OpenAI and Microsoft’s Grand Bargain, Sam Altman’s Next Three Years, A New Humanoid Robot

Channel: Alex Kantrowitz

Published at: 2025-11-03

YouTube video id: La6OMtnzGzA

Source: https://www.youtube.com/watch?v=La6OMtnzGzA

Microsoft and OpenAI have met and signed
a peace treaty so this entire generative
AI business can move [music] forward.
Sam Alman plots out OpenAI's next three
years. Big tech earnings come in and
there's a new humanoid robot in town.
That's coming up on a very spooky Big
Technology Podcast Friday edition right
after this. Welcome to Big Technology
Podcast Friday edition where we break
down the news in our traditional
coolheaded and nuanced format. We have a
great show for you today because finally
the AI deal of the century is done.
Microsoft and OpenAI have a resolution.
OpenAI might be heading towards an IPO
and finally all the funding that OpenAI
has been promised can start pouring in
because the structure is here. We're
also going to talk about Sam Alman's
vision for the next 3 years of the
company, how OpenAI has become maybe
[music] a little too metified. uh big
tech earnings and of course there is a
new humanoid [music] robot that Ranjan
and I are both eager to discuss. So uh
joining us as always on Friday is Ranjan
[music] Roy. Ranjan, happy Halloween.
>> Have you uh have you ordered pre-ordered
your Neo robot just yet?
>> No. To me that is of all spooky
headlines this week that is the
spookiest and I'm staying far away. How
about you?
>> I have not pre-ordered it either. But
but I don't know. I don't think it's
that spooky. I want a robot standing
right next to me just, you know,
handling all of my work, all of my
household tasks.
>> All right, we will save that debate for
the end because I think you and I will
have diverging opinions on this
>> pro
as part of our service here to the
public discussing the benefits and the
downsides of AI and robotics. Uh, we'll
have to get into it. But first, I think
the biggest story of the week and really
the biggest story of the quarter if you
ask me, maybe this entire half of the
year is the fact that OpenAI and
Microsoft have a deal. This is from the
Wall Street Journal. OpenAI has
successfully converted to a more
traditional corporate structure. A move
that cleared an obstacle for a potential
initial cleared an obstacle for a
potential initial public offering and
pushed the valuation of longtime partner
Microsoft above 4 trillion. The
artificial intelligence startup has
turned its for-profit uh subsidiary into
a public benefit corporation of which
Microsoft will own 27%. The conversion
will grant OpenAI's nonprofit parent a
stake in the forprofit worth 130 billion
with the ability to get more ownership
as the for-profit becomes more valuable.
The deal ends nearly a year of wrangling
with attorneys uh the attorneys general
of California and Delaware who have the
power to regulate nonprofits as well as
with the broader philanthropic community
and key investors over the question of
whether the AI juggernaut could remain
true to its mission while transforming
into a more conventional company. Rajan
first question for you here is well open
eyes made this deal so like Microsoft
becomes a more traditional investor in
the company but it's still controlled by
a nonprofit uh and it it was I mean if
you think about the conversion itself
it's going from um an initial it's sorry
it's going from a uh from a for-profit
subsidiary into a public benefit
corporation. Um what is the big deal
here? This is actually I'm going to ask
back to you. Can you explain in clear
language what you think this means?
Because I still have a hard time trying
to understand again that the nonprofit
has a stake in the forprofit worth 130
billion and will get more ownership as
the forprofit becomes more valuable. It
still felt very open AI corporate
structurey to me like it didn't help
necessarily clear up a lot of the what's
been vague about everything so far. Do
you do you understand what is going on?
>> So I'll be the first to admit not fully.
Um okay why I think this is important
this is why I think it's the most
important story is because Microsoft's
ownership in Open AI is finally cleared
up. And yes, there's going to always be
some weirdness with OpenAI when it comes
to the nonprofit uh and the for-profit,
the public benefit corporation. Uh but
up until this point, you had Microsoft
in this position where it get it was
entitled to 50% of OpenAI's profits. Uh
it had an unlike determined ownership
percentage of the company. Um, and it
also had access to all of its IP rights
and it could potentially lose access to
all that if OpenAI just said, "Hey,
we've hit AGI." So, I think that to me
is the biggest thing that needed sorted
out. And until that became sorted out,
until this sort of conversion was
complete and Microsoft was baked in as a
traditional investor, the company
couldn't move forward. And now it's been
settled and now it can move forward. So,
to me, that that is the biggest point.
And of course, we know Microsoft had
been holding it holding it up because it
was trying to negotiate and strongarm
OpenAI for a large percentage uh uh even
larger I think than it got of this
public benefit corporation. That is the
big deal to me. What do you
>> Okay, I I I'll agree with you. I mean,
if nothing else, the fact that this
opens up the conversation around an open
AI IPO, that is massive news. I think as
well uh one of my favorite parts of the
agreement was you'd mentioned IP.
Microsoft still will have exclusive IP
rights to open AI technology until 2032.
So so there's a long time that whatever
OpenAI makes Microsoft still has access
to. But I also loved that the deal has
OpenAI committing to purchase an
additional $250 billion worth of Azure
cloud computing services. So of course I
mean Microsoft I think came out pretty
good here maintaining 27% ownership
getting just a massive uh uh commitment
in terms of compute and uh compute spend
over the next number of years
maintaining exclusive IP rights. So I
guess it does appear everyone everyone
does seem to win here.
>> That's right. And I think the Microsoft
thing is is quite interesting. uh for
Microsoft you are leaving this it's the
great it's an amazing deal for Microsoft
I mean of course Microsoft sort of had
open AAI in a strangleold right and
again that's why I think this is
important and now it's sort of loosened
that a little bit it's allowing open AAI
for instance to you know it's no long it
no longer has like first right of
refusal to um work with open AI on
infrastructure pro projects but it still
keeps 27% of the company it still keeps
the IP as you mentioned till 2032 uh and
so what I think Sat Nadella is doing
here is threading the needle. And
actually, if folks listened to the Scott
Guthrie uh interview here on the show a
couple weeks ago, you sort of saw it
trending this way, which is Satya is
saying you may be able to build, you
know, an amazing company um or or this
sort of unprecedented new technology by
scaling up what you have today. And
Microsoft doesn't want to bet its entire
company facilitating that for OpenAI. So
it relinquishes some of its stake in
what OpenAI was going to get uh in the
future for two things. First of all,
this 27% of the company which is massive
uh and then um well and the IP rights.
