March 14, 2025 • 34 mins
In the second of this week's two-part series, Ed Zitron walks you through how SoftBank is mortgaging its future to keep OpenAI alive - and the dangerous consequences of over-funding an unprofitable, unsustainable company.
---
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
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 greetums. I'm at xytron and this is better offline.
In my last episode, I started by telling you about
a report from the analyst wing of a major bank,
(00:24):
TD Cohen that revealed how Microsoft had drastically pulled back
on its plans to build new data center capacity. At
a bare minimum, Microsoft had canceled the equivalent of every
data center in London or Tokyo at least, that the
real figure is actually probably much much higher, and as
I pointed out, that this is a real pale horse,
a harbinger of bad times through generative AI and open
(00:45):
AI especially, and an indicator that the bubble is imploding
or popping or I really don't want to say it's popped,
and I actually don't know if it's necessary to say
when it does, it's never going to be one thing.
I also mentioned that some may interpret this movie as
a sponds to open AI's Stargate project, which aims to
build hundreds of billions of dollars of data centers and
power generation facilities in the US, all to power generator
(01:08):
FAI apps and tools that I'm not sure actually there's
really the demand for. But this begs the question how
feasible is the Stargate project. Let's start with a bit
of background. The Stargate project was officially announced that the
very beginning of the Trump presidency. Is open AI tried
to court favor from the notoriously transactional and praise hungry
person of the United States of America, despite that the
(01:28):
project has been in the works for some time and
open AI had previously courted the Biden administration. Sam Allman's
previous pitch to that administration late last year was that
it was necessary to build a five gigaw what data center.
We don't know how big Stargate will be, just that
it will initially involve spending one hundred billion dollars due
and I quote developed data centers for artificial intelligence in
the US according to the information, with the project potentially
(01:51):
scaling to five hundred billion dollars, a truly fucking, astoundingly
stupid number. Stargate's first and only data center deal currently
signed is in Abilene, Texas, and it's expected to be
operational in mid twenty twenty six. Though these centers usually
become operational in phases, this is especially likely to be
the case here considering that, according to the information. Open
(02:11):
Ai plans to have access to a gigawat of power
on hundreds of thousands of GPUs. As part of this,
the Stargate project will construct a three hundred and sixty
point five megawatt natural gas power station. And as I
said last episode, that one's about power. This power station is,
as far as I can tell, still in the permitting phase.
It'll be some time before Stargate breaks ground on the facility,
(02:31):
let alone starts to actually generate power. Now, as for funding,
things have got a little weird. Both open ai and
soft Bank have committed to putting in either eighteen billion
or nineteen million dollars each into the project. I've seen
both numbers reported, by the way. Regardless, it's not that
big a difference to worry about, especially with well the
fact that open ai does not really have the money
(02:53):
either way. What's a billion dollars when you don't have anything.
The company is currently trying to raise forty billion dollars
at a two hundred sixty billion dollar valuation, with the
quote CNBC part of the funding expected to be used
for open AI's commitment to Stargate. Now, I'm old enough
to remember when this round was previously rumored to be
valuing open Ai at three hundred and forty billion dollars
and also at three hundred billion dollars, and soft Bank
(03:16):
appears to be taking full responsibility for raising the round,
including syndicating as much as ten billion dollars of the amount,
which means that it would include a group of other investors. Nevertheless,
it certainly seems SoftBank will provide the majority of the
capital thirty billion dollars, with CNBC reporting that it will
be paid out over the next twelve to twenty four months,
with the first payment coming as soon as spring. Soft
(03:36):
Bank also has another problem, and this one's this one.
I think even the least technical of you can understand.
They also appear to not have the money. They don't
have the money. They not the money to give open Ai,
not the money for Stargate. It's a little bit of
worrying math issues with this. According to the Information, SoftBank
CEO Masayoshi's Sun is planning to borrow sixteen billion dollars
(03:57):
to invest in Ai, and may borrow another eight billion
dollars next year. The following points, by the way, are
drawn from the Information's reporting, and I give serious props
to Duro Osawa and Corey Weinberg for their hard work here.
I attack the information sometimes with some of the framing,
but they actually do some of the best reporting in
tech journalism. Now let's do some maths. Soft Bank currently
only has thirty one billion dollars in cash and its
(04:17):
balance sheet as of December twenty twenty four. It's net debt,
which despite what you think, does not measure total debt,
but rather represents its cash minus any debt liabilities, sounds
at twenty nine billion dollars. They plan to use the
loaning question to finance part of their investment in open Ai,
and there are acquisition of chip design firm Ampere. According
to SoftBank's reported assets, its holdings are worth about two
hundred and nineteen billion dollars, so thirty three point sixty
(04:39):
six trillion yen for those of you who deal with yen,
including stock in companies like Alibaba and ARM. On a
side note, not every company in soft Bank's portfolio is
an arm or an Alibaba or like a good one.
There are some real stinkers too, and I'm not just
talking about the Morrobon carcass of we work. I recommend
you look at and I'll link to the list in
the episode of notes SoftBank's recent group report, which is
(05:02):
linked in the spreadsheet. Like I just said, which is
me reading a script out a little behind the curtain
for you there. In particular, got to twenty nine in
the report, which lists the ten largest publicly traded companies
in soft Bank's Vision Fund portfolio. Now how all of them,
without exception trade at a significant fraction of their peak market cap.
In simpler terms, the ten biggest companies in soft Bank's
(05:22):
flagship tech fund are worth far far less than their
all time high, and in some cases with less than
one fifth of their all time high. If soft Bank
liquidated its assets, and I admit this is a big
ear for most likely a worst case scenario situation, how
big would their losses be? Separately, SoftBank has committed to
a joint venture called SBO open Ai Japan and to
spend three billion dollars a year on open Ai. Is
(05:44):
take for the various companies across its group. We'll talk
about that later, but there's some napkin mass. It is
what soft Banks agreed to sixteen billion dollars sorry, eighteen
billion dollars or nineteen billion dollars really do not know
in funding for the Stargate Data Center project, three billion
dollars a year in spend on open ai software, and
as much as thirty billion dollars in funding for open
Ai paid over twelve to twenty four months, according to
(06:06):
the Information, and by the way, twenty five billion has
also been reported. But the Information reports that open Ai
has told investors that soft Bank will provide it at
least thirty billion dollars of the forty billion dollars they need.
Jesus fucking Christ. Can you imagine if this went It's
something useful anyway, What I'm getting at is that soft
Bank has effectively agreed to bankroll the entirety of open
AI's future, signing up for over forty six billion dollars
(06:29):
of investments over the next few years, and does not
appear to be able to do so without selling its
current holdings in valuable companies like arm or taking at
least sixteen billion dollars of debt this year, representing a
fifty five percent increase of its current liabilities. Work still,
thanks to disagreement, open AI's future both in its ability
to expand its infrastructure, which appears to be entirely contingent
(06:49):
on the construction of Stargate with Microsoft pulling out, and
its ability to raise funding, which is also entirely dependent
on soft bank and soft Bank in this case, where
open ai is entirely dependent on soft Bank to live,
must borrow money to give to open Ai, a company
that only only loses money. On top of that, open
(07:11):
Ai also on the money losing front, anticipates it will
burn as much as forty billion dollars a year by
twenty twenty eight, and projects to only turn a profit
by the end of the decade after the build out
of Stargate, which I add is, like I said, almost
entirely dependent on soft bank, which has to take on
debt to fund both open Ai and the project required
(07:33):
to theoretically make open ai profitable. How the fuck does
this work? How does this work? How does this work?
How does this work? How the fuck does this work?
Open Ai, a company that spent nine billion dollars to
lose five billion dollars in twenty twenty four, requires so
much money to be its obligations, both to cover its stupid, ruinous, unprofitable,
unsustainable operations and the eighteen billion to nineteen billion dollars.
