Microsoft also has access to sell
open ais and I quote pre agi products, which is
all of their products right now, and full access to
all of their research, as well as and I quote
certain rights to open AI's intellectual property. Unless open ai
(33:47):
is capable of delivering something meaningfully different, something that Dela
can wave in front of analysts faces and show them
that they can automate all of these jobs away, its
existence is just another cost center and one that could
get eliminated in a tough quarter. And to be clear,
saturn Adella has been extremely frank about his leverage over
open ai and it's effective ownership of their tech or
(34:08):
be it buffeted with the usual bullshit platitudes and niceties
that you'd expect from a shape shifting tech CEO. Monster.
In a late twenty twenty interview with PR person Kara Swisher,
sach an Adella said, if open ai disappeared tomorrow, I
don't want any customer of ours to be worried about it,
quite honestly, because we have all the rights to continue
the innovation, not just to serve the product, but we
(34:28):
can go and just do what we were doing in
partnership ourselves. We have the people, we have the compute,
we have the data, we have everything. Mister Nadella, that
is an insane thing to say, and the kind of
thing you'd only say to someone like Kara Swisher. Swisher
should have made that the headline that Microsoft effectively owns
open ai. I miss this at the time I found
(34:49):
out about this quote, like a week or two ago.
It's crazy to me. It's crazy to me that this
was said, and it's not front page news. Microsoft owns
open ai. Who gives a shit about the nonprofit structure.
But it actually gets better because he followed up that
quote with this. Microsoft's investment of thirteen billion dollars in
open ai gives us significant rights, as I said, and
(35:12):
also this thing, it's not hands off, right. We are
in there, we are below them, we are above them,
we are around them. We do the kernel optimizations. Very
funny if you consider the CrowdStrike thing. By the way
we build tools, we build the infrastructure. So that's why
I think a lot of the industrial analysts are saying,
oh wow, it's really a joint project between Microsoft and
open ai. The reality is we are as I said,
(35:34):
very self sufficient in all of this. Well, I'm not
saying it's for certain. Microsoft can very clearly if they
so please, simply drop open AI and keep all of
their stuff. Nevertheless, as the market's sour and jennit of AI,
I think it's time for big tech to prove something.
It's time for them to prove that this wasn't all
a huge, big, stupid, fucking waste of time and money
(35:57):
to continue with their current paths. Summed up is sh
of Google and sach and Adella of Microsoft. As the
loudest of the publicly traded generative AI hype men, will
have to show everybody something remarkable. But even then they
will have to show how it actually makes money, and
if they can't, at least in the case of sach
and Adella, they can still walk away. As well as
(36:18):
in the best case scenario, these companies would either have
to dramatically reduce their costs, something they've really shown they're
not capable of doing and in fact have said they'll
do the opposite of or show ways to make generative
AI so significantly efficient that it somehow finally balances itself out.
What we're really waiting for now is for somebody to blink.
If none of the major cloud companies change course over
(36:40):
the next few months, we could potentially see this boom
continue for another quarter. And just to be clear, they
usually in lockstep. They love holding up each other's oligopolies
and monopolies. They love being a weird cartel that sells data.
To continue, though, they'll have to commit billions of dollars
in at the very least cloud credits for open ai
and Anthropic, as neither of those companies can survive without
(37:02):
further cash infusions. They'll also have to find some way
to make any of this useful and profitable, two challenges
the generative AI has never been able to deal with,
not even once. And I just as an aside, I've
had people email and saying, well, I use it for this.
I was on a show the other days that he
uses it for show notes. These are all fine whatever,
(37:23):
These are like little use cases. These people need this
to make a trillion dollars. Show notes aren't going to
do it. Shitty translations are not going to do it.
A picture of Garfield with a gun is not going
to do it. And big tech is going to have
to figure out a way to support these companies in
the face of rising disquiet among shareholders, who in the
most recent route of AI companies lost them about two
(37:46):
point six trillion dollars in market capitalization. As I pointed
out in the Shareholders Supremacy, investors don't really care about
the future or innovation. They just want profits and their
portfolios to get bigger. And they want to see the
number for profit go larger and the costs go lower,
and absolutely nothing else matters at all, which is why
(38:07):
they love generative AI at first, because if this actually
made a bunch of money, this is a way that
Microsoft could just route cloud costs from themselves to themselves.
I'll get to that in a future episode about monopolies, though,
And I think that if generative AI becomes an albatross
around the next of Google, Microsoft, and Amazon, which I
really believe it will, these companies are going to face
(38:27):
really nasty pressure to curtail their investments and their capex.
Investors will ask why these companies are spending a collective
one hundred two hundred billion dollars by the end of
twenty twenty five on a technology that likely won't drive
much revenue growth when it could be used for things
like stock dividends and stock buybacks, and I don't know
thrones for the demon lords that live underneath the offices.
(38:49):
I don't know how Microsoft truly makes money. That's a joke,
don't email me about it. But I believe in any case,
the next month is going to be critical to the
future of the AI boom, and I believe this is
a time for industry wide introspection and to really consider
everyone's roles and why this bubble existed in the first place.