So as we have like we enter the most
critical buildout uh of generative AI uh
Microsoft will be the one that o owns
you know the IP rights of the most
important company. Um and then it says
basically if you want to go build go
build with others. So it prevents itself
from getting the downsides of like let's
say facilitating this whole uh buildout
and having to you know effectively
mortgage your company's financial future
to do it. So it's I think a terrific
move for Microsoft. Well, you know,
Satcha is feeling good because he he in
that coverage he starts talking about
how they're a platform company and
Satcha has the quote, "My mindset is all
platform. I'm happy with OpenAI. I'd
love to have anthropic gro anyone. If
Google wants to put Gemini on Azure,
please do so." He's feeling pretty good
right now, I think. And for for good
reason.
>> That was a very important quote from
Satya. It's him washing his hands of
OpenAI in a way.
>> Yep. Yeah,
>> I mean you you realize that Azure is
going to be OpenAI's models will only be
available on Azure and on uh and through
OpenAI's API that's it unless it gets
SA's blessing. U and so now SA is saying
well we're going to already have this by
the way we're going to get to earnings I
guess but we probably spread them
throughout Microsoft Azure grew 39%
quarter over year-over-year in this in
this most recent quarter. I mean that's
astonishing. Uh and so why wouldn't
Satya then you know now that he's like
effectively he has his ownership stake
in OpenAI he has uh and he has the
exclusive rights why wouldn't he you
know continue to welcome all the other
models and make Azure the go-to place
for anybody looking to build with this
technology.
>> Yeah I think again and we'll get more
into earnings. We'll get more into the
potential for an AI bubble or a wobble
as one of my favorite uh phrases of the
week came out, but [snorts] 39% on a
business of this scale. Like I've worked
at, you know, like midund million
revenue startups. 39% is a massive
number and Azure is just incredibly
wellplaced. And uh yeah, I think this
idea of just openly talking about
platform, even kind of throwing shade
and just saying if Google wants to put
Gemini on there, I think they're they
are in a very good place right now.
>> Now, even though their nonprofit will
control uh will appoint the board of the
for-profit side of OpenAI, I think by
all uh intents and purposes, it's become
one entity. there's only one board
member on the nonprofit that's not on
the for-profit side. Uh, and that's
where a lot of folks are saying, you
know, this is effectively the end of the
nonprofit, even though OpenAI did make
some nice announcements saying that the
nonprofit was going to like invest in
some AI safety initiatives. Here's uh
here's from this uh one AI watcher, Svi
Mauowitz. He says they're they're moving
to complete the greatest theft in human
history or perhaps perhaps second
business if you count what happened
around the dissolution of the USSR. Are
we going to let them get away with this?
>> Uh what do you think about the fact that
this is effectively
>> I didn't quite understand. I know you
had you had thrown that the tweet into
our uh into our prep dock. I didn't
fully understand what is he calling the
theft there. the theft would be the
basically that the public was entitled
to whatever opening I was building
because it operated as a nonprofit and
now clearly the upside will be handed to
Microsoft and the investors.
>> Uh okay. I I don't feel too bad about
that because anyone who's lived under
the illusion for the past few years that
anything was actually meant for the
public or open AI was truly open and
about this mission, come on. No one
really believed that, did they?
>> No. But I think if you start a company
as a nonprofit, there should be some
obligation to hold to that that mission.
I don't know. I'm I'm going to be less
cynical here. I'll be more idealistic.
Ranja, maybe
>> keep going for a moment.
>> Yeah. [laughter]
>> Come on. Like if there laws are laws and
if you you started this way, you
obviously got some of the startup
capital from Elon to do it. Um, I don't
think it should be that easy to be like,
nah, we're forprofit now.
>> No, I mean I at a at a truly human
level, I agree with you in terms of like
how
>> I'll take that a truly human level.
Truly human level,
>> we have to go that far.
>> We have to go that that's as far as I
can go with it. But I mean again like
you had mentioned the board structure.
There's the the cloud compute uh
commitment. So much of this still feels
so open AI like in the from the
information open AAI foundations board
controls the OpenAI corporations board
through its ability to name and remove
its directors which could upend the
company. So the nonprofit part of it
still has a good deal of influence, but
everyone on the foundation board is also
on the for-profit companies board with
the exception of Carnegie Melon
University's machine learning department
head Ziko Kelter. Like I mean you don't
get more open AI in terms of corporate
governance and structure than that
announcement. And I also wonder why did
Ziko not get the call for uh for the
for-profit board? Like that's pretty odd
that uh one person gets gets uh gets a
stiff arm on that announcement.
>> I read that and I was just like I got to
get Ziko on the podcast.
>> There will be an invite in Ziko's inbox
probably by the time you listen to this
show. So
>> I I think you had mentioned this at the
top. One of the kind of also interesting
subtexts of this for me is does this
mean AGI is nowhere on the horizon?
Because remember Sam gave Satcha a
pretty sweetheart deal here and all he
had to ever do was just say AGI and
everything would have gone his way. Does
this mean Sam's not going to be
declaring AGI anytime soon because now
he doesn't have to? He's already started
backing off of it for a while now. Yeah,
I think that that we can all put that
silly little acronym to the side for a
while because if it ever had meaning,
that meaning has been sort of completely
uh dissolved. And I will put the death
date on AGI uh back to I think you know
what I'm about to say, Sam Alman's
appearance on Theoon where he basically
said GPT5 was going to be AGI without
using the term and then he rolled it out
and everyone's like, "Huh?"
the historic date of Theo or a historic
presence of Theo Von in this
conversation and and also remember it's
Halloween. You can easy date to remember
that that's the day that from a
corporate structure standpoint open AI
made us all be able to move on from the
absurd concept and notion of what AGI is
or could be. So, let's retire it. It's
time. Well, we can still talk about ASI
and super intelligence. But before we
retire it though, we should actually
talk about there's some juicy details uh
in this information story about the uh
agreement between OpenAI and Microsoft
that that term did come up. AGI did come
up and was used as a negotiating tactic.