(07:55):
A committed to keep growing that it has to raise
more money than any startup ever raised in history, forty
billion fucking dollars, with the cast iron guarantee that it
will need even more money within a year. Every goddamn month,
Sam Wilton has to go to someone and say, my huge,
beautiful company is so powerful, but it's weak and sick
and frail. I need more money than you have today,
and I'll need more money than you'll have tomorrow. My
(08:17):
company is so sick and weak, but it's the most
powerful company of all time. Sam flipping off pisces me off.
Imagine if this money went literally anywhere else, you could
set it on fire at least be fucking warm. Soft Bank.
On top of the thirty billion dollars of funding and
three billion dollars a year of revenue it's committed to,
open ai itself also has to cough up eighteen billion
(08:38):
dollars for Stargate, the data center project that open ai
will run and get this. Soft Bank will take financial
responsibility for forty eight billion dollars in cash three billion
dollars in revenue. You the latter of which, like all
open AI's offerings, will lose the company money. Open Ai
has no path to profitability guaranteeing it will need more
(08:58):
cash and right now at the time time, and it
needs it more than it's ever needed it. Soft Bank,
the only company willing to provide it and possibly the
only company with the money to do so, has proven
that it will have to go to greet possibly ruinous
lengths to do so. If open ai needs forty billion
dollars in twenty twenty five, how much will it need
in twenty twenty six fifty billion, one hundred billion dollars.
Where is that money going to come from? While soft
(09:21):
Bank might be able to do this once, what happens
when open ai needs money in six to twelve months.
Soft Bank made about fifteen billion dollars a profit in
the last year on about forty six billion dollars of revenue.
Three billion dollars is an absolutely obscene amount to commit
to buying open ai software annually, especially when some of
it is allegedly for access to open AI's barely functional
operator and mediocre deep research products. As per my previous
(09:44):
podcasts and pieces, I do not see how open ai survives,
and soft Bank's involvement only gives me further concern. While
soft banker theoretically burn its current holdings to fund open
Ai in perpetuity, its ability to do so is cast
into doubt by them having to borrow money from other
banks to get into this funding ground and to get
Stargate done. Open Ai burned five billion dollars in twenty
(10:05):
twenty four, a number it will likely double in twenty
twenty five. And remember the information reported that open ai
was projected to spend thirteen billion dollars on compute alone
with Microsoft in twenty twenty five, and has no path
to profitability. SoftBank has already had to borrow to fund
this round, and the fact that it had to do
so suggests its inability to continue supporting open Ai without
(10:26):
accruing further debt. Open Ai. As a result of Microsoft's cuts,
the data center capacity now only has one path expansion
once it runs through whatever remaining buildout Microsoft has planned.
That is, and that's Stargate, a project funded by open
AI's contribution, which it's receiving from SoftBank and SoftBank which
is also having to take out loans to meet its share.
(10:47):
How does this work exactly? How does this continue? Do
you see anyone else stepping up to fund this. Who
else has got thirty to forty billion dollars to shit
out every year? While the answer is Saudi Arabia. SoftBank's
CEO Massioshi Son recently said that he had had and
I quote not given Saudi ruler Muhammed bin Salmon in
our return, adding that he still owed him. That's really
(11:10):
not the ideal thing you want to say after naming MBS.
Nothing about this suggests that Saudi money will follow soft
banks in anywhere near the volume necessary. As to the
m ratis. They're already involved through the MGX fund, and
it's unclear how much more they'd be willing to commit. Really, really,
are really, mate, buddy fellow? How does this? How does
(11:31):
this work? In my opinion, this open Ai soft Bank
deal is wildly unsustainable, dependent on soft Bank continuing to
raise both debt and funnel money directly into a company
that burns it burns it by the billions every year,
and it's said to only burn more thanks to the
arrival of its latest bullshit model. And if it had
a huge breakthrough that would change everything. Wouldn't Microsoft want
to make sure they were building the data center capacity
(11:52):
to support him? Hey, maybe they're not. Now. Perhaps this
crazy level of spending would be necessary if open ai
was still the leader in generator of AI and it
(12:13):
was still meaningfully improving its capabilities year after year or
two year after two years. And I think we all
know that that isn't the case. A few weeks ago,
open ai launched GPT four point five, it's latest model
that well, this isn't brilliant, sam Alman says, and I quote,
is the first model that feels like talking to a
thoughtful person, which, by the way, is really funny. It's
(12:35):
really funny to say it, to be like, yeah, I
have spent billions of dollars telling you that this bullshit
is like a person, but this one really is. The
other one's not so much. I did not think that
in the past. And it's not obvious what GPT four
point five does better, or even really what it does,
other than Altman saying it is a different kind of
intelligence and there's a magic to it that he has
(12:58):
not felt before. Wow, wow, really really inspirational stuff. Sammy
fucking idiot. This was, by the way, in Altman's words,
the good news. The bad news was that and I
quote GPT four point five is a giant expensive model,
adding the open Ai was out of GPUs, but proudly
declaring that it had added tens of thousands of GPUs
(13:19):
in the week following and would roll out to open aiyes,
twenty dollars a month plus here and that they would
be adding hundreds of thousands of GPUs soon. Excited. Well,
you shouldn't be on top of a vague product set
and interterminately high compute cost. GPT four point five costs
developers an incredible seventy five dollars put million input tokens
in those are the prompts and data pushed into the model,
(13:40):
and of absolutely astounding one hundred and fifty dollars per
million output tokens. That's the output it creates. And that's
roughly three thousand percent more for input tokens and fifteen
hundred percent more expensive for output tokens than GPT four
to zero For results that open Aiyes, andre J. Caapathy
described as a little bit better and awesome, but also
not exactly that are trivial to point to, and one
(14:02):
developer described GPT four point five to our technica as
a lemon when comparing its reported performance to its price.
Arstentnaca also reported the GPT four point five was terrible
for coding relatively speaking, and other tests showed that the
model's performance was either slightly better or slightly worse across
the board. With that according to Ours Technica, one success
(14:22):
metric being the open ai found open human evaluators preferred
GPT four point five's responses of a GPT four to
zero in about fifty seven percent of interactions. Wow, that's
very underwhelming. So, just to be crystal clear, the biggest
AI company's latest model appears to be even more ruinously
expensive than its last one, while providing modest at best
(14:43):
improvements and performing worse than several benchmarks done competing models
very good. Despite these piss poor results, Sam Wultman's reaction
was to bring in hundreds of thousands of GPUs as
a means of exposing as many as people as possible
to his mediocre, ultra expensive model. And the best that
Altman has to offer is that this is the first
time people have been emailing with such passion asking open
(15:04):
ai to promise to never stop offering a specific model.
I am just going to say this. That is a tweet,
and it never happened, or it happened like once. This
is some girlfriend in Canada share Sam Altman is washed
when all of this falls apart, Remember I said he
was washed? Now, Remember how I talked about open AI's
lack of meaningful improvement as a reminder, GPT four point
(15:26):
five was meant to be GPT five, but, according to
the Wall Street Journal, continually failed to make a model
that advanced enough to justify the enormous cost, with a
six month training run costing five hundred million dollars and
GPT four point five requiring multiple runs of different sizes.
So yeah, open ai spent hundreds of millions of dollars
to make this great stuff. And I haven't even mentioned
(15:47):
the company's supported agent products. And no, I'm not talking
about Operator, which is also dogshit. By the way, open
ai wants to create tiers of AI agents, with the
cheapest costing two thousand dollars a month and capable of
handling it administrative tasks, and the most expensive costing twenty
thousand dollars in having PhD level, Look, I wrote the
script right, and I'm gonna be honest. I can't even
(16:08):
read that sentence with a straight face. Operator cannot even
search trip Advisor properly. It can't even do a thing
that let me google that for you does. And these
chunder fucks want to charge two thousand dollars a month
for an agent that does what does some sort of
what's it dou?
Speaker 2 (16:27):
Oh?