In the next episode, I'm going to take a closer
(39:11):
look into open Aiy's likely demise, what it looks like,
and why the company's view avenues for salvation in fact
look like dead ends. And I'll get into how fucking
ridiculous this whole bubble is and what its potential collapse
spells for the future of the tech industry. Thank you
(39:35):
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 at Matasowski
dot com, M A T T O. S O. W
Ski 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.
(39:56):
I also really recommend you go to chat dot Where's
youoeed dot com to visit the discord, and go to
our slash Better Offline to check out I'll 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 Coolzonemedia dot com, or check us out
on the iHeartRadio app, Apple Podcasts, or wherever you get
(40:17):
your podcasts.
All Zone Media. Hello, and welcome to Better Offline Live
and direct, coming at you from the middle of nowhere
the center of everywhere, and I'm your host ed zetron.
What else it feels like forever since I talked directly
(00:26):
to you, my dear listeners. So I'll get right to it.
And as a reminder, I'll be including links to everything
I'm talking about in the episode notes. And if you're
just catching up, all you really need to know here
is that open ai makes large language models like GPT
and of course the product chat GPT and throp it
makes claude a similar product, and that generative AI products
(00:47):
have yet to really prove a use case that justifies
them losing money on every single transaction. Let's begin. August second,
twenty twenty four was Black Friday for the AI boom,
as a week of rough earnings from big tech led
to what felt like the entire media industry asking one question,
is the AI bubble popping? And that's the question I'm
(01:08):
going to try and answer for you today and in
the next episode two, the Guardians sought to answer why
the big seven tech companies had been hit with AI
boomed outs. CNN asked has the AI bubble burst, and
The Atlantic suggested several months too late that the generative
AI revolution may indeed be a bubble. The Financial Times
reported hedge fund Elliott Management told investors that Nvidia was
(01:32):
a bubble, and Bloomberg reported that big tech had failed
to convince Wall Street that AI was paying off. It's
all part of a growing trend where people are suddenly
realizing what I've been saying for months, that this unprofitable,
energy hungry technology that creates mediocre outputs is not actually
the future. It's just another cloud do hickey. I had
(01:53):
not a great way of putting it, I realized, but
this is my nice way of saying. Generative AI isn't
completely useless, but the large language models they kind of
are based on the costs. These articles were for the
most part, talking about the sudden and violent declines and
the share prices of companies like Microsoft, Amazon, and Google,
which for the most part haven't really recovered and as
(02:15):
of about August tenth, are down significantly over the past month.
Although these articles didn't say anything particularly new or reveal
any individual missteps or scandals that might have prompted the slide,
they hinted to a broader awareness among Wall Street that
the AI ambitions of these companies and by which I
really mean generative AI, will require these massive upfront investments
(02:36):
and the payoff might not n actually be there, and
at least if it is, it's not going to be
there for a while. We're talking years, if not a decade,
and once the narrative gets settled, is very, very very
hard to move in. While a stock market isn't always rational,
check out Tesla for more. The one day sell off
of these stocks and the comments from analysts and industry
(02:57):
figures that followed, suggest that Wall Street is growing in
increasingly uncomfortable with the vast amounts of money required to
build and grow generative AI into whatever the fuck that's
meant to be. These companies, while still making over ten
billion dollars in profit, of course, referring to Microsoft, Google, Amazon,
and the like in the last quarter alone, have also
spent an absolute shit ton of money on infrastructure to
(03:19):
capture the so called demand for cloud services from generative AI. However,
one little problem, none of them seem to actually be
making that much money from the thing they're investing in.
In the last fiscal year, Microsoft's capex the capital expenditures
was about fifty five point seven billion dollars, which is
up seventy five percent year over year, with more than
(03:42):
one third nineteen billion dollars spent in the last quarter
and in June thirtieth, twenty twenty four. This is reportedly
split fifty to fifty between infrastructure and technology, which suggests
an aggressive data center build out, with Chief financial officer
for Microsoft Amyhood saying that the company expects capital expenditures
to increase on a sequential basis, giving cloud and the
(04:02):
and I quote AI demand that, as I've repeatedly said,
is not really there. And what exactly does increasing on
a sequential basis mean in dollar terms? I have absolutely
no idea. It's vague and perhaps vague enough to rattle
the markets, particularly when Microsoft will be starting from an
already quite high valuation. We're still, Hood added, and I
(04:24):
quote Microsoft's earnings calles that AI related spend represented nearly
all of our total capital expenditures, with roughly half for
infrastructure needs that will support monetization over the next fifteen
years and beyond ugh. In essence, Microsoft spent nineteen billion
dollars in the last quarter on cloud and AI expenses
and has made it clear that it's not done spending
(04:44):
more money than it's ever spent before on a technology
that neither makes Microsoft nor the people paying them that
much money. It's very stupid and for context, Microsoft made
twenty two point zero four billion dollars in profits in
Q two twenty twenty five. Is this really worth sinking
an entire quarters worth of profits into? Let me give
(05:05):
it another way to look at it. Microsoft's net profit
margin has dropped from thirty nine point four four percent
in Q three twenty twenty three to thirty four point
four percent in Q two twenty twenty four, meaning that
it's taking home less money than it usually would because
they're really investing in this thing that only loses money.
It's very good, it's very good that this is happening.