Leading up this from the story leading
up to the restructuring, Microsoft and
Open AI were often at odds over how long
Microsoft will get access to the
startup's intellectual property. OpenAI
had threatened to declare it had reached
artificial general intelligence or AI
that could handle most economically
valuable work, which could have abruptly
ended Microsoft's rights to reuse the
startup's technology. The negotiations
were also complicated because the prior
OpenAI structure promised to pay out
future profits to Microsoft before other
investors got a share of profits, a
potential advantage that's no longer in
place. Under the a new agreement,
Microsoft will keep getting access to
OpenAI's models after AGI is reached
with appropriate safety guardrails and
OpenAI will continue to have the
exclusive rights to res Oh, oh, I think
it's um yeah, and OpenAI will continue
to have the exclusive rights to resell
OpenAI's models on its Azure cloud
service until AGI is reach. I think
that's a typo. I think that was
Microsoft. Uh so yeah, AGI did come up
in the negotiations. These negotiations
got must have been amazing to just
>> open. Oh, I would love to have been
there. Opening I threatened to call
antitrust regulators to come into this
after the way that Microsoft held it
over to them.
>> Yeah. I mean and imagine lawyer just
sitting across the table. Hey, we can
just say AGI at any time. But again,
that is how far away from AGI GPT5 must
have been even internally in their own
understanding of it if they gave up
because that's a pretty good card to
have in your back pocket. So to just
give that up, it means you're not going
to be saying it anytime soon. But what I
keep [snorts] thinking about and I'm
very excited about, I want this IPO to
happen because I want like an actual GAP
compliant or at least you know GAP
adjusted reading of what are OpenAI's
financials because even like uh in a
live stream in the information reporting
it talks about how Sam Alman is talking
about they have $1.4 4 trillion dollar
worth of financial obligation from all
of these data center commitments. I
mean, throw another 250 bill on with
Microsoft even though their revenue is
going to be 13 billion this year and I
think it's 20 billion they're losing
this year. Like, I just want to see
these financials actually just laid out
clearly. I've loved all the reporting
and the leaks and whatever else around
trying to understand the actual
economics of their business, but that's
going to give us a better look at how is
this all working and we need that as an
industry. So, so thank you for the
public service whenever this happens.
>> When I read that they're thinking about
an IPO. And folks, by the way, MG
Seagler is going to be on the show next
week for his first week of the month
appearance. I think we'll probably do it
on Wednesday. Uh but stay tuned for
that. Uh we'll we'll talk about the open
AI potential open AAI IPO at length but
uh Ron just to address it here the one
thing that I just thought was so funny
was or I don't know absurd is have we
ever had a company that is going to go
for a trillion dollar IPO on 13 billion
in revenue. I mean then of course that's
this year's revenue so maybe they'll
grow it they'll I'm sure they'll grow it
next year. The question is by how much
but then 120 billion in losses over a
three or four year period. I mean has
that ever happened? You know what? I I
cannot think about from a multiple
perspective. I mean, that's already just
insane. But I'm sure we've seen that at
some point from a like loss perspective.
Certainly not. Cuz I I can't imagine at
these scale any company ever losing that
much money and being able to like
confidently even think about an IPO. So,
do you think it happen? When do you
think they go for it?
I think it's going to be a while. I
mean, again, just some so so Alman uh
did tweet out he did this live stream
this week. He tweeted he um shared some
perspective there and then tweeted out
some notes about it. I mean on the live
stream he said OpenAI has 1.4 trillion
worth of financial obligation. Uh and
it's made commitments to use 30 gawatts
of data center capacity. Uh, but
OpenAI's revenue is going to be 13
billion this year and we already have
800 million people using chat GPT. I
just don't see how you go from, and I
could be wrong, but I don't see how you
go from OpenAI's uh base of revenue now
to uh $1.4 trillion that it can spend
and and ever have a a economically
viable company. I mean, we could go back
to this and I'll look like an idiot when
OpenAI pulls it off. Um, but to me the
financial picture just seems absurd.
>> Well, do you know what could help them
in kind of creating a more viable
financial picture?
>> Erotic chat GPT.
>> Well, certainly erotic chat GPT. I was
going to say advertisements, but erotic
[snorts] chat GPT can also work there.
>> Yeah. So we we will talk about the
advertising and uh and there also for
folks we've talked about this study that
there's 95% of AI pilots are um are not
generating an ROI from from MIT. There
was a study this week that said 40 74%
of enterprises are actually finding an
ROI from uh from Wharton. So some
dueling studies another data point. Um,
but I think before we get there, we
should talk a little bit about this Sam
Alman memo and what he uh basically
spoke about in terms of OpenAI's uh
roadmap. So, let me just run this by
you. He says, "We have internal goals of
and this is might be another way that
they end up making the money. We have
internal goals of having an automated AI
research intern by September of 2026
running on hundreds of thousands of GPUs
and a true automated AI researcher by
March of 2028. We may totally fail at
this goal, but given the extraordinary
potential impact we think uh we think
given sorry given the extraordinary
potential impacts, we think it is in the
public's interest to be transparent
about this. Okay, hold on. This was like
totally overlooked in this week's news,
but I think this is the most important
piece of uh OpenAI. Well, not the most
important, but a very important piece of
of information on Open AI. They're
trying to cause an intelligence
explosion, and they believe they're
going to get there uh within 3 years.
Wait, I I need you to explain this one
for me because when I read that part of
the memo, that was my biggest I Sam
Ultimate eye roll moment because again,
one talking about the public interest
and the impact on the public interest.
Sam, you don't have to do that anymore.
It's a like you're forprofit now. It's
okay. Those kind of, you know, bombastic
statements around public interest, you
it's over. It's okay. You don't have to
do that. But but also like vague
language around what is an intern versus
a researcher? Why does a research intern
need hundreds of thousands of GPUs when
actually I mean deep research for across
all platforms is very good right now
like like what do you think he meant by
intern verse researcher and what is a
true automated AI researcher?
>> Okay, I will take a swing at this. I
think an intern is an AI program that
can do work for you and then come back
and you have to really supervise it a
lot.
>> Um, so if you were, let's say,
>> can't you do that now?