Speaker 1 (16:27):
Twenty thousand dollars for something with PhD level capabilities. I
think all the people with PhDs listen to this have
just stood up and gone, I have an idea. And
this is insane on many levels, not simply because the
base product is undercut by actual human workers in many
parts of the world, and even PhD students are typically
I only paid twenty thousand to thirty thousand a year
on average, and even people with actual doctorates in industries
(16:50):
really earn twenty thousand dollars a month unless they're working
in Silicon Valley occupying a C suite job. But forget
all about that. What does it mean to have a
PhD level agent? Remember? Other lambs are guessing machines. They
don't know anything or even understand the concepts behind the
words they spit onto a page. No, seriously, Sammy, what
does it mean? I'm fucking waiting you damp Goblin you pissant. Ah,
(17:15):
maybe I shouldn't just site here instarting him wanker. Anyway, This,
by the way, is the company that is about to
raise forty billion dollars led by a Japanese bank that
has to go into debt to fund both their operations
and the infrastructure necessary for them to grow any further. Again,
as we started with, Microsoft is canceling plans to massively
(17:36):
expand its data center capacity right at the time when
open ai just released its most computationally demanding model. Ever,
how do you reconcile those two things without concluding either
that Microsoft expects GBT four point five to be a
flop although it's simply unwilling to continue bankrolling open AI's
continued growth, or perhaps it's having doubts about the future
of generative AI in general. Hmm. Maybe now I have
(18:02):
been and remain hesitant to call the bubble bursting, because
bubbles do not burst, really, they certainly don't burst in
neat little events. Nevertheless, my pale horses I've predicted in
the past were led by one specific call that reduction
and capital expenditures by a hyperscaler was a sign that
things were collapsing. Microsoft walking away from over a gigawatt
of data center plans equivalent to as much as fourteen
(18:23):
percent of its current data center capacity, is a direct
sign that it does not believe that growth is there
in generative AI, and thus they are not building the
infrastructure to support it, and indeed may have overbuilt something,
as I've mentioned that Microsoft CEO Sachinidella has directly foreshadowed
in his interview with Dwarkesh and otherwise extremely boring waste
of an hour of your life. The entirety of the
(18:45):
tech industry and the AI bubble has been built on
the assumption that generative AI was the next big growth
vehicle for tech, and if Microsoft, the largest purchaser of
Nvidia GPUs in the most aggressive builder of AI infrastructure,
is reducing capacity, it heavily suggests that the growth is
not there. Microsoft has, by the looks of things, effectively
given up on further data center expansion, at least at
(19:07):
the breakneck pay suit runs promised, and that even suggests
that generator if AI will be a thing in a
few years, definitely not the scale it is right now.
AI boosters will email me and they'll say there's something
that I don't know that in fact, Microsoft has some
greater strategy and some efficiency play. But answer me this,
Why is Microsoft canceling over a giga? What of data
set expansion? And again this is the most conservative estimate.
(19:30):
The realistic number is much much higher. Do you think
it's because it expects there to be this dramatic demand
for the AI services? Do you think it's reducing supply
because of all the demand? Now you might think that
this is an efficiency play they're playing with deep SIG right, No, sorry,
that doesn't matter. Even if deep CG was this magical
(19:50):
efficiency play, which it may or may not be. I
actually think it is more efficient, Like that's true, but
we actually don't know if it's profitable. Even then, they've
been talking about not having the capacity to deal with demand.
They've been talking about how incredible this is. They've been
talking about how big this is gonna be. This sounds
like they don't think it's gonna big, gonna be big.
Even it's gonna big, they're gonna get my ass in
(20:11):
the comments on that one. Now, one might argue that
Microsoft's reduction and capacity build out it's just the sign
that open AI is moving its compute elsewhere. Maybe that's true.
And if starguate ever gets built, which I question anyway,
here are some questions to ask. Microsoft still sells access
to open AI's API through asure. Does it not see
the growth in that product? Do they not see it?
(20:34):
They're not expanding? Is the growth not there? And Microsoft still,
one would assume, makes money off of open AI's compute expenses,
right or is that not the case due to the
vast seventy five percent discount the open ai gets on
using its services. I have been told that it's very
close to the wire by sources, but I can't say it.
But you'd have to just look at the fact that
they do actually do that was reported by the information.
(20:56):
Look look what that Microsoft making. So which a material
pullback on data center expansions suggests that the growth in
generative AI products, both those run on Microsoft servers and
those sold as part of Microsoft's products, do not have
the revolutionary growth trajectory that both CFO Amhood and CEO
of sach Inadella have been claiming. And this is all
(21:16):
deeply concerning, while also calling into consideration the viability of
generative AI as a growth vehicle for any hyperscaler. If
I am correct, Microsoft is walking away not just from
expansion of its current data center operations, but from generative AI,
writ large. I actually believe it will continue selling this unprofitable,
unsustainable software because the capacity it has right now is
(21:36):
more than sufficient to deal with the incredible lack of demand.
It's time for investors in the general public to begin
demanding tangible, direct numbers on the revenue and profits related
to GENERATIVEAI, as it is becoming increasingly obvious that the
revenues are small and the profits are non existent. A
gigawatt of capacity is huge, and walking away from that
much capacity is a direct signal that Microsoft's long term
(21:59):
plans do not include needing a great deal of compute.
One counter could be that it's waiting for more of
the specialized nvideo GPUs to arrive, to which the response
is Microsoft still wants to build the capacity, so it
has somewhere to fucking put them again, These facilities take
anywhere between three and six years to build. Do you
really think black will be so delayed that they won't
arrive until what twenty twenty eight? Ways, what are you
(22:20):
talking about? From even alive then? Anyway, One counter could
be that there isn't the power necessary to power these
data centers, and if that's the case, it isn't but
let me human the idea.
Speaker 2 (22:30):
Then.
Speaker 1 (22:30):
The suggestion is that Microsoft is currently changing its entire
data center's strategy so significantly they now has to issue
over a giga what's worth of statements of intent across
the country to different places because of more power. There's
more power in those places, less than did they make
like a giga what worth of mistakes? Now? Okay. Another
(23:08):
counter is that I'm only talking about leases and not purchases.
In that case, I'll refer you to this article from CBRE,
which is linked in the spreadsheet for this episode, and
which includes a looser dating read on how hyperscalers actually
invest in data center infrastructure. Leases tend to account for
the majority of spending simply because it's less risky. A
specialist takes care of the tough stuff location buying, land handling, construction,
(23:29):
and the hyperscaler isn't left trying to figure out what
to do with the facility when it reaches the end
of it's the useful life cycle. I also expect someone
to chime in and say well, that's this Microsoft, What
about Google and Amazon? Get out of my house. I'd
counter and say that these companies are comparatively less exposed
to general IFAI. Amazon has invested eight billion dollars in Anthropic,
which is a bit less than half of what Microsoft
(23:50):
is reportedly invested in open Ai, which amounted to about
fourteen billion dollars as of December. When you consider the
discounted asure rates Microsoft offers to open ai too, the
real number is probably my much much higher. Google also
has three billion dollars in Anthropic, in addition to its
own AI services like Gemini. Open Ai, as I noted
in my last newsletter, is pretty much the only real
(24:10):
generative AI company with market share and significant revenue, although
I once again remind you that revenue is not the
same thing as profim and this is true across mobile
web and likely it's APIs two. Similarly, nobody has quite
pushed generative AI as aggressively as Microsoft, which has introduced
it to an overwhelming number of its paid products, hiking
prices for customers as it goes. I suppose you could
(24:32):
say that Google has pushed GENII into its workspace products
as well as its sarch products, but the scale and
aggression of Microsoft's push feels different that, and as I've
mentioned repeatedly, they are the largest purchaser of Nvidia GPUs
by nearly twice as many four hundred and eighty five
thousand as its nearest competitor to Meta, which brought two
hundred and twenty four thousand in twenty twenty four. Ultimately,
(24:53):
Microsoft has positioned itself at the heart of generative AI,
both through its own strategic product decisions and its partnership
with open AI. And the fact that it's now scaling
back on the investment required to maintain that momentum is
I believe pretty significant. I also recognize that all of
this is a big juicy steake for someone some people
call a pig or an animal or a monster or
an AI cynic. Look, I've pulled over this data repeatedly
(25:16):
and done all that I can to find less convenient
or satisfying conclusions. Let us off intent and likely the
weakest part of my argument. These are serious documents, by
the way, but they're not always legally binding. Neither are
those statements of qualifications. But as Tdcoen pointed out, sqqs
are generally treated as the green lights stop working on
construction even though a formal lease agreement hasn't yet been signed,
(25:37):
and to be clear, Microsoft let an interminate amount of
sqqs go. Nevertheless, it's incredibly significant that Microsoft is letting
so many the equivalent of as much as fourteen percent
of its current data set of capacity at a bare minimum.
