AI Revenue Explodes, Dario’s Memo, McDonalds’ CEO’s Baby Burger Bite

Channel: Alex Kantrowitz

Published at: 2026-03-09

YouTube video id: Kb54fo-1wOE

Source: https://www.youtube.com/watch?v=Kb54fo-1wOE

AI revenue is exploding as OpenAI and
Anthropic's businesses skyrocket.
Anthropic is continuing to talk to the
Pentagon, even as CEO Daario Ammoday
writes some choice words in an internal
memo. And the McDonald's CEO takes a
very cute baby bite of a big burger.
That's coming up on a Big Technology
Podcast Friday edition right after this.
Welcome to Big Technology Podcast
[music] Friday edition where we break
down the news in our traditional,
coolheaded, and nuanced format. Well,
this week we've gotten a peak or
actually a big look at the state of the
AI business and it is booming. We have
the numbers. We're going to talk about
them. We'll also bring you [music] the
latest in the anthropic Pentagon spat,
some incredible choice words from CEO,
Anthropic CEO Dario Amod about Open AI
and [music] the Pentagon itself. And
then of course, we'll bring you the most
critical news of the week. [music] uh
the state of fast food CEOs taking bites
out of their burgers, how big those
bites are and what they mean for the
global economy. Joining us as always on
Friday is Ranjan Roy of Margins. Ranjan,
welcome back.
>> I I will admit I'm going to have a big
arch this weekend. I I'm going to do it.
I'm ready for it. He got me.
>> This is This is the C the the burger
that the CEO of McDonald's uh held
pensively for quite some time before
being the key tiny bite. [laughter]
Yeah. And I'm going to take a giant bite
of that big arch. I'm ready for it.
>> After three very serious episodes in a
row, we will add some levity for you at
the end of the show and talk about not
only the bite that the McDonald's CEO
took of his burger, but also the Burger
King and Wendy's CEOs and the madness
that ensued. So, Burgergate at the end.
But first, let's go to the main course
here. We have some revenue numbers for
OpenAI and Anthropic. uh and we come
back to these numbers again and again
because they give us a a way to look
into the state of AI adoption and the
business and where things are going and
also they they just move so fast so
quickly uh that it's important to stay
on top of them. So here this is from the
information. OpenAI tops 25 billion in
annualized revenue as anthropic narrows
the gap. OpenAI topped topped 25 billion
in annualized revenue at the end of last
month. That's a 17% increase from the
21.4 billion in annualized revenue the
company was generating at the end of the
year. OpenAI is still gener generating
more revenue than its younger rival
Anthropic. Though the difference between
the arch rivals has been narrowing.
Anthropic annualized revenue recently
topped 19 billion up nearly three times
from the end of last year and up 36%
from 2 weeks ago. Uh so we have
basically these two companies who were
doing zero in 2022 are now doing 40
billion plus in annualized revenue in
2026. What do you make of the magnitude
of this increase when you look at it on
the chart? Uh, it actually does resemble
a hockey stick.
>> Oh, it does resemble a hockey stick. But
I I'm going to come out today and say
let's all start talking about annualized
revenue. A arr annualized recurring
revenue. Like if you look at this graph
and ju just to confirm, I mean, you had
said it's tripled from the end of last
year being 2025, right? Like we're
talking 2 months of growth. It's
tripled, which is insane. Like I
completely agree the magnitude, the
scale, it's pure hockey stick, but
extrapolating these numbers always times
12 I don't think makes sense because the
the sheer like gravity of these moves
means that like trying to extrapolate it
as just a natural curve when in you know
I mean even in all of 2025 things grew
pretty dramatically but not at the scale
they have in the last two months. We are
right in the middle of just a massive
gold rush in terms of this kind of token
consumption, this this API consumption
especially for anthropic. So I think
trying to extrapolate this out is not
the right way to approach this.
>> All right, let me take the other side
>> please.
>> All right, so uh this is the anthropic
revenue trajectory if I have it right.
100 million in 2023, a billion in 2024,
uh 9 billion of ARR at the end of 2025,
now they're up to 19, right? So even if
it is just a month, you have them doing
more revenue uh in a month than they did
a year and a half ago in the entire
year.
>> No, no, no. I agree. I will say the last
two months have been absolutely wild,
but it's more it's starting to feel like
COVID era extrapolation of every single
business that we talked about that the
whole world is going to move towards
like remote work only and Zoom is going
to be worth god knows how much money.
Like I I feel it those kind of uh like
trying to actually look at these numbers
in that kind of way. Everyone just so
casually throws out annualized numbers.
Again, it's n it's not they're making
$19 billion.
If you take potentially, I don't know
how they're calculating it, whether it's
January and February time 6, whether
it's February time 12, like