>> I think you can, but it's an a AI
research intern. Um, so I think he is
putting it on. I know you're laughing at
it, but I think he's putting it on uh
what they're seeing today and I'm taking
it seriously because I don't think
OpenAI is the only one trying to do
this. Also, like Anthropic is also
trying to do this. All these labs are
trying to effectively cause an
intelligence explosion by building AI
that can improve itself. And so my
reading of this is an AI research intern
uh is going to be something that's like
not autonomous. uh and a true AI
automated AI researcher by March of 2028
means effectively you you give it tasks
and it just will go and in an automated
fashion be able to improve your models,
come up with new methods to improve
models and then the models will just
start improving themselves. I I think
that Sam caveats it in a good way. We
may totally fail at this goal. Um but I
also think that it's important for them
to be out in the open saying that this
is what they want to do. This is
obviously the path they think is going
to get them to I won't even say AGI but
like human level intelligence or super
intelligence. And uh and that's why I
titled the section of this p section of
the podcast Sam Alman's next three years
because these three years are going to
be very interesting.
>> That was a noble effort. I I I'll give
you that. I think okay if we're reading
it as some kind of self- evvolving model
that's like improving at the model level
rather than just doing the task I think
that can get us to something that starts
to be interesting. But but I think I'm
going to say this is one of Sam Alman's
[clears throat] lesser or interesting
memos that I've read. I feel like I
don't know he he's in the past there was
much more kind of concrete visions of
what the future looks like and now he's
giving vague notions and just saying
buzzwords like researcher. He goes on to
talk about like an AI cloud that enables
huge businesses. Like I I don't know. Do
you do you believe genuinely believe
this is like a pretty important Sam
Ultman memo or do you think that
actually does this show he's losing a
bit of steam?
>> No, I wouldn't say he's losing steam at
all. I have a completely opposite read
than you. I think an a automated AI
researcher is not like it it's not
anywhere close to pie in the sky or
amorphous. It's an obvious thing that
they want these uh self-recursive uh
recursively self-improving AI systems
and they're going out to build them. And
that that to me is if you want to see
fast progress in AI uh if you want to
see a payoff of all these billions maybe
trillion plus dollars that they are
going to uh you know invest in this
stuff you almost need that and we don't
it's right we don't have a real picture
of what that looks like today uh but the
fact that that is the north star I think
is very important because if they do
pull it off and they may not but if they
do pull it off you know the ride that
we're on up until this point is going to
feel like you know, a small children's
train compared to what we'll be on, you
know, should this vision
>> Japanese bullet train. I think I'm going
to give you again credit for
extrapolating into something that is
actually a lot more, I don't know,
exciting. But to me, again, the actual
words he used in this, like he goes on
to say, in 2026, we expect that our AI
systems may be able to make small new
discoveries. In 2028, we could be
looking at big ones. It just I I still
don't understand because like we talked
the other day about deep mind and like
the cellto structure model that's making
massive advances in cancer research like
the generative [clears throat] AI and
large language models are already making
big discoveries and exciting
discoveries. But then again once he goes
on to say we think that science and the
institutions that let us widely
distribute the fruits of science are the
most important ways that the quality of
life improves over time. Like again
bringing it back to quality of life and
the public interest. Sam, you don't have
to do that anymore. Just show us some
good numbers.
>> Sam has definitely had some fortune
cookie stuff and he always does. He But
he has had
>> but there was good fortune cookie stuff
before. He had good fortune cookies.
>> Yeah. I mean, he had a tweet this week,
all palaces are temporary palaces. All
theories are provisional theories. And
someone quote tweeted and was just like,
"Can you just focus on building
>> that? That's what I want to hear. That's
what I want to at least make them good
as opposed to kind of amorphous vague
statement. Public interest. Public
interest.
>> Okay.
I'm standing on the table for my my
perspective here and I hear your
skepticism, but I think the entire AI
industry is trying to move this way.
They're trying to build AI models that
improve uh the standard AI models. Just
as the cell to sentence model with
Google was able to look at a bunch of
different treatments and make hypotheses
about like which would get the cancer
cells to raise their hands to the immune
system. uh they're going to try to get
AI models to make a bunch of different
hypotheses about the way to build better
models and then at some point act on
them and check their work. Uh, and so I
think we're living in this moment where
AI has started to code really well and
people are like, well, how what sort of
use cases does that actually have,
right? And well, okay, it can help
coders build stuff, but it's also like
that is the foundational layer for
anything you want to build. And once AI
starts to code well, you can start
thinking about, you know, AI then
stepping in where researchers are. I I'm
not saying Sam's going to pull it off.
I'm just saying that this is this is the
vision. This is the ultimate vision. And
I know the fact that he has a timeline
on it. We'll be able over the next three
years to check his progress.
>> I agree with all of everything you said.
Sam didn't say it in such words. He
could have. I feel it's that that line
between
overly v like kind of like the tweet you
just read that's like classic Sam or
just lay out a really concrete vision
that's like saying exactly
self-recursive improvement in models
like that. Just say that if that's what
the vision is but
>> okay with you look
>> he obviously should have hired the caner
and Roy marketing agent.
>> That was the mistake. That was the
mistake.
>> Mistake.
>> Yeah. Um but let's talk brief before we
go on to some other stuff. Let's talk
briefly about this AI cloud that enables
huge business. It is interesting like
they are building with the support of
the AI clouds that enable huge
businesses uh in Nvidia, Oracle and
Microsoft. Um but now he's saying they
want to build it uh themselves. On the
product side like you said we're trying
to move towards a true platform where
people and companies build on top of our
offerings will capture the most value.
Today people can build on our API and
apps and chat GPT. eventually want to
offer an AI cloud that enables huge
businesses. What do you think about
that?
>> Okay. Okay. You're right. You know, now
rereading that, my first read of that
was kind of like that already exists. So
again, as this kind of like grandiose
futuristic vision that wasn't that
interesting, but maybe that is kind of a
dig at Microsoft and Google and Oracle
even and saying like, you know, your big
Azure business that just grew 39%
yearonear and is at the the growth
engine of your entire business. Is he
saying that OpenAI is going to be moving
in that direction? That okay, that
that's interesting.