On top of the couple hundred so at least two
hundred megabats of data center leases become canceled. I do
not know why nobody else has done this analysis. I've
(25:58):
now read every single piece about the td co and
report from every single outlet the covenant. I've read some
weird seo stuff. It's it's not good, and I'm under
I'm just kind of astounded by the lack of curiosity
as to what one GW plus means and the report
the meaningfully moved markets, as I'm equally astonished by the
lack of curiosity to contextualize most tech news. It's as
(26:21):
if nobody wants to think about this too hard, like
nobody wants to stop the pie. Nobody wants to accept
what's being staring us in the face since last year,
if not earlier, and we're given the most egregious, glaring evidence.
People still must find ways to dismiss it or ignore
it rather than give it the energy it deserves. Far
more resources were dedicated to finding ways to gussy up
the releases of anthropics claudes on at three point seven
(26:44):
or open ais GPT four point five. And we're given
the report from an Investment Banks research wings that the
largest spender and generityve Ai, the largest backer for now
of open ai at least, is massively reducing its expenditures
in data centers required for the industry and for open Ai,
a company of stems worth one hundred and fifty seven billion
dollars to expand. Microsoft's stake in open Ai is a
(27:06):
bit fuzzy, is open Ai doesn't issue traditional equity, and
there's a likelihood it may be diluted as more money
comes in. It reportedly owned forty nine percent in twenty
twenty three, though assuming that's still the case, are we
to believe that Microsoft is willing to strangle and that's
it worth at least seventy five billion dollars several times
more than its investment today by canceling a few leases.
(27:26):
How many more alarms do we need to go off
before people recognize that something bad is happening. Why is
that tangible, meaningful evidence that we're in a bubble and
possibly a sign that it might be popping less interesting
than the fact that Claudes on it three point seven
can think longer. And if you're listening to this and
you think I'm talking about you, I fucking am. I
am sick of this shit. I am absolutely sick of
(27:49):
this shit. I'm sick of reading articles like that when
far more important and scary and damning things are happening,
things that actually matter. I don't care if Warrio Ama
Day has allowed you to make it compute for longer.
It doesn't are compared to this. It does not matter
compared to a gig or what or more of capacity going.
And I'm sick of this. I'm sick of me having
to be the guy. Sometimes. I do not say these
(28:10):
things to be right. I don't want to be a
cynic or a hater. I say this all because I
am trying to understand what's going on, and if I
do not, I will actually go insane. Every time I
sit down to write my newsletter or record this podcast,
I'm doing it because I'm trying to understand what's happening
and how I feel about it, And these are the
only terms to dictate my creativity. It just happens that
(28:30):
I've stared at the tech industry for too long and
now I can't look away. Perhaps it's driving me mad,
or maybe I'm getting smarter, or maybe it's an arm
or palmer of the two. But what comes out of
my work is not driven by wanting to go viral
or having a hot take, or be a renowned skeptic,
or being right or being anything, because such things suggest
that I would do this differently if three people listened
(28:52):
versus actually can't say the amount of people that do.
I have rules, but fifty seven thousand people subscribe to
my newsletter extra. From there, I would do the same
goddamn thing. And in fact, if you look back in
my work, I've done the same thing from the beginning. Now,
I'm not saying that Microsoft is dying or making any
grandiose claims about what happens next. What I am describing, however,
(29:13):
is the material contraction of the largest investor in data centers,
according to TD Cohen, potentially at a scale that suggests
the Microsoft has meaningfully reduced its interest in further expansion
of data centers. Writ large, this is a deeply concerning move,
one that suggests that Microsoft does not see demands sustain
the current expansions, which has greater ramifications beyond generative AI,
because it suggests that there isn't any other reason for
(29:35):
it to expand the means of delivering software. What has
sat in Adella scene? What is Microsoft cfo amy Hood doing?
What is the plan here? And really what's the plan
with open Ai? SoftBank has committed over forty billion dollars
of cost that it currently cannot afford, taking on as
much as twenty four billion dollars in debt in the
(29:55):
next year to help sustain one more funding round and
the construction of data centers. Open Ai, a company that
loses money on literally every single customer. To survive, open
Ai must continue raising more money than any startup has
ever raised before, and they are only able to do
so from SoftBank, which in turn must take on debt.
Open Ai burned five billion dollars in twenty twenty four
and more likely burn eleven billion dollars or more in
(30:17):
twenty twenty five, and will continue burning money in perpetuity
and to scale further will require funding for a data
center project funded partially by a funding from a company
that's taken on debt to fund them. And when you
put this all together, all I can see is calamity.
Generative AI does not have meaningful mass market use cases,
and while chat GPT may have four hundred million weekly
(30:37):
active users, there doesn't appear to be meaningful consumer adoption
outside of chat GPT, mostly because almost all AI media
coverage inevitably ends up marketing one company Open AI. Argue
with me all you want about your personal experiences with
chat GPT or how you personally found it useful. I
don't care. I stopped listening a while ago. Your points
(30:57):
never prove anything that doesn't make it a product with
mass market utility or enterprise utility, or worth the vast
sums of money being plowed into it. Worse still, that
doesn't appear to be any meaningful revenue. As discussed in
my last episode, Microsoft claims thirteen billion dollars in annual
recurring revenue not profit on all AI products combined on
over two hundred billion dollars of capital expenditures since twenty
(31:19):
twenty three, and no other hyperscaler is willing to break
out any AI revenue at all. Not Amazon, no matter,
not Google, Nobody does that. Not worry anyone. Is anyone
listening to this who actually deals with the economy? Can
you please listen to me? So we don't Actually I
don't know what we do. I actually don't know. I
have no idea. Where's the growth? Where's the money? Where's
(31:44):
the money? Sammy? Where is it? Where's my money? Honey,
give me the money, Sam Mormon? Where's my money? Why
is Microsoft canceling a gigawat of data center capacity while
telling everybody that it didn't have enough data centers to
handle demands for its AI products. I suppose this is
one way of looking at it. Microsoft may currently have
a capacity issue, but soon one, meaning that further expansion
(32:04):
is unnecessary. And that's the case, It'll be interesting to
see whether their peers follow suit either way. Look, I
see nothing that suggests that there's future growth in general
of AI. In fact, I think it's time for everybody
to seriously consider the big tech burn billions of dollars
on something that nobody ever wanted or would pay for.
If you listen to this and scoff, I don't know
(32:25):
what should I have talked about anthropic adding a sliding
thinking bar to a model GPT four point five. Who
gives a shit? Can you even tell me what it
does differently to GPT four to zero? Can you explain
to me why it matters? Or are you more interested
in nakedly captured imbeciles like Ethan Mollock sweatly oinking about
how powerful the latest ask and shit. It's like no
(32:46):
one's interested in the things happening in the real world.
It's like nobody's thinking about the silicon and the infrastructure
and how things actually get built. Wake the fuck up, everybody.
Things are on fire. Thank you for listening to Better Offline.
(33:07):
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 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
(33:29):
you go to chat dot Where's youreaed dot at to
visit the discord and go to our slash Better Offline
to check out our reddit. Thank you so much for listening.