This is what you love to see. The other cloud
(05:27):
providers aren't really doing much better. Google's capital expenditures are
forecast to be fifty billion dollars in twenty twenty four,
and it's spent eleven billion dollars in Q four twenty
twenty three, driven by and I quote mostly by technical
infrastructure meaning servers and data centers, and twelve billion dollars
was what they spent in Q one twenty twenty four.
(05:48):
The reason I'm not breaking things out much with Google
is because it hasn't been extremely guarded about its AI expenses,
probably because they're really high and not making any money
from it. Amazon's similarly guarded, with its capex last year
hanging somewhere around forty eight point four billion dollars. It's
spent thirty point five billion dollars so far in twenty
twenty four, an absolutely ridiculous amount considering its profit for
(06:11):
Q two twenty twenty four was thirteen point four to
eight billion. They're just sinking their profits into these things,
and there's really no sign that anything changes. Every hyperscaler
has said that they intend to keep spending all of
this money on AI, and I haven't even mentioned companies
like Oracle, which expects to spend ten billion dollars on
cloud infrastructure this year, with much of that new capacity
(06:32):
going to support Microsoft and Google Cloud. I don't know
what happens if they don't need it anymore. Anyone think
about anyway anyway, But no Black Friday. It was a
collective realization of the scale and the cost of AI
and the first signs that the markets are starting to
ask those annoying little questions about whether it's actually worth it.
(06:53):
Yet the real chaos, and one that comparatively speaking slipped
under the radar, came in the form of one of
my pale horses. Last month, I put out an episode
called top Culture where I suggested that the first signs
of the AI bubbles collapse would be in the failure
of a major AI company, though one not operating at
the scale of say Open Ai, and I specifically picked
(07:14):
one out character dot Ai, which raised one hundred and
fifty million dollars in funding, and the information had already
hinted might sell itself to one of the big tech companies. Now.
The reason I picked them is that their app which
allows you to talk to AI chatbots of Elon Musk
and Staragojo from jiu Jitsu Kaizen, the anime manga which
we all know I'm tired of waiting for. That's not
(07:37):
a business, by the way, and they never really had
a sustainable business model or really a meaningful product. And
there one hundred and fifty million dollar raise was the
first big stupid capital raise in the generative AI boom.
Early twenty twenty three, when nobody else could raise at all.
In my opinion, their continued existence, other than being a
disgusting insult to company building and startups everywhere, was proof
(07:59):
that the bubble existed, and their death or whatever form
this really is, is a sign that the tolerance for
bullshit is leaving the market, as my blunt force for
shadowing suggests. On Friday August second, and yes, most of
this stuff happened. On that Friday, Google said that it
would license Character AI's technology and hire the company's leaders,
Noam Shazia and Daniel Defritas, along with their research team
(08:22):
of thirty people, to work at Google's DeepMind AI division.
The fates of the other one hundred and forty employees
remain uncertain. I assume they're alive, but I imagine some
will remain on as staff as the character dot ai
app remains operational. It's a strange, sad end for a
company that never really had any business existing, and in
many ways feels like a con happening in broad daylight.
(08:45):
Shazir and Defritas are both former Google employees, having left
the company in twenty twenty one to create Character dot ai.
It's unclear whether given their previous employment, they'll be required
to wear the new GLO propeller hat that you get
if you're new to Google. I actually don't know if
they still do that. Email me is easy at better
offline dot com. That's the letter E than the letter
(09:05):
Z or Z for my British and Canadian fans. If
you know about this propeller or indeed can get me one,
I would very much like one anyway. While this is
being framed as a typical licensing employee poaching deal where
nothing acquisition adjacent has happened, it is actually an acquisition.
Google's paying two point five billion dollars to investors. Employee
(09:26):
stock options will vest, and that means when you get
stock in a company, it usually takes time for you
to actually earn it, so that you can't just take
it and leave it immediately, as I would, And that
will keep vesting until July twenty twenty six at the
acquisition rate of eighty eight dollars a share paid for
by the money from the licensing deal. That is not
an acquisition from Google. But after that point your shit
(09:47):
out of luck. It's no longer guaranteed. This is a
terrible situation. All around for everyone other than the earliest investors.
It's another great situation where the bad guys win. In
essence anyone who has options that aren't fully there. By
July twenty twenty six, maybe shit out of luck and
character AI's original funding valued them a billion dollars, making
this a situation where and Resent Horowitz aw a one
(10:09):
point five x return on their investment in a company
that never really did anything. Employees kind of got screwed.
Maybe they didn't, It's unclear, But what is clear is
that as of now, character ai is effectively dead. The
company will shift from using their own models their literal
only thing that they did, to publicly available open source
large language models like metas Lama, and the original engineering
(10:32):
team is effectively gone. The company will likely shamble a
long life lessly until it curls up in the corner
like an old cat, Except an old cat has more
purpose and meaning in the universe than the bullshit chatbot company.
So why does this matter? On one level, the death
of character dot ai is an indicator of the unsustainability
of many of these generative AI applications and companies. Even
(10:55):
with one hundred and fifty million dollars in funding, which
is a decent amount of money for a normal company.