>> Yeah, let me make a point here. On top
of this moment that OpenAI has ushered
in, uh, Microsoft has become a a$ four
trillion dollar company. Nvidia has
become a five trillion dollar company.
Google is nearing 4 trillion. Even
Google, the much forlorn Google is
getting close to 4 trillion with just an
absolutely fantastic earnings report
this week. And OpenAI is like looking
around and it's being like, well, Nvidia
builds infrastructure and Nvidia designs
large language models, the Neotron
models. It's saying, well, Microsoft
does has this cloud business and
Microsoft is building its own LLM.
Amazon has this cloud business and
Amazon is building its own LM. Google
has its cloud own cloud business and
Google is building its own LLM. Everyone
seems to be doing well. We're building
LLMs. Why aren't we a cloud?
>> That's what Sam's saying.
>> All right, listeners. I will step back
from my earlier uh statements. I think
Alex might have convinced me. Okay, I
think there's something here. I think
you're right that he is at least hinting
at that they want a piece of that entire
infrastructure layer. What that looks
like, it's remains to be seen, but and
and I'll say we're going to get into the
ads conversation. To me, this is a lot
more interesting and exciting if they're
actually going to be going after all the
large cloud providers versus just kind
of giving us Sora personalized ads. this
could be something that he starts
leaning into a lot more now that we
think about like what they have what
what picture they need to paint to
actually make it to an IPO.
>> Now that I have made the case that this
is the smart bold business decision and
you know they may not make it make it
there but this is certainly where
they're heading. I want to take a break
and then on the other side of this break
sort of allow myself to unfortunately uh
make the counterargument because we've
been hinting at ads this entire time but
there is a conversation that OpenAI is
going through a Facebookification
process where all the executives have
come from Meta uh they have putting in
Facebook processes and they're chasing a
Facebook business model and so it is
interesting to see on the ground how
different this is uh compared to that
memo that Sam Alman wrote. So, let's
cover that right after this. And we're
back here on Big Technology Podcast
Friday edition with Ron John Roy of
Margins. As we do every week, we break
down the week's news in the first half.
We went through the Microsoft and OpenAI
agreement. Micros and Sam Alman's big
plan for the future of OpenAI. And I
stood on the table here and said they
are making bold bets. They're trying to
be a massive business, autonomous
research, their own cloud. And then you
look at what's happening on the ground
inside OpenAI. And what we see there is
a mirroring of Facebook in many ways
that are making OpenAI's employees
themselves uh quite uncomfortable. So
here's uh from the information OpenAI
readies itself for its Facebook era. In
the three years since OpenAI introduced
Chet, the company has seen an influx of
leadership and staff from Meta. Even as
Meta has poached liberally from OpenAI
in recent months, of the people who list
OpenAI as their current employer on
LinkedIn, approximately 630 previously
worked for Meta, which represents 20% of
the roughly 3,000 total employees at
OpenAI. The contingent is so large that
OpenAI's Slack channel, OpenAI's Slack,
has a channel just for former Meta
employees, including one uh uh sorry,
according to one of the current
employees. As the Meta alums have
arrived, it's become evident that some
of OpenAI's latest strategies and
initiatives do resemble the tactics Meta
used to grow a corporate juggernaut.
OpenAI itself is keenly interested in
growing in a into a similarly uh
gigantic form, an effort to satisfy
investors and justify the half trillion
valuation it received a few months ago.
So, Ranjan, I mean, we've talked about
it, you know, at length about all the
open AI or all the Facebook executives
now working within OpenAI. Um, does what
they're actually doing in their hiring
practices show a very different
direction than the one Sam is talking
about in his tweets and live streams? I
think certainly yes. I think it's
telling and almost I don't want to say
funny, but like the I mean, and we we
we've talked about this since the moment
Fiji Sema was hired. um when the moment
Pulse came out, certainly when Sora came
out, like they're just laying this
groundwork and infrastructure for just
lots of ads. And I actually think
they're going to have to release this
ahead of any kind of IPO because, you
know, online advertising, especially
whatever that next wave of AI platform
advertising is going to be very
different than it looks today across
like any meta platform, Facebook,
Instagram or even a Tik Tok. So that is
the next big market. They're incredibly
well positioned, but it certainly
doesn't square with any of the big like
no one no one is going to like to hear
when you're painting this futuristic
vision that it's just going to be a
bunch of ads stuffed into whatever your
whatever your feed even though now it's
not going to be a feed, but it's going
to be like a chat conversation. So it's
definitely not as grandiose, but it's
certainly what's going to happen.
>> Yeah. And when you look at the and this
is again I'm like I made this whole case
for like all right opening eye is going
for the gold and maybe they are but I
also think they need to build an
intermediate business and I I don't know
if this is going to really work out if
they end end up being an ads business
above all else. Uh but it certainly
seems like they're going that way. We've
talked about Pulse which is this morning
update that is available to um TouchPT
Pro users right now. That is like an
obvious media or ad product coming up.
Um, and there's also, this is from the
information story, uh, Open Eye employee
worries, uh, range from the perceived
shallowess and potential abuse of Sora,
the company's new video app, which
doubles as a social network and has
rocketed to the top of the app store
charts, to recent comments from OpenAI's
leaders, suggesting a growing openness
to advertising as a revenue stream. I
mean, Sora is almost case in point of
like what happens when you merge
Facebook ethos with an OpenAI ethos,
right? It had uh some a kind of
disregard for um you know typical norms
like privacy uh some good engagement
stuff to get you like interacting with
your friends um social components
notifications
uh it felt like it they almost did the
Facebook version of AI better than
Facebook did. I mean not almost they did
and you saw how flat Facebooks fell with
vibes. So
>> yeah look even the actually in terms of
uh the Logan Paul like being able to use
him as a cameo on Sora.
>> Jake Paul. Oh, sorry. Jake Paul. All
right. Always get my Pauls wrong. Um,
even Jake Paul like being used as a
cameo. And I remember like at first
people were It was actually brilliant.
Like at first a lot of people were kind
of horrified that someone would let
themselves have their likeness be used
so openly and people kind of making fun
of it. And actually to Sam's credit too,
he did that and it was such a perfect
seeding of a growth hack because like
when you open that blank start problem
of what do I create but suddenly being
able to make fun of Jake Paul and Sam
Alultman is uh is what everyone wants to
do anyways and is a deep human uh trait
of all of us. So so to enable that was a
great growth hack and really feels like
just classic Facebook product. So they
they're doing well.