Speaker 2 (33:38):
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.
Speaker 1 (34:02):
F
Media greetums. I'm at xytron and this is better offline.
In my last episode, I started by telling you about
a report from the analyst wing of a major bank,
(00:24):
TD Cohen that revealed how Microsoft had drastically pulled back
on its plans to build new data center capacity. At
a bare minimum, Microsoft had canceled the equivalent of every
data center in London or Tokyo at least, that the
real figure is actually probably much much higher, and as
I pointed out, that this is a real pale horse,
a harbinger of bad times through generative AI and open
(00:45):
AI especially, and an indicator that the bubble is imploding
or popping or I really don't want to say it's popped,
and I actually don't know if it's necessary to say
when it does, it's never going to be one thing.
I also mentioned that some may interpret this movie as
a sponds to open AI's Stargate project, which aims to
build hundreds of billions of dollars of data centers and
power generation facilities in the US, all to power generator
(01:08):
FAI apps and tools that I'm not sure actually there's
really the demand for. But this begs the question how
feasible is the Stargate project. Let's start with a bit
of background. The Stargate project was officially announced that the
very beginning of the Trump presidency. Is open AI tried
to court favor from the notoriously transactional and praise hungry
person of the United States of America, despite that the
(01:28):
project has been in the works for some time and
open AI had previously courted the Biden administration. Sam Allman's
previous pitch to that administration late last year was that
it was necessary to build a five gigaw what data center.
We don't know how big Stargate will be, just that
it will initially involve spending one hundred billion dollars due
and I quote developed data centers for artificial intelligence in
the US according to the information, with the project potentially
(01:51):
scaling to five hundred billion dollars, a truly fucking, astoundingly
stupid number. Stargate's first and only data center deal currently
signed is in Abilene, Texas, and it's expected to be
operational in mid twenty twenty six. Though these centers usually
become operational in phases, this is especially likely to be
the case here considering that, according to the information. Open
(02:11):
Ai plans to have access to a gigawat of power
on hundreds of thousands of GPUs. As part of this,
the Stargate project will construct a three hundred and sixty
point five megawatt natural gas power station. And as I
said last episode, that one's about power. This power station is,
as far as I can tell, still in the permitting phase.
It'll be some time before Stargate breaks ground on the facility,
(02:31):
let alone starts to actually generate power. Now, as for funding,
things have got a little weird. Both open ai and
soft Bank have committed to putting in either eighteen billion
or nineteen million dollars each into the project. I've seen
both numbers reported, by the way. Regardless, it's not that
big a difference to worry about, especially with well the
fact that open ai does not really have the money
(02:53):
either way. What's a billion dollars when you don't have anything.
The company is currently trying to raise forty billion dollars
at a two hundred sixty billion dollar valuation, with the
quote CNBC part of the funding expected to be used
for open AI's commitment to Stargate. Now, I'm old enough
to remember when this round was previously rumored to be
valuing open Ai at three hundred and forty billion dollars
and also at three hundred billion dollars, and soft Bank
(03:16):
appears to be taking full responsibility for raising the round,
including syndicating as much as ten billion dollars of the amount,
which means that it would include a group of other investors. Nevertheless,
it certainly seems SoftBank will provide the majority of the
capital thirty billion dollars, with CNBC reporting that it will
be paid out over the next twelve to twenty four months,
with the first payment coming as soon as spring. Soft
(03:36):
Bank also has another problem, and this one's this one.
I think even the least technical of you can understand.
They also appear to not have the money. They don't
have the money. They not the money to give open Ai,
not the money for Stargate. It's a little bit of
worrying math issues with this. According to the Information, SoftBank
CEO Masayoshi's Sun is planning to borrow sixteen billion dollars
(03:57):
to invest in Ai, and may borrow another eight billion
dollars next year. The following points, by the way, are
drawn from the Information's reporting, and I give serious props
to Duro Osawa and Corey Weinberg for their hard work here.
I attack the information sometimes with some of the framing,
but they actually do some of the best reporting in
tech journalism. Now let's do some maths. Soft Bank currently
only has thirty one billion dollars in cash and its
(04:17):
balance sheet as of December twenty twenty four. It's net debt,
which despite what you think, does not measure total debt,
but rather represents its cash minus any debt liabilities, sounds
at twenty nine billion dollars. They plan to use the
loaning question to finance part of their investment in open Ai,
and there are acquisition of chip design firm Ampere. According
to SoftBank's reported assets, its holdings are worth about two
hundred and nineteen billion dollars, so thirty three point sixty
(04:39):
six trillion yen for those of you who deal with yen,
including stock in companies like Alibaba and ARM. On a
side note, not every company in soft Bank's portfolio is
an arm or an Alibaba or like a good one.
There are some real stinkers too, and I'm not just
talking about the Morrobon carcass of we work. I recommend
you look at and I'll link to the list in
the episode of notes SoftBank's recent group report, which is
(05:02):
linked in the spreadsheet. Like I just said, which is
me reading a script out a little behind the curtain
for you there. In particular, got to twenty nine in
the report, which lists the ten largest publicly traded companies
in soft Bank's Vision Fund portfolio. Now how all of them,
without exception trade at a significant fraction of their peak market cap.
In simpler terms, the ten biggest companies in soft Bank's
(05:22):
flagship tech fund are worth far far less than their
all time high, and in some cases with less than
one fifth of their all time high. If soft Bank
liquidated its assets, and I admit this is a big
ear for most likely a worst case scenario situation, how
big would their losses be? Separately, SoftBank has committed to
a joint venture called SBO open Ai Japan and to
spend three billion dollars a year on open Ai. Is
(05:44):
take for the various companies across its group. We'll talk
about that later, but there's some napkin mass. It is
what soft Banks agreed to sixteen billion dollars sorry, eighteen
billion dollars or nineteen billion dollars really do not know
in funding for the Stargate Data Center project, three billion
dollars a year in spend on open ai software, and
as much as thirty billion dollars in funding for open
Ai paid over twelve to twenty four months, according to
(06:06):
the Information, and by the way, twenty five billion has
also been reported. But the Information reports that open Ai
has told investors that soft Bank will provide it at
least thirty billion dollars of the forty billion dollars they need.
Jesus fucking Christ. Can you imagine if this went It's
something useful anyway, What I'm getting at is that soft
Bank has effectively agreed to bankroll the entirety of open
AI's future, signing up for over forty six billion dollars
(06:29):
of investments over the next few years, and does not
appear to be able to do so without selling its
current holdings in valuable companies like arm or taking at
least sixteen billion dollars of debt this year, representing a
fifty five percent increase of its current liabilities. Work still,
thanks to disagreement, open AI's future both in its ability
to expand its infrastructure, which appears to be entirely contingent
(06:49):
on the construction of Stargate with Microsoft pulling out, and
its ability to raise funding, which is also entirely dependent
on soft bank and soft Bank in this case, where
open ai is entirely dependent on soft Bank to live,
must borrow money to give to open Ai, a company
that only only loses money. On top of that, open
(07:11):
Ai also on the money losing front, anticipates it will
burn as much as forty billion dollars a year by
twenty twenty eight, and projects to only turn a profit
by the end of the decade after the build out
of Stargate, which I add is, like I said, almost
entirely dependent on soft bank, which has to take on
debt to fund both open Ai and the project required
(07:33):
to theoretically make open ai profitable. How the fuck does
this work? How does this work? How does this work?
How does this work? How the fuck does this work?
Open Ai, a company that spent nine billion dollars to
lose five billion dollars in twenty twenty four, requires so
much money to be its obligations, both to cover its stupid, ruinous, unprofitable,
unsustainable operations and the eighteen billion to nineteen billion dollars.