Character likely couldn't keep the lights on for very long,
which kind of made it necessary for them to be
absorbed or acquired before they shut down and embarrassed everyone involved.
It's also just like a Charlatan's Olympics. It's this company sucked.
It never made money, it didn't really do anything that different,
(11:19):
but they just get bailed out. These fucking start up people.
They always go on about the meritocracy. They're always going
on about, Oh, it's about working hard, It's not about
your circumstances. These jack holes, these fucking idiots. They got
to sell their bullshit company to Google, a company they
left the found it. It's just so annoying. And this
agreement also strongly resembles an earlier one between Inflection AI
(11:42):
and Microsoft, where Microsoft bought Inflections technology and team without
actually acquiring its equity, so the actual stock in the company.
Founded by DeepMind co founder mister Fassuliman and LinkedIn co
founder Reid Hoffman in twenty twenty two, Inflection was, in
Layman's terms, intended to be a more emotion aware version
of CHATGBT just one year after its creation, they'd raised
(12:04):
one point three billion dollars at evaluation of four billion.
As I noted previously, I believe that Microsoft's acquisition of
Inflection was an attempt to avoid regulatory scrutiny by structuring
an acquisition as a transfer of tech and talent, rather
than buying the company where the tech and the talent live.
And yet I'm also convinced that Inflection would struggle to
(12:24):
last out on its own, even with its smaller user
base and with a one point three billion dollar war chest,
and eventually it would have had to get acquired or
just die. Regardless, I'm beginning to see it patten AI's
point of failure is shifting, or more accurately centralizing. At
one point, the burden of this entire generative AI fast
(12:46):
was shouldered by a large and disparate group of startups
and investors, and it's now moving to a few shoulders,
those of giants like Microsoft, Google, Amazon, Meta and to
a much lesser extent, Apple. The two most prevalent large
language models outside of course of Metaslama open AI's GPT
and Anthropics claud are effectively big tech welfare recipients receiving
(13:06):
billions of dollars in cloud credits to run their extremely
expensive models without having to build their own infrastructure. It's
also kind of a con because that's not real money,
that's chuckie geese tokens, and how are those valued. Ten
billion dollars goes to open Ai, mostly in cloud credits,
and they get equity in return. That doesn't make it.
I can't invest in companies and air miles. Why can
(13:27):
Microsoft do it? Anyway? Microsoft invested billions of dollars in
open Ai, and Google and Amazon are propping Upanthropic too,
with the latter completing its four billion dollar investment in
the ai company earlier this year. I need to be
clear that open Ai and Anthropic are wholly reliant on
the mercies of a handful of trillion dollar companies who
are themselves at the mercy of the rot economy in
(13:49):
the public markets, and especially in the case of open
AI's relationship with Microsoft, they're not just dependent on big
tech firms financial support, but also on their ability to
procure and build the actual infratre structure of these products.
Open Ai does not have their own service, neither does Anthropic.
Perhaps they have some a few, but the majority of
their actual processing is done by big tech firms that
(14:13):
paid them in Chuck e GE's tokens in cloud credits
to run their services. Who knows how you even amortize that.
It's completely insane. It drives me insane that they're able
to do this every time they find a horrible little
thing to do. But me, I'm at the mercy of
one of the fine products that you'll hear after this
ad break, one that will no doubt echo my exact thoughts, feelings,
(14:36):
and desires. Buy their things or don't. That's up to you,
and we're back. As I was previously saying, Friday, August second,
twenty twenty four is a rough day for AI, and
(14:57):
it somehow got worse. That's SAE. The Information reported that
Nvidia has reportedly told Microsoft and another unnamed cloud customer
that its next generation Blackwell chips, which are designed primarily
for accelerating AI compute tasks, will be delayed by three
months due to an unspecified design floor identified by its
contract foundry TSMC. It's entirely possible that this floor takes
(15:21):
a little more than three months to resolve. Semiconductor manufacturing
is really, really, really really hard and issues that initially
see minor can rapidly spiral out of control. On top
of that, you can't rush them in the same way
you can with software. You can't just throw a bunch
more money. TSMC has plenty. But on top of that,
(15:42):
this isn't a money problem. This is one of those
scientific ones that you can't just skip over. And with
black Whill representing a major leap forward in capabilities both
in the terms of power efficiency and sheer compute power,
this is a huge blow to open ai and it's
many many competitors, by which I mean it's other competitor,
which is also bankrolled by a big tech sugar daddy.
(16:05):
It's also likely to delay some of open AI's grander
ambitions for an indeterminate amount of time, which in turn
kind of dampens its appeal to potential investors. And to
be clear, open ai basically has to raise in the
next six months, as the Information reported that open ai
could lose as much as five billion dollars this year alone.
I will get to that later, but just to be clear,
(16:27):
that's not five billion in losses plus revenue. That is,
after you take the money out that they made. It's
really bad it's really really bad. Though. It's all so
very bad. I'm not sure why more people aren't freaking out.