>> Now remember, OpenAI started as a
research lab and there have been people
that have like sort of stayed there from
those research days. Uh but as they
bring in these meta executives, some of
their perspectives have flown in. Um
here this is again from the story in
2024. A long memo from Kevin Wheel, a
former Meta executive uh who was then
serving as the chief product officer
caused a bit of a stir shortly after he
joined. In the document, Wheel, who was
known at Meta for his lengthy missives,
detailed his goal of getting chat GPT to
a billion weekly active users and having
it pass the toothbrush test, a term
former Google CEO Larry Page coined to
refer to products consumers use daily.
Some executives, including Mira Moratti,
who was then OpenAI's chief technology
officer, chafed at Whale's memo and felt
the naked emphasis on see simply
increasing users rather than trying to
build a quality product that could
attract users was the wrong strategy.
Very interesting culture clash going on
within that company and Meera is no
longer there. So, it seems obvious which
elements of the company have won.
>> Yeah, I like the toothbrush test. a
product you use daily is a good kind of
barometer or metric for this. And I I
mean, ChachiBT certainly has passed the
toothbrush test. There's times when I
actually wish they did not kind of
engagement farm the hell out of it
because it's a pretty compelling daily
use product for anyone, but obviously
they need to kind of increase time on
site total chat interactions and you
know they're going to have those KPIs in
a dashboard. But my favorite part of
this is the idea that maybe Mera Moratti
this is why she doesn't even need a
product and is raising a billion dollars
with no product. So, so it does not
actually
>> if you can get a billion dollars for no
product, go get it.
>> Yeah. And that way the product is not
going to be used in the wrong way and
not going to be used for engagement
farming because there's no product. It's
it's brilliant. It's brilliant.
>> I swear any VC who bought that I mean we
will see something but um I don't know.
>> I think it's ever has a similar pitch.
The purest form of product that avoids
all of this kind of like sullying of
greatness is no product at all. That's
>> we'll try to do that with the podcast.
You know, we will just not play and not
record any episodes uh so as not to
upset the balance of the universe.
>> I I I think that's the only way. I think
>> advertisers would be into that.
>> I mean, they must they or maybe we'll
just get some VC funding. No, we could
just get uh we should just get Mera and
Elas to advertise on our non-show.
[laughter]
>> No product, no show.
>> No ads. God bless.
>> As we're all just downloading Dave's Hot
Chicken app for free sliders, reference
to last week's
>> this week, Jensen and uh the CEOs of
Samsung and I think one other company,
they were eating fried chicken.
>> Uh longtime readers of margins will know
of my I I love fried chicken and I use
it to explain. I've written a lot about
fried chicken and private equity with
Popeye's. It's a longunning thing. I
love when it shows up at the center of
the great business stories and it
continues to do so.
>> Yes,
>> that's forget forget uh AGI or ASI,
fried chicken is the barometer for where
greatness in tech lies.
>> So, one more part of this story that was
really interesting to me uh is at the
very end, super buried, but I I would
have written an entire story based out
of this. I don't know if they didn't
have the sourcing there or what happened
but uh some employees have bristled at
how po post training when a model learns
to follow instructions and respond how
humans prefer has started to emphasize
engagement metrics according to former
employee uh according to a former
employee they have felt uh the new
emphasis is another sign that open AI is
beginning to become another meta we
don't want to become engagement farmers
a current employee said wait a second
post training core of the model they're
building in engagement hacking
or emphasizing engagement metrics.
That's a huge story.
>> Yeah. I mean, but anyone who uses chat
GPT sees it baked into the product.
We've talked about this a lot. I've
complained about this a lot that you can
ask a very straightforward question that
has a very straightforward answer and it
has to ask you, would you like to do
these other three things? And here's
another chart or a slide or so like you
know like the entire product is it's
it's so clear what is happening and I'm
sure it works and I guess that that
really does kind of like sol solidify
the metification of open AI where you
know you can when it's so baked into the
product that you can viscerally feel it
I think it shows that you you've been
metified
>> right and you had talked pretty recently
about how like OpenAI just wants to use
its compute. And we had some readers
being like, what is Ron John talking
about? And I think we're starting to get
some illumination here, which is that
they are chasing users. They want
engagement. More engagement probably
leads to more investment. They think
they'll get the cost down eventually.
And they have an IPO and they sort of
will figure it out after the IPO.
>> Well, yeah. No, because the 1.4 four
trillion in financial commitments around
data centers and compute. They have to
show that it's being leveraged and
utilized in some way. Like we had talked
about this in the Sora context which is
a very heavy compute user uh usage. But
to me, the pulse where it kind of works
all night and keep to give you an update
of things in the morning and it's just
draining compute even as you sleep like
all of these products. Typically, you
would think a company that's operating
with some kind of margin is actually
worried about are we overusing compute.
Every other enterprise in the world is
definitely having conversations when it
comes to AI around are we doing this in
an economically viable way, but OpenAI,
they're building things that just
they're not worried about efficiency.
They're worried about actually does it
use more compute because that's going to
look great to the market to as as
counterintuitive as it shows or as it
may seem, it'll look good to the market.
I'm convinced to show we're actually
using this kind of percentage of our
overall capacity.
>> And now on the flip side, you wonder,
well, what's happening at Meta? Because
Mark Zuckerberg's sitting over there in
Menllo Park and he's like, wait a
second, OpenAI has the fastest growing
consumer application of all time. It has
lots of my uh executives. It has um a
bot that's become not only informational
but a something that's having
relationships with many of its users uh
and uh and people are getting less and
less interested in having relationships
with other people on social media and uh
and they're very well capitalized and so
I've actually gone from being like well
you know Meta not Google might be the
most under threat from this generative
AI moment and Meta to Mark Zuckerberg's
is going to spend to try to stay in the
game. And when it told Wall Street this
week that it was going to spend a lot
more in order to do it, uh, Wall Street
did not like that. And Meta Stock is
down 12% uh, this week as of this
recording. But I'm actually kind of
curious what your perspective is on this
Meta situation. Is it an existential
threat? And how does the business get
through if it's going to actually fight
this way?