(07:55):
A committed to keep growing that it has to raise
more money than any startup ever raised in history, forty
billion fucking dollars, with the cast iron guarantee that it
will need even more money within a year. Every goddamn month,
Sam Wilton has to go to someone and say, my huge,
beautiful company is so powerful, but it's weak and sick
and frail. I need more money than you have today,
and I'll need more money than you'll have tomorrow. My
(08:17):
company is so sick and weak, but it's the most
powerful company of all time. Sam flipping off pisces me off.
Imagine if this money went literally anywhere else, you could
set it on fire at least be fucking warm. Soft Bank.
On top of the thirty billion dollars of funding and
three billion dollars a year of revenue it's committed to,
open ai itself also has to cough up eighteen billion
(08:38):
dollars for Stargate, the data center project that open ai
will run and get this. Soft Bank will take financial
responsibility for forty eight billion dollars in cash three billion
dollars in revenue. You the latter of which, like all
open AI's offerings, will lose the company money. Open Ai
has no path to profitability guaranteeing it will need more
(08:58):
cash and right now at the time time, and it
needs it more than it's ever needed it. Soft Bank,
the only company willing to provide it and possibly the
only company with the money to do so, has proven
that it will have to go to greet possibly ruinous
lengths to do so. If open ai needs forty billion
dollars in twenty twenty five, how much will it need
in twenty twenty six fifty billion, one hundred billion dollars.
Where is that money going to come from? While soft
(09:21):
Bank might be able to do this once, what happens
when open ai needs money in six to twelve months.
Soft Bank made about fifteen billion dollars a profit in
the last year on about forty six billion dollars of revenue.
Three billion dollars is an absolutely obscene amount to commit
to buying open ai software annually, especially when some of
it is allegedly for access to open AI's barely functional
operator and mediocre deep research products. As per my previous
(09:44):
podcasts and pieces, I do not see how open ai survives,
and soft Bank's involvement only gives me further concern. While
soft banker theoretically burn its current holdings to fund open
Ai in perpetuity, its ability to do so is cast
into doubt by them having to borrow money from other
banks to get into this funding ground and to get
Stargate done. Open Ai burned five billion dollars in twenty
(10:05):
twenty four, a number it will likely double in twenty
twenty five. And remember the information reported that open ai
was projected to spend thirteen billion dollars on compute alone
with Microsoft in twenty twenty five, and has no path
to profitability. SoftBank has already had to borrow to fund
this round, and the fact that it had to do
so suggests its inability to continue supporting open Ai without
(10:26):
accruing further debt. Open Ai. As a result of Microsoft's cuts,
the data center capacity now only has one path expansion
once it runs through whatever remaining buildout Microsoft has planned.
That is, and that's Stargate, a project funded by open
AI's contribution, which it's receiving from SoftBank and SoftBank which
is also having to take out loans to meet its share.
(10:47):
How does this work exactly? How does this continue? Do
you see anyone else stepping up to fund this. Who
else has got thirty to forty billion dollars to shit
out every year? While the answer is Saudi Arabia. SoftBank's
CEO Massioshi Son recently said that he had had and
I quote not given Saudi ruler Muhammed bin Salmon in
our return, adding that he still owed him. That's really
(11:10):
not the ideal thing you want to say after naming MBS.
Nothing about this suggests that Saudi money will follow soft
banks in anywhere near the volume necessary. As to the
m ratis. They're already involved through the MGX fund, and
it's unclear how much more they'd be willing to commit. Really, really,
are really, mate, buddy fellow? How does this? How does
(11:31):
this work? In my opinion, this open Ai soft Bank
deal is wildly unsustainable, dependent on soft Bank continuing to
raise both debt and funnel money directly into a company
that burns it burns it by the billions every year,
and it's said to only burn more thanks to the
arrival of its latest bullshit model. And if it had
a huge breakthrough that would change everything. Wouldn't Microsoft want
to make sure they were building the data center capacity
(11:52):
to support him? Hey, maybe they're not. Now. Perhaps this
crazy level of spending would be necessary if open ai
was still the leader in generator of AI and it
(12:13):
was still meaningfully improving its capabilities year after year or
two year after two years. And I think we all
know that that isn't the case. A few weeks ago,
open ai launched GPT four point five, it's latest model
that well, this isn't brilliant, sam Alman says, and I quote,
is the first model that feels like talking to a
thoughtful person, which, by the way, is really funny. It's
(12:35):
really funny to say it, to be like, yeah, I
have spent billions of dollars telling you that this bullshit
is like a person, but this one really is. The
other one's not so much. I did not think that
in the past. And it's not obvious what GPT four
point five does better, or even really what it does,
other than Altman saying it is a different kind of
intelligence and there's a magic to it that he has
(12:58):
not felt before. Wow, wow, really really inspirational stuff. Sammy
fucking idiot. This was, by the way, in Altman's words,
the good news. The bad news was that and I
quote GPT four point five is a giant expensive model,
adding the open Ai was out of GPUs, but proudly
declaring that it had added tens of thousands of GPUs
(13:19):
in the week following and would roll out to open aiyes,
twenty dollars a month plus here and that they would
be adding hundreds of thousands of GPUs soon. Excited. Well,
you shouldn't be on top of a vague product set
and interterminately high compute cost. GPT four point five costs
developers an incredible seventy five dollars put million input tokens
in those are the prompts and data pushed into the model,
(13:40):
and of absolutely astounding one hundred and fifty dollars per
million output tokens. That's the output it creates. And that's
roughly three thousand percent more for input tokens and fifteen
hundred percent more expensive for output tokens than GPT four
to zero For results that open Aiyes, andre J. Caapathy
described as a little bit better and awesome, but also
not exactly that are trivial to point to, and one
(14:02):
developer described GPT four point five to our technica as
a lemon when comparing its reported performance to its price.
Arstentnaca also reported the GPT four point five was terrible
for coding relatively speaking, and other tests showed that the
model's performance was either slightly better or slightly worse across
the board. With that according to Ours Technica, one success
(14:22):
metric being the open ai found open human evaluators preferred
GPT four point five's responses of a GPT four to
zero in about fifty seven percent of interactions. Wow, that's
very underwhelming. So, just to be crystal clear, the biggest
AI company's latest model appears to be even more ruinously
expensive than its last one, while providing modest at best
(14:43):
improvements and performing worse than several benchmarks done competing models
very good. Despite these piss poor results, Sam Wultman's reaction
was to bring in hundreds of thousands of GPUs as
a means of exposing as many as people as possible
to his mediocre, ultra expensive model. And the best that
Altman has to offer is that this is the first
time people have been emailing with such passion asking open
(15:04):
ai to promise to never stop offering a specific model.
I am just going to say this. That is a tweet,
and it never happened, or it happened like once. This
is some girlfriend in Canada share Sam Altman is washed
when all of this falls apart, Remember I said he
was washed? Now, Remember how I talked about open AI's
lack of meaningful improvement as a reminder, GPT four point
(15:26):
five was meant to be GPT five, but, according to
the Wall Street Journal, continually failed to make a model
that advanced enough to justify the enormous cost, with a
six month training run costing five hundred million dollars and
GPT four point five requiring multiple runs of different sizes.
So yeah, open ai spent hundreds of millions of dollars
to make this great stuff. And I haven't even mentioned
(15:47):
the company's supported agent products. And no, I'm not talking
about Operator, which is also dogshit. By the way, open
ai wants to create tiers of AI agents, with the
cheapest costing two thousand dollars a month and capable of
handling it administrative tasks, and the most expensive costing twenty
thousand dollars in having PhD level, Look, I wrote the
script right, and I'm gonna be honest. I can't even
(16:08):
read that sentence with a straight face. Operator cannot even
search trip Advisor properly. It can't even do a thing
that let me google that for you does. And these
chunder fucks want to charge two thousand dollars a month
for an agent that does what does some sort of
what's it dou?
Speaker 2 (16:27):
Oh?