I'm mostly chilling just because they don't really care if
they live or die. I feel bad for the people
(16:47):
with the jobs, but I'm rambling. We'll get back to it,
and this is a problem that expands to Google and
Microsoft too. In Nvidious Blackwell chips was supposed to be
a major technological leap for generative AI, designed with both
influence the way that large language models generate answers in
training in mind, and thus providing vastly more compute parerent
energy efficiency necessary to make this shit move. But as
(17:10):
of now, these chips have been delayed, with their availability
not expected into the first quarter of twenty twenty five
at the earliest, and the Information reports that open Ai
probably won't get access to them until March twenty twenty four,
and Nvidia might be forced to do new test runs
with its foundry partner before scaling up to mass production.
One particularly worrying quote, and I kind of just hinted
(17:32):
in it there, was that Microsoft managers had planned to
make Blackwell powered servers available to open Ai by January,
but like I said, may need to plan for March
or early spring, and that was said by a person
with knowledge of the situation. Beautiful name the reason that's bad, Well,
it's not really just one. First, open ai desperately needs
(17:52):
something new. They need something that will show both investors
and the media that open ai is building something meaningful.
And while it's possible it will be able to deliver
GPT five, which is their next model, in the near
term future, even potential customers don't believe the jump will
be significant enough from GPT four. Second, for reasons I've
alluded to at the start of this episode, investors are
(18:13):
getting a touch itchy with the AI hype boom, and
more specifically with the giant tech companies that are bankrolling it.
The street needs to see results, or at the very
least ap plausible case for the future profitability and marketability
of generative AI, a thing that it's genuinely bad at.
In vidious Blackwell chips could help open ai models run
faster and ingest training data faster, as well as, even
(18:36):
if it's not possible, fooling investors into believing that open
ai had the latest tech and thus could build the
average general intelligence that sam Ortmand's been lying about and
allow him to continue his con regardless, the article from
the Information put it very plainly open AI only gets
access to the latest technology as fast as Microsoft permits,
(18:58):
or as fast as Microsoft is able to if I
did on their own, even one of these events would
be a deeply worrying sign of the bubble popping, and
together they threatened to begin a collapse that I've been
predicting since March, where I said that AI companies had
about three quarters to prove themselves and I quote savaging
the revenues of the biggest companies in tech when things
don't work out. But I've realized now that it isn't
(19:20):
really super useful to attach things to time. Though I
stand by bud prediction, and thus I think it's more
illustrative to suggest what the terms of the bubble popping
actually are, So let's define them. What do I mean
when I talk about the bubble popping? For the sake
of clarity, I'm defining it as the major cloud companies
reducing capital expenditures related to generative AI in a public
(19:43):
and significant manner, as in an actual statement from satching
Adelaer at Microsoft or Sun Duppishi at Google or one
of the major large language model companies Anthropic or open
AI collapsing in some way. No cheap thrills, no half measures.
We're not calling this fucking thing until it actually pops,
and it will not pop into one or both of
these companies go tits up. To be clear, when I
(20:06):
talk about Anthropic and Open Air collapsing, I don't actually
mean in the Enron style way. These companies are probably
not going to fall apart, shut down, close their doors suddenly.
It could be that they're absorbed into another company on
unfavorable terms, you know, like character AI and inflection, like
(20:26):
the things that just happened. And in that case, I
think investors are going to take a massive loss. Open
AI or Anthropic might be forced to radically limit their operations,
either by shutting down their free chatbots or limiting access
to enterprise customers, or by slowing down development massively, which
by the way, will be death's embrace. It will be over.
(20:47):
We could see massive layoffs, or we could see these
companies pivot to a less capital intensive business model, such
as licensing their patterns and IP to other technologies or companies,
but not really offering an actual tangible product to consumers
or business customers, as opposed to what they're doing today,
which is offering a tangible thing that doesn't really doom
that much. I'm rambling again. Yeah, I can imagine all
(21:08):
sorts of terminal endings for Generative AI is big two,
and some of them aren't even legally actionable. But I
believe that the collapse or absorption of these companies is
the one critical sign to look for. Open ai began
this hype cycle and Sam Altman is definitely the pt
barnum of the large language model circus, and open ai
has absorbed more money and attention than any other startup
(21:29):
of the last few years, and outside of Uber, I
think they might be getting the most all time. God,
I hate it. It's symbolic of the excess in the
waste of the generative AI boom and it's death or
as I mentioned, some alternative collapse is the sign that
we're done here in the same way that FTX signaled
the end of the cryptocurrency boom. And if open AI's
collapse marks the apocalypse for generative AI, I believe it
(21:52):
will be accompanied by one or several of the following
pale horses, real tangible signs that things are falling apart,
things you can send to your friends, family and laugh at.
Let's take a look, shall we. Number one, Any significant
price changes by any large language model company are a
bad sign. Desperate times require desperate measures, and any oscillation
(22:13):
in pricing is a bad sign. We could see a
race to the bottom, which is kind of already happening
with model developers releasing increasingly cheaper options to grow revenue
and build their customer bases, which is an idea that
only makes sense if the underpinning technology is profitable, which
Generative Ai is not, and it's already kind of begun
with open AI's reduced price GPT four oh mini in
(22:33):
Google Gemini's Google Flash one point five I believe it's called,
and they just reduce that pricing to compete with GPT
four oh mini's cheapest option. This is totally unsustainable, which
means it fits right in with the rest of the
hype cycle. Conversely, if prices start increasing, this is a
sign that the company is getting really, really desperate and
(22:53):
have to find a way to start recouping the massive
costs of the unsustainable shit show they've been participating in well.