>> Well, I I I was definitely thinking a
lot about that. why Meta it was a
negative uh signal to the market that
they're increasing their capex spend
whereas Microsoft and Google it was
almost a show of strength to kind of
continue that story on capex spending
and I think it's got to be because when
Google or Microsoft spend more and even
Amazon on capex part of the assumption
embedded in that is companies will use
that and give them money for that com
that capacity like it's not just
training your own models. It's not just
feeding your own product. It's you're
building an entire infrastructure that
people will spend money to be part of.
Whereas with Meta, every dollar they
spend, it only sees a return when it
actually flows into one of their
products. So, it's a very different kind
of spending than the the other tech
giants. So, so that was my read. I don't
know. How did you how did you think of
>> No, that that makes sense. But then it's
also interesting. I'm actually curious
what you think about the product side of
meta also
>> because it's you're going to get less of
an immediate reward for your spending
and your product is under threat.
>> Yeah, it's a good point, but to me I
don't know. I think okay, the product
side one, we've talked about this a
bunch recently. Like I think Meta is
going into the hardware business in a
big way. I think they're going to go
actually start competing with Apple. Um,
so the way AI flows into actual kind of
more from the hardware side, I think
they're going to have like a pretty
interesting place in all of this
competitively. But you can also look at
like remember when iOS 14.5 came out,
meta, everyone said their ads business
was dead. And they leverage like classic
AI and machine learning to actually
solve their ads problem and their ad
targeting problem. So they have actually
shown using AI to actually improve your
own existing business. They probably
showed that better than anyone. It
wasn't an add-on or kind of like a
tackon feature. Literally transformed
and saved the core advertising business.
So So I think that they're in two
businesses right now, advertising and
potentially hardware. And AI can help
both. Here's my new take is that Meta
might want to consider not developing
its own models and just building
products on top of others.
>> I I think I'm I agree with that because
Llama always almost felt like almost a
defensive posture or or not defensive
like Trojan horse just to kind of scare
the open AIs of the world. Open source
here you go. Um, so, so I think it's
never really been clear and clearly all
the upheaval with at fair and now what's
it called again? TBD.
>> Oh yeah, that's the super intelligence
lab is now called
>> Yeah, the super intelligence lab. Like
it's clear that there has not been a
straightforward straightline path over
there in terms of model development,
>> right? Okay. Uh we should also talk
about this um AI wobble uh because you
know this week obviously we got a lot of
numbers about whether uh the AI
investment is like seeing uh return and
I mean some investors are making this
argument that like well unlike let's say
the com boom where you laid all that
fiber uh and 95% of it was dark now all
the GPUs are being used uh so therefore
no bubble but what's your perspective on
the AI I wobble.
>> Yeah. My This is my favorite phrase of
the week. I hope everyone It was just in
the Goldman Sachs morning briefing
newsletter from like a a managing
partner and CIO of a fund I had not
heard of. Um AI wobble. The idea that
like you know so much of the current
market action and growth is completely
centered around AI but the idea that
there is a much more immediate actual
kind of return potential here I think is
is how I view the market. So more
comparing this to internet bubble 1.0
has never been the like correct
correlary for me. So I'm loving this AI
wobble. It's not a bubble. We're gonna
get a bit of a wobble maybe when OpenAI
announces their trillion dollar IPO and
we actually look at their balance sheet
and P&L, but I think in going to be a
wobble.
My one question is I mean this is going
to come up in a conversation that Nick
Kle will have here on the convers on the
podcast coming up in a couple weeks, but
and he's the formal former president of
global affairs at Meta. all this money
is going into this AI model development
with the uh belief that if you reach AGI
or super intelligence or whatever it is
that you can you will and you and you
alone will possess it and there's been
no evidence to show that like once
company gets builds a model of that
nature that they can hoard that
intelligence. So
>> well that's why Sam's doing it for the
public good. It's for the public good.
But if that's the case, then it's then
all that money is going to light on fire
if you can't hoard intelligence.
Interesting.
>> Well, yeah. I think I mean you Alex had
a tweet in our prep doc um that was very
interesting to me because Sundar
basically like proudly saying 150 Google
Cloud customers processed one trillion
tokens and then had some kind of
standard case study metrics about email
open rates and stuff like creating
campaign efficiency. But but when you
actually map out the the numbers there,
it's that's less than a million dollars
using their own token cost per
enterprise. So you're at 150 million in
revenue, which is.3% of GCP's overall
annual revenue. So the scale is
interesting to me where like I think
that it just reflects like no one has
any idea what the economics of any of
this are. Like what's it? We've talked
about this at length. How do you value a
company? This isn't software. This is
something different. Maybe it's more
industrials. Maybe it's more traditional
retail like apparel retail even the
actual kind of economics of generative
AI. No one knows. And this was such a
perfect encapsulation of it where Sundar
is just like, you know, bragging on
Twitter and then when you actually just
calculate it out based on their own
numbers, it's kind of underwhelming.
Okay. Yeah, it's going to be fascinating
to watch. I mean, obviously the stakes
just get they get raised higher every
week. This week, we even talk about
Nvidia hit 5 trillion. It's like an
afterthought.
>> After it took three three months to go
from 4 trillion to 5 trillion.
>> I'm more concerned about Jensen's fried
chicken order in in Korea than that in
market cap.
>> All right, so let's close with this
story about um this new humanoid robot,
the 1X Technologies Neo. Uh Joanna Stern
from the Wall Street Journal wrote, "I
tried the robot that's coming to live
with you. It's still part human." Uh she
describes her experience with it. The
five foot 6-in robot shuffled to the
dishwasher, pulled the door handle, and
slid a fork. Uh tightens up naturally
into the silverware holder. Then it
grabbed a towel to wipe the counter.
Later, it folded my sweater, and fetched
a bottle of water from the fridge. It
was wild to watch. Sure, Neo nearly
toppled over while closing the
dishwasher. took two minutes to fold a
shirt and twisted its arm attempting to
dance the Macarena, but
uh oh, she goes, "I didn't mention Neo
had a human puppet master controlling it
with a VR headset." Um, you like this?
It's a person using VR to control a
robot in your house. You like this?