Speaker 1 (16:27):
Twenty thousand dollars for something with PhD level capabilities. I
think all the people with PhDs listen to this have
just stood up and gone, I have an idea. And
this is insane on many levels, not simply because the
base product is undercut by actual human workers in many
parts of the world, and even PhD students are typically
I only paid twenty thousand to thirty thousand a year
on average, and even people with actual doctorates in industries
(16:50):
really earn twenty thousand dollars a month unless they're working
in Silicon Valley occupying a C suite job. But forget
all about that. What does it mean to have a
PhD level agent? Remember? Other lambs are guessing machines. They
don't know anything or even understand the concepts behind the
words they spit onto a page. No, seriously, Sammy, what
does it mean? I'm fucking waiting you damp Goblin you pissant. Ah,
(17:15):
maybe I shouldn't just site here instarting him wanker. Anyway, This,
by the way, is the company that is about to
raise forty billion dollars led by a Japanese bank that
has to go into debt to fund both their operations
and the infrastructure necessary for them to grow any further. Again,
as we started with, Microsoft is canceling plans to massively
(17:36):
expand its data center capacity right at the time when
open ai just released its most computationally demanding model. Ever,
how do you reconcile those two things without concluding either
that Microsoft expects GBT four point five to be a
flop although it's simply unwilling to continue bankrolling open AI's
continued growth, or perhaps it's having doubts about the future
of generative AI in general. Hmm. Maybe now I have
(18:02):
been and remain hesitant to call the bubble bursting, because
bubbles do not burst, really, they certainly don't burst in
neat little events. Nevertheless, my pale horses I've predicted in
the past were led by one specific call that reduction
and capital expenditures by a hyperscaler was a sign that
things were collapsing. Microsoft walking away from over a gigawatt
of data center plans equivalent to as much as fourteen
(18:23):
percent of its current data center capacity, is a direct
sign that it does not believe that growth is there
in generative AI, and thus they are not building the
infrastructure to support it, and indeed may have overbuilt something,
as I've mentioned that Microsoft CEO Sachinidella has directly foreshadowed
in his interview with Dwarkesh and otherwise extremely boring waste
of an hour of your life. The entirety of the
(18:45):
tech industry and the AI bubble has been built on
the assumption that generative AI was the next big growth
vehicle for tech, and if Microsoft, the largest purchaser of
Nvidia GPUs in the most aggressive builder of AI infrastructure,
is reducing capacity, it heavily suggests that the growth is
not there. Microsoft has, by the looks of things, effectively
given up on further data center expansion, at least at
(19:07):
the breakneck pay suit runs promised, and that even suggests
that generator if AI will be a thing in a
few years, definitely not the scale it is right now.
AI boosters will email me and they'll say there's something
that I don't know that in fact, Microsoft has some
greater strategy and some efficiency play. But answer me this,
Why is Microsoft canceling over a giga? What of data
set expansion? And again this is the most conservative estimate.
(19:30):
The realistic number is much much higher. Do you think
it's because it expects there to be this dramatic demand
for the AI services? Do you think it's reducing supply
because of all the demand? Now you might think that
this is an efficiency play they're playing with deep SIG right, No, sorry,
that doesn't matter. Even if deep CG was this magical
(19:50):
efficiency play, which it may or may not be. I
actually think it is more efficient, Like that's true, but
we actually don't know if it's profitable. Even then, they've
been talking about not having the capacity to deal with demand.
They've been talking about how incredible this is. They've been
talking about how big this is gonna be. This sounds
like they don't think it's gonna big, gonna be big.
Even it's gonna big, they're gonna get my ass in
(20:11):
the comments on that one. Now, one might argue that
Microsoft's reduction and capacity build out it's just the sign
that open AI is moving its compute elsewhere. Maybe that's true.
And if starguate ever gets built, which I question anyway,
here are some questions to ask. Microsoft still sells access
to open AI's API through asure. Does it not see
the growth in that product? Do they not see it?
(20:34):
They're not expanding? Is the growth not there? And Microsoft still,
one would assume, makes money off of open AI's compute expenses,
right or is that not the case due to the
vast seventy five percent discount the open ai gets on
using its services. I have been told that it's very
close to the wire by sources, but I can't say it.
But you'd have to just look at the fact that
they do actually do that was reported by the information.
(20:56):
Look look what that Microsoft making. So which a material
pullback on data center expansions suggests that the growth in
generative AI products, both those run on Microsoft servers and
those sold as part of Microsoft's products, do not have
the revolutionary growth trajectory that both CFO Amhood and CEO
of sach Inadella have been claiming. And this is all
(21:16):
deeply concerning, while also calling into consideration the viability of
generative AI as a growth vehicle for any hyperscaler. If
I am correct, Microsoft is walking away not just from
expansion of its current data center operations, but from generative AI,
writ large. I actually believe it will continue selling this unprofitable,
unsustainable software because the capacity it has right now is
(21:36):
more than sufficient to deal with the incredible lack of demand.
It's time for investors in the general public to begin
demanding tangible, direct numbers on the revenue and profits related
to GENERATIVEAI, as it is becoming increasingly obvious that the
revenues are small and the profits are non existent. A
gigawatt of capacity is huge, and walking away from that
much capacity is a direct signal that Microsoft's long term
(21:59):
plans do not include needing a great deal of compute.
One counter could be that it's waiting for more of
the specialized nvideo GPUs to arrive, to which the response
is Microsoft still wants to build the capacity, so it
has somewhere to fucking put them again, These facilities take
anywhere between three and six years to build. Do you
really think black will be so delayed that they won't
arrive until what twenty twenty eight? Ways, what are you
(22:20):
talking about? From even alive then? Anyway, One counter could
be that there isn't the power necessary to power these
data centers, and if that's the case, it isn't but
let me human the idea.
Speaker 2 (22:30):
Then.
Speaker 1 (22:30):
The suggestion is that Microsoft is currently changing its entire
data center's strategy so significantly they now has to issue
over a giga what's worth of statements of intent across
the country to different places because of more power. There's
more power in those places, less than did they make
like a giga what worth of mistakes? Now? Okay. Another
(23:08):
counter is that I'm only talking about leases and not purchases.
In that case, I'll refer you to this article from CBRE,
which is linked in the spreadsheet for this episode, and
which includes a looser dating read on how hyperscalers actually
invest in data center infrastructure. Leases tend to account for
the majority of spending simply because it's less risky. A
specialist takes care of the tough stuff location buying, land handling, construction,
(23:29):
and the hyperscaler isn't left trying to figure out what
to do with the facility when it reaches the end
of it's the useful life cycle. I also expect someone
to chime in and say well, that's this Microsoft, What
about Google and Amazon? Get out of my house. I'd
counter and say that these companies are comparatively less exposed
to general IFAI. Amazon has invested eight billion dollars in Anthropic,
which is a bit less than half of what Microsoft
(23:50):
is reportedly invested in open Ai, which amounted to about
fourteen billion dollars as of December. When you consider the
discounted asure rates Microsoft offers to open ai too, the
real number is probably my much much higher. Google also
has three billion dollars in Anthropic, in addition to its
own AI services like Gemini. Open Ai, as I noted
in my last newsletter, is pretty much the only real
(24:10):
generative AI company with market share and significant revenue, although
I once again remind you that revenue is not the
same thing as profim and this is true across mobile
web and likely it's APIs two. Similarly, nobody has quite
pushed generative AI as aggressively as Microsoft, which has introduced
it to an overwhelming number of its paid products, hiking
prices for customers as it goes. I suppose you could
(24:32):
say that Google has pushed GENII into its workspace products
as well as its sarch products, but the scale and
aggression of Microsoft's push feels different that, and as I've
mentioned repeatedly, they are the largest purchaser of Nvidia GPUs
by nearly twice as many four hundred and eighty five
thousand as its nearest competitor to Meta, which brought two
hundred and twenty four thousand in twenty twenty four. Ultimately,
(24:53):
Microsoft has positioned itself at the heart of generative AI,
both through its own strategic product decisions and its partnership
with open AI. And the fact that it's now scaling
back on the investment required to maintain that momentum is
I believe pretty significant. I also recognize that all of
this is a big juicy steake for someone some people
call a pig or an animal or a monster or
an AI cynic. Look, I've pulled over this data repeatedly
(25:16):
and done all that I can to find less convenient
or satisfying conclusions. Let us off intent and likely the
weakest part of my argument. These are serious documents, by
the way, but they're not always legally binding. Neither are
those statements of qualifications. But as Tdcoen pointed out, sqqs
are generally treated as the green lights stop working on
construction even though a formal lease agreement hasn't yet been signed,
(25:37):
and to be clear, Microsoft let an interminate amount of
sqqs go. Nevertheless, it's incredibly significant that Microsoft is letting
so many the equivalent of as much as fourteen percent
of its current data set of capacity at a bare minimum.