This I think is probably one of the later ones.
If they do this, if they even get to it,
it's just the sign that things are falling apart. I
also think we're going to see stories about general discord
in AI investment. Any articles you see about investors fleeing
(23:13):
private aideals as they did with the metaverse, or a
sign that things are falling apart. And the things I'm hearing,
by the way, is that this has already begun, but
that's anecdotal. Don't take my word for it. Just watch
the media. We might also see stories about open ai
having trouble raising money. Up until now, there haven't been
any rumors about them having trouble, despite the fact that
they're expected to lose like I mentioned, five billion dollars
(23:36):
in twenty twenty four. They need to raise billions of
dollars and they have to do it soon. And if
they can't raise, nobody can. But moving on, there's another
harbinger to look out for, and it would be evidence
that Google or Microsoft is reducing the capex as discussed,
because it's really important to realize that venture capital and
Silicon Valley really are not They're not the ones propping
(23:58):
this up. They might be the year leaders, they might
be the ones talking about how important is But Google
and Microsoft are the ones holding this up, and Amazon
as well. They're the ones who are bankrolling Anthropic and
open Ai. Microsoft as I will get to basically owns
open ai. If those companies decide they don't want this
to continue, it will stop. Venture capital cannot afford to
(24:20):
keep open ai alive. But moving on, you should keep
out and beady little I for any discord within any
of the major players in generative AI. But I'm not
talking about open ai and Anthropic, talking about companies like
Scale Ai, who are training data company that's raised over
one on a half billion dollars and cohere, another large
language model company that recently raised another four hundred and
(24:41):
fifty million dollars from Nvidia and Salesforce and Salesforce up.
They've been talking about doing AI for like a decade.
I don't know what product they sell. I'd watch those
two very carefully. They're seen as stable, revenue generating yet
as discussed, unprofitable and sillery firms that will likely need
to raise again in the next twelve months. If we
(25:02):
hear about problems such as employees being unhappy, or trouble's
fundraising or making money. This means that people on the
inside are making nasty little noises because they know things
are falling apart. And the same, by the way, goes
for any discord in open Ai or Anthropic. This one's obvious.
If we hear there are problems they're worried, or if
they're getting worried, if they can't raise money, if they're
(25:24):
losing people, which is kind of already happening, if they
can't seem to make a deal work, if they're having
trouble progressing their software. Anything about these companies is probably
a bad sign. If I were you, I'd keep an
eye on the Washington Post. Natasha Tiku's done great work there,
as the people at Shera A. V. Day and Jeffrey Fowler.
But Natasha's done some great work. Or mister Altman, the
(25:47):
Information has got some great people doing work on this
Reuter's Tech Crunch for or film media. They're the ones
that keep an eye out for. We've already seen a
little bit of this happening already. A few days ago,
open Ai lost two of its co founders, John Shulman,
left for competitor anthropic and chief operating officer Greg Brockman
has gone on something he's calling extended leave. As on aside,
(26:10):
this is objectively one of the funniest things it's anyone
has ever done in tech. Yeah, man, nothing big going on.
Enjoy a vacation, mate, Not like the entire industry is
falling apart, not like the bubble might be popping. Enjoy
carbo Anyway. As things get especially bad, I think we're
going to see discussions of layoffs based on people I've
talked to. Generative AI companies are paying these massive salaries
(26:33):
in stock and cash, and they're not run particularly efficiently
because they are really driven by dreamers, by which I
mean conn artists. If AI jobs are no longer these
cushy little Heidi holes for the most expensive engineers in
the valley, things are going to sour fast because they
need the best to do this mediocre crab that you
see every day. But I'd argue the biggest and most
(26:54):
desperate of these pale horses would be some sort of big,
stupid magic trick. As these companies get desperate, expect someone,
especially Open AI, to try and show something new and
crazy as a means of trying to turn this narrative around.
When or if this happens, look very carefully at what
they say about the products availability, what it can actually do,
(27:16):
or who they show it to. Is it available publicly?
Are there limitations on it? And are there any weird
terms and conditions to look out for. If open ai
gets desperate, it may move up the public launch of
soro It's generative video product. Doing those really only going
to cause more problems. There isn't a chance in hell
that this thing is profitable, and I'm one hundred percent
(27:37):
confident that it's much much, much, much more expensive than
the video and the text based generation tools they already have,
and I imagine its visual ginconsistencies and hallucinations would make
for some entertaining content for YouTubers and tech reporters, which
will mean that they just have hundreds thousands, hundreds of thousands,
perhaps the people requesting shitty videos off of this thing,
(27:58):
burning servers quite literally in this case, losing the money.
And for what how much can they charge for this
thing because it's not going to be enough. The common
thread between all of these events is that they're all
expressions of desperation fear and a total lack of confidence
in the underlying company. So far, and this is Napkin maths,
(28:19):
I'd estimate that a total of two hundred billion dollars
has been spent to get Generative AI to this point
in infrastructure and funding and energy in so many different
meaningless ways, or to get us to the point that
we have a tool that's really good at generating things
that aren't as good as what a humor could make,
and only sometimes are they good enough to actually use.