>> No, I actually don't like it. I was just
saying that to for the sake of argument.
>> Ranjan, you got Now you have to make the
case for it. You got the listeners on
the line the entire show saying you were
going to like this.
>> I Okay, here's what I like about it.
>> Okay, now now we're talking.
>> Yeah. [laughter]
To me, I respect and like a startup
going for it. Everyone has painted the
picture Elon more than others that
humanoid robots. Remember Optimus?
Haven't heard much about that recently.
This is the future. I like that a
startup is actually like pre-order. It's
not perfect. There's a person who's just
going to be looking inside your house
all day long and controlling this
remotely, but we're going for it. We're
doing it. If this is truly the future,
we're going to at least take a shot and
uh and try to do it. That's my positive
case. And and I think it's I respect it.
>> People allow real humans in their house
uh during the day. So this idea, I mean
the the thing that got me was he was
like this robot is training. It needs to
learn the use cases in your house. We'll
collect data and then eventually we'll
be able to automate that. It's the
second part of that that I'm not 100%
sure. Like people have done this with
chatbots. The the pre-LM chatbots were
like we just need to collect data on
what people want and then we'll find a
way to build that. And they couldn't
until the technology breakthrough
happened. So this idea that they're
building a company with a direct shot to
like a humanoid robot that works to me
is is not promised. But yeah, I guess
okay, I'm coming around to you here.
Kudos for trying. Well, do you know
what? But to take the other side because
that's really where my mind is. I think
I have I've said this in past episodes.
I just never thought that the form
factor of robotics needs to be humanoid.
like it makes everything more difficult
and I would rather I wish and I know
I've lost this battle. I've given up.
Everyone wants a humanoid robot. I
rather would have a laundry folding
robot in one corner of my house. Like
the Roomba vision like you you didn't
need a humanoid robot to hold a vacuum
cleaner that you already have to vacuum
your house. Just make a little disc-like
thing and it actually works great. So, I
would rather have a series of robots
around the house specialized for their
task rather than needing this kind of
actually it looks kind of cool. I'll
give them it's like looks like what we
all imagine a humanoid robot to look
like with like a this weird mask type
blank mask face that someone somewhere
is looking into your house through and
controlling through VR. It looks cool,
but I just no humanoid robots. It's
maybe it's not too late.
>> Yeah. I mean, I will tell you that my
reaction when so Joanna made a great
video about this and my reaction when
like the human behind the robot started
speaking was like, "Can you can you
please shut up?" Like the guy's name was
Turing. That was his first name.
>> Wow.
>> It was his real first name was Touring.
>> Oh, really?
>> Um Okay.
>> Yeah. Okay.
>> The tele operator. But I'm also
concerned about this future where um
people in wealthy countries are buying
these bots for like $30,000 and then
paying like a service fee of like $100 a
month. And then on the other end, you
have just like warehouses filled with
like thousands of people in the
Philippines that are just like operating
these bots for them as like virtual
housekeepers. And I I I mean maybe that
gives you know some uh you know creates
some jobs in the Philippines but it just
feels dystopian to me. I don't like it.
>> I mean I'm not not to be even overly
cynical but that's how a lot of
globalization worked. I mean call
centers were basically like a early
precursor to that. So, so I think that
side, but I don't know [clears throat]
the the ability like
how easily could the person on the other
end just [ __ ] up your house like
>> it's [laughter] a very good question.
Like
>> now they say
>> what's stopping them is there.
>> Yeah. Now they say they say they have
zones um they have like zones that the
robot can't go into and like the robot
can is able to lift certain like very
heavy weights but like
>> you know they say it's [laughter] not
going to be able to they're not going to
let it let's be honest like there's
probably a way around it. If you give a
person with a robot in your house like
free reign around your house, there's
definitely going to be one who's like I
don't like the person who's like
employing me. I'm going to smash some
frames and stuff like that for sure. And
then of course, you know, we have cloud
outages all the time. So for me, the
biggest worry is like, you know,
humanoid robot holding my baby gets hit
with an AWS.
>> Anyone who's let
>> Wait, I'm not done with the tweet. AWS
cloud outages happen and the baby drops.
[laughter]
>> Anyone who lets that happen, I'm just
telling you, don't don't let me into
your house through the guise of a VR
controlled humanoid robot cuz we'll
smash your [ __ ] up. I I am not the tea
operator. You are [laughter]
only only mischief can take place at
that point, but I again respect them
actually trying to make this a real
thing rather than just endless talk and
demos. They're trying. So if you came
here next technologies,
>> if you came in here trying to make a
convincing argument for the Neo robot,
um I'm not buying it. especially not
letting you in my house with uh the
ability to lift heavy things and smash
my stuff.
>> Don't. But but you know what? One more
positive thing. The company's name is
kind of awesome too. On 1x Technologies,
is that like a play on like everyone
wants 10x, everyone wants 100x, but you
know what?
>> 1x
>> or 1x.
>> Dude, that is a great name. Okay, I
suddenly believe in this technology.
>> Yeah, I think 1x. You know what? It's
real. It's on the ground. It's
happening. We don't need your 10 or
100x. Our robots are only 1x and we're
okay with that.
>> Are you going to dress up as the Neo for
Halloween?
>> I wonder if there's going to be if we're
going to see a Neo robot out there. But
>> that's
to whoever is dressing up as Neo,
meaning you put together that costume in
the last two days, I I salute you more
so than the robot itself.
>> You know, I've had a week to prepare, so
I can tell you what my costume's going
to be. The Dave hot [clears throat]
chicken sandwich. [laughter]
>> My wife will be the ketchup bottle.
>> The app.
>> The app.
>> How about [laughter]
I do the new couple Halloween uh one two
is one person is the Dave's Hot Chicken
app and the other person is Sora.
>> He's the slider. Oh, Sora. Okay.
>> Sor one. Number one and number two. All
right, Raj. I think we should we should
uh say farewell before this really goes
off the rails.
>> I think it's time. 1X Technologies.
>> All right, everybody. have a great
weekend. We'll be back on the feed next
with [music] MG Seagler talking about
potential OpenAI IPO and everything else
around that. So, thanks to you, Ranjan.
Thanks to everybody [music] for
listening and we'll see you next time on
Big Technology podcast.