On top of the couple hundred so at least two
hundred megabats of data center leases become canceled. I do
not know why nobody else has done this analysis. I've
(25:58):
now read every single piece about the td co and
report from every single outlet the covenant. I've read some
weird seo stuff. It's it's not good, and I'm under
I'm just kind of astounded by the lack of curiosity
as to what one GW plus means and the report
the meaningfully moved markets, as I'm equally astonished by the
lack of curiosity to contextualize most tech news. It's as
(26:21):
if nobody wants to think about this too hard, like
nobody wants to stop the pie. Nobody wants to accept
what's being staring us in the face since last year,
if not earlier, and we're given the most egregious, glaring evidence.
People still must find ways to dismiss it or ignore
it rather than give it the energy it deserves. Far
more resources were dedicated to finding ways to gussy up
the releases of anthropics claudes on at three point seven
(26:44):
or open ais GPT four point five. And we're given
the report from an Investment Banks research wings that the
largest spender and generityve Ai, the largest backer for now
of open ai at least, is massively reducing its expenditures
in data centers required for the industry and for open Ai,
a company of stems worth one hundred and fifty seven billion
dollars to expand. Microsoft's stake in open Ai is a
(27:06):
bit fuzzy, is open Ai doesn't issue traditional equity, and
there's a likelihood it may be diluted as more money
comes in. It reportedly owned forty nine percent in twenty
twenty three, though assuming that's still the case, are we
to believe that Microsoft is willing to strangle and that's
it worth at least seventy five billion dollars several times
more than its investment today by canceling a few leases.
(27:26):
How many more alarms do we need to go off
before people recognize that something bad is happening. Why is
that tangible, meaningful evidence that we're in a bubble and
possibly a sign that it might be popping less interesting
than the fact that Claudes on it three point seven
can think longer. And if you're listening to this and
you think I'm talking about you, I fucking am. I
am sick of this shit. I am absolutely sick of
(27:49):
this shit. I'm sick of reading articles like that when
far more important and scary and damning things are happening,
things that actually matter. I don't care if Warrio Ama
Day has allowed you to make it compute for longer.
It doesn't are compared to this. It does not matter
compared to a gig or what or more of capacity going.
And I'm sick of this. I'm sick of me having
to be the guy. Sometimes. I do not say these
(28:10):
things to be right. I don't want to be a
cynic or a hater. I say this all because I
am trying to understand what's going on, and if I
do not, I will actually go insane. Every time I
sit down to write my newsletter or record this podcast,
I'm doing it because I'm trying to understand what's happening
and how I feel about it, And these are the
only terms to dictate my creativity. It just happens that
(28:30):
I've stared at the tech industry for too long and
now I can't look away. Perhaps it's driving me mad,
or maybe I'm getting smarter, or maybe it's an arm
or palmer of the two. But what comes out of
my work is not driven by wanting to go viral
or having a hot take, or be a renowned skeptic,
or being right or being anything, because such things suggest
that I would do this differently if three people listened
(28:52):
versus actually can't say the amount of people that do.
I have rules, but fifty seven thousand people subscribe to
my newsletter extra. From there, I would do the same
goddamn thing. And in fact, if you look back in
my work, I've done the same thing from the beginning. Now,
I'm not saying that Microsoft is dying or making any
grandiose claims about what happens next. What I am describing, however,
(29:13):
is the material contraction of the largest investor in data centers,
according to TD Cohen, potentially at a scale that suggests
the Microsoft has meaningfully reduced its interest in further expansion
of data centers. Writ large, this is a deeply concerning move,
one that suggests that Microsoft does not see demands sustain
the current expansions, which has greater ramifications beyond generative AI,
because it suggests that there isn't any other reason for
(29:35):
it to expand the means of delivering software. What has
sat in Adella scene? What is Microsoft cfo amy Hood doing?
What is the plan here? And really what's the plan
with open Ai? SoftBank has committed over forty billion dollars
of cost that it currently cannot afford, taking on as
much as twenty four billion dollars in debt in the
(29:55):
next year to help sustain one more funding round and
the construction of data centers. Open Ai, a company that
loses money on literally every single customer. To survive, open
Ai must continue raising more money than any startup has
ever raised before, and they are only able to do
so from SoftBank, which in turn must take on debt.
Open Ai burned five billion dollars in twenty twenty four
and more likely burn eleven billion dollars or more in
(30:17):
twenty twenty five, and will continue burning money in perpetuity
and to scale further will require funding for a data
center project funded partially by a funding from a company
that's taken on debt to fund them. And when you
put this all together, all I can see is calamity.
Generative AI does not have meaningful mass market use cases,
and while chat GPT may have four hundred million weekly
(30:37):
active users, there doesn't appear to be meaningful consumer adoption
outside of chat GPT, mostly because almost all AI media
coverage inevitably ends up marketing one company Open AI. Argue
with me all you want about your personal experiences with
chat GPT or how you personally found it useful. I
don't care. I stopped listening a while ago. Your points
(30:57):
never prove anything that doesn't make it a product with
mass market utility or enterprise utility, or worth the vast
sums of money being plowed into it. Worse still, that
doesn't appear to be any meaningful revenue. As discussed in
my last episode, Microsoft claims thirteen billion dollars in annual
recurring revenue not profit on all AI products combined on
over two hundred billion dollars of capital expenditures since twenty
(31:19):
twenty three, and no other hyperscaler is willing to break
out any AI revenue at all. Not Amazon, no matter,
not Google, Nobody does that. Not worry anyone. Is anyone
listening to this who actually deals with the economy? Can
you please listen to me? So we don't Actually I
don't know what we do. I actually don't know. I
have no idea. Where's the growth? Where's the money? Where's
(31:44):
the money? Sammy? Where is it? Where's my money? Honey,
give me the money, Sam Mormon? Where's my money? Why
is Microsoft canceling a gigawat of data center capacity while
telling everybody that it didn't have enough data centers to
handle demands for its AI products. I suppose this is
one way of looking at it. Microsoft may currently have
a capacity issue, but soon one, meaning that further expansion
(32:04):
is unnecessary. And that's the case, It'll be interesting to
see whether their peers follow suit either way. Look, I
see nothing that suggests that there's future growth in general
of AI. In fact, I think it's time for everybody
to seriously consider the big tech burn billions of dollars
on something that nobody ever wanted or would pay for.
If you listen to this and scoff, I don't know
(32:25):
what should I have talked about anthropic adding a sliding
thinking bar to a model GPT four point five. Who
gives a shit? Can you even tell me what it
does differently to GPT four to zero? Can you explain
to me why it matters? Or are you more interested
in nakedly captured imbeciles like Ethan Mollock sweatly oinking about
how powerful the latest ask and shit. It's like no
(32:46):
one's interested in the things happening in the real world.
It's like nobody's thinking about the silicon and the infrastructure
and how things actually get built. Wake the fuck up, everybody.
Things are on fire. Thank you for listening to Better Offline.
(33:07):
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 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
(33:29):
you go to chat dot Where's youreaed dot at to
visit the discord and go to our slash Better Offline
to check out our reddit. Thank you so much for listening.
Speaker 2 (33:38):
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.
Speaker 1 (34:02):
F
Host
Ed Zitron
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
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.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com