To be clear, we're not quite at the bubble popping yet.
(28:40):
The reason I'd chose the collapse of open ai as
the event is because it will mean that Microsoft decided
to cut them off and that there wasn't enough banker
or venture capital interest to prop them up further. Open
AI's collapse would be both financial and symbolic, the sign
that the valley would let a company die and that
this idea was not good enough for everybody to stake
their fetes on it. In any case, things could change,
(29:03):
the bubble could stay inflated. But while you're thinking about how,
why don't you take a breather to hear from someone
one of our wonderful beautiful, precisely chosen advertisers that you
simply must pay money to, and we're back now. I
(29:31):
don't want to give you any homework. I no, you're
currently holding up a convenience store or maybe trying to
get the hub cap off of a car wheel. You
hear me speak enough. But a few weeks ago I
wrote a newsletter called how does open ai Survive? And
eight thousand or so word analysis of how difficult it's
going to be to keep that company going at its
current burn rate, to simmer it down as quickly as possible.
Open ai costs about eight point five billion dollars a
(29:53):
year to run and only makes about three and a
half billion to four and a half billion a year
in revenue, a number, by the way that I'm that
does not sound right, but it's what's being reported. Open
AI's costs are increasingly nearly and generative AI, as I've
said a few times, is deeply unprofitable and unsustainable. Making
open AI's models better, which doesn't necessarily mean they're more
(30:13):
valuable and more useful, also costs hundreds of millions, if
not billions, of dollars. As a result, open ai needs
to raise more money than anybody has ever raised in
the history of the valley in the next twelve to
twenty four months, or launch a products so significant that
it blows chet GPT out of the water and actually
is profitable. And these things are unlikely but possible. I
(30:36):
guess Microsoft, Google, and Amazon could somehow change the narrative,
but doing so too would require some sort of technological
breakthrough or a product, a very obvious product that would
make people money, more money than is being spent on
the thing. And the markets are also capricious, and thus
I really wouldn't count on there being a narrative turnaround,
(30:57):
especially with the shifting sands of a vestor sentiment. As
I've mentioned previously. On August second, Elliott Management, the hedge
fund with over seventy billion dollars in assets under management,
said in a letter to clients that AI is overhyped
and the Nvidia is a bubble, adding that AI had
not delivered value commensurate with the hype. While it's possible
there are others that dissent against the rapidly forming consensus
(31:19):
against generative AI to prop things up, which I do
think is possible, as especially if you deepen the whole
with Microsoft or and Video. It's worth putting Elliott Management's
note in a broader context. It's only the latest voice
in a loud or at least increasingly louder, chorus of
critics whose members include esteamed, analyst houses, and some of
(31:39):
the largest investment banks in the world. Gold and Sax
put out a report in July that you've heard in
the pop culture episode that said generative AI was too
expensive and didn't solve the complex problems that would need
to justify their expenses. At the end of July, Gartner
put out a report predicting that thirty percent of generative
AI products would be abandoned entirely after a proof of
(32:00):
concept by end of twenty twenty five, and The Washington
Post reports that Barklay's bank thinks that and I quote
Wall Street and Lists are expecting big tech companies to
spend around sixty billion dollars a year on developing AI
models by twenty twenty six, but reap only twenty billion
dollars a year in revenue from AI by that point.
The lastra is gone there for you. That does say revenue,
(32:21):
that doesn't say profim and by the way, that's only
a billion dollars more than Microsoft's entire capital expenditures in
the last quarter alone. Microsoft also has a storied history
of pumping and dumping ideas for Microsoft. Augmented reality was
and I quote, an absolute breakthrough in twenty nineteen before
it was quietly shoved in a corner with mass layoffs
(32:42):
and a business line enclosures a few years later in
twenty twenty one, satchur Innidella, their CEO, couldn't overstate how
much of a breakthrough the metaverse was yet two years later,
and big props the Preston Growler of Computer World for
saying this in February twenty twenty three, he laid off
most of the people involved in everything on artificial intelligence
and that everything thing that's a headline from Wired magazine.
(33:06):
Very good. It's also important to remember that Microsoft effectively
owns open ai as part of their twenty nineteen funding round,
which I believe was a billion dollars. Microsoft is full
access to all of open AI's intellectual property and research,
which makes thorny little question why at this point would
Microsoft give open ai any more money other than to
(33:27):
save face. That's a very real possibility that Satch in
the Dela is surrounded by yes men who don't know
what they're talking about. Microsoft also has access to sell
open ais and I quote pre agi products, which is
all of their products right now, and full access to
all of their research, as well as and I quote
certain rights to open AI's intellectual property. Unless open ai
(33:47):
is capable of delivering something meaningfully different, something that Dela
can wave in front of analysts faces and show them
that they can automate all of these jobs away, its
existence is just another cost center and one that could
get eliminated in a tough quarter. And to be clear,
saturn Adella has been extremely frank about his leverage over
open ai and it's effective ownership of their tech or
(34:08):
be it buffeted with the usual bullshit platitudes and niceties
that you'd expect from a shape shifting tech CEO. Monster.
In a late twenty twenty interview with PR person Kara Swisher,
sach an Adella said, if open ai disappeared tomorrow, I
don't want any customer of ours to be worried about it,
quite honestly, because we have all the rights to continue
the innovation, not just to serve the product, but we
(34:28):
can go and just do what we were doing in
partnership ourselves. We have the people, we have the compute,
we have the data, we have everything. Mister Nadella, that
is an insane thing to say, and the kind of
thing you'd only say to someone like Kara Swisher. Swisher
should have made that the headline that Microsoft effectively owns
open ai. I miss this at the time I found
(34:49):
out about this quote, like a week or two ago.
It's crazy to me. It's crazy to me that this
was said, and it's not front page news. Microsoft owns
open ai. Who gives a shit about the nonprofit structure.
But it actually gets better because he followed up that
quote with this. Microsoft's investment of thirteen billion dollars in
open ai gives us significant rights, as I said, and
(35:12):
also this thing, it's not hands off, right. We are
in there, we are below them, we are above them,
we are around them. We do the kernel optimizations. Very
funny if you consider the CrowdStrike thing. By the way
we build tools, we build the infrastructure. So that's why
I think a lot of the industrial analysts are saying,
oh wow, it's really a joint project between Microsoft and
open ai. The reality is we are as I said,
(35:34):
very self sufficient in all of this. Well, I'm not
saying it's for certain. Microsoft can very clearly if they
so please, simply drop open AI and keep all of
their stuff. Nevertheless, as the market's sour and jennit of AI,
I think it's time for big tech to prove something.
It's time for them to prove that this wasn't all
a huge, big, stupid, fucking waste of time and money
(35:57):
to continue with their current paths. Summed up is sh
of Google and sach and Adella of Microsoft. As the
loudest of the publicly traded generative AI hype men, will
have to show everybody something remarkable. But even then they
will have to show how it actually makes money, and
if they can't, at least in the case of sach
and Adella, they can still walk away. As well as
(36:18):
in the best case scenario, these companies would either have
to dramatically reduce their costs, something they've really shown they're
not capable of doing and in fact have said they'll
do the opposite of or show ways to make generative
AI so significantly efficient that it somehow finally balances itself out.
What we're really waiting for now is for somebody to blink.
If none of the major cloud companies change course over
(36:40):
the next few months, we could potentially see this boom
continue for another quarter. And just to be clear, they
usually in lockstep. They love holding up each other's oligopolies
and monopolies. They love being a weird cartel that sells data.
To continue, though, they'll have to commit billions of dollars
in at the very least cloud credits for open ai
and Anthropic, as neither of those companies can survive without
(37:02):
further cash infusions. They'll also have to find some way
to make any of this useful and profitable, two challenges
the generative AI has never been able to deal with,
not even once. And I just as an aside, I've
had people email and saying, well, I use it for this.
I was on a show the other days that he
uses it for show notes. These are all fine whatever,
(37:23):
These are like little use cases. These people need this
to make a trillion dollars. Show notes aren't going to
do it. Shitty translations are not going to do it.
A picture of Garfield with a gun is not going
to do it. And big tech is going to have
to figure out a way to support these companies in
the face of rising disquiet among shareholders, who in the
most recent route of AI companies lost them about two
(37:46):
point six trillion dollars in market capitalization. As I pointed
out in the Shareholders Supremacy, investors don't really care about
the future or innovation. They just want profits and their
portfolios to get bigger. And they want to see the
number for profit go larger and the costs go lower,
and absolutely nothing else matters at all, which is why
(38:07):
they love generative AI at first, because if this actually
made a bunch of money, this is a way that
Microsoft could just route cloud costs from themselves to themselves.
I'll get to that in a future episode about monopolies, though,
And I think that if generative AI becomes an albatross
around the next of Google, Microsoft, and Amazon, which I
really believe it will, these companies are going to face
(38:27):
really nasty pressure to curtail their investments and their capex.
Investors will ask why these companies are spending a collective
one hundred two hundred billion dollars by the end of
twenty twenty five on a technology that likely won't drive
much revenue growth when it could be used for things
like stock dividends and stock buybacks, and I don't know
thrones for the demon lords that live underneath the offices.
(38:49):
I don't know how Microsoft truly makes money. That's a joke,
don't email me about it. But I believe in any case,
the next month is going to be critical to the
future of the AI boom, and I believe this is
a time for industry wide introspection and to really consider
everyone's roles and why this bubble existed in the first place.
In the next episode, I'm going to take a closer
(39:11):
look into open Aiy's likely demise, what it looks like,
and why the company's view avenues for salvation in fact
look like dead ends. And I'll get into how fucking
ridiculous this whole bubble is and what its potential collapse
spells for the future of the tech industry. Thank you
(39:35):
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 at Matasowski
dot com, M A T T O. S O. W
Ski 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.
(39:56):
I also really recommend you go to chat dot Where's
youoeed dot com to visit the discord, and go to
our slash Better Offline to check out I'll 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 Coolzonemedia dot com, or check us out
on the iHeartRadio app, Apple Podcasts, or wherever you get
(40:17):
your podcasts.
Host
Ed Zitron
Popular Podcasts
24/7 News: The Latest
Today’s Latest News In 4 Minutes. Updated Hourly.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.