>> whether it's a week in February time,
>> whether it's a day time 365, like
[laughter]
yeah, this is what I mean that that
again there's no absolutely no doubt in
my mind that the both of these
companies, especially Anthropic, is
absolutely taking off. It's more we've
seen this in cursor. There was like some
interesting news this week again around
their numbers, but like every startup
over the last year and a half or so, has
been making these extrapolations and
like you it's very very difficult to and
they have done an amazing job over the
last 12 months, let's say, in terms of
actually kind of growing at that kind of
that scale and that speed and that rate
of growth rather than just the growth
alone. But but are we assuming do you
think it's going to continue at that
rate for the next 10 months?
So this brings up a really interesting
question because the second thing I was
planning to bring up today and I guess
we're here now is what the projections
look like and if you go from that like
uh exponential graph that we were
talking about where does it lead and a
lot of the financial activity we've seen
around these companies the fact that
Nvidia is going to invest 30 billion uh
in a company like OpenAI or 10 billion
in a company like Anthropic
comes from a belief that not only are
these numbers real, but they're going to
continue to accelerate, not just fast,
but exponential. And I think this
distinction between fast and exponential
is going to be really important as these
companies near the public market as this
technology takes off. We're going to
find out whether this is just a fast
growing technology or an exponential
growing technology. And the information
did have some interesting uh numbers
last week about where OpenAI expects to
go from the revenue side. So notably in
2025, this was the projection was 13
billion. I think they ended up going uh
a little bit higher than that. I think
they were close to 20 billion in 2025.
They expect to be 30 billion in revenue
this year, 62 billion in revenue next
year, 113 billion in 2028,
184 billion in 2029,
and
buckle your seat belt for this one. 284
billion in 2030. And I I saw these
numbers and I had to bring them up
because look, I I will be the first one
to say the revenue has accelerated like
crazy and certainly in an exponential
way over the past 3 years. But to go
from where we are now to 284 billion in
2030 and by goodness, I'm sure this clip
will come out if they actually end up
pulling it off, it sounds impossible to
me and the numbers seem fake.
>> I I agree. [laughter] I mean, I'm going
to have to there's there's no other way
to look at it's it's it the way we
talked a lot about this at the beginning
of 2025. And to Anthropic's credit, they
made this incredible I don't know if
we're going to call it a pivot, but they
went all in on coding APIdriven
enterprise business, APIdriven business
overall. They essentially were seating
the consumer market and it was a big
risk and it paid off in a big way. But
they're going to have to continue doing
this like it those kind of growth rates
essentially do assume that they're going
to swallow up just massive other parts
of the overall digital economy. And now
I mean certainly like SAS stock prices
maybe not in the last 6 days but over
the last two months have reflected that
belief in some kind of way. But I don't
know. I think like it really I it's
there's a very clear story for it, but
the probability if you're assigning a
probability of it happening, I don't
think it's like above 50%. I think it's
a reasonable probability, but I don't
think it's like the it would not be the
expected outcome in my mind. And I did
just ask Claude to take our graph from
the information and extrapolate it out.
And uh Claude tells me that Claude will
be at between 28 and 32 billion by the
end of this year. So
>> well that's that's small potatoes
compared to 2027. I mean to go from and
maybe it will happen but to go it this
just assumes that the absolute it's even
beyond the best case scenario. And as
that happens guess what costs are going
to go up and the spending you can I
guess you can control right? the
spending is much more predictable
because you're going to spend it. The
question is whether you can turn that
predictably into revenue and notably
these companies will have to do that
about two years in advance. So it so
open right we're in 2026 now they're
already building for the capacity
assuming they're going to meet that $13
billion threshold in 2028 which means
the losses are going to be wild. Uh here
this is from that information story. As
revenues climb rising compute costs will
weigh on OpenAI's bottom line. Last
year, the company burned $8 billion in
cash. However, the company expects to
burn 25 billion this year and 57 billion
next year. About 30 billion more than
the total previously predicted. Holy
I mean, [laughter]
I think so, okay, let's take open AI
because I would actually say Anthropic
has a much cleaner business right now
and a cleaner line in that direction. If
we dig into Open AI, I mean, is it a
consumer business? Are they going to
realize their enterprise business? Is
that device that uh did you see the
who's what's the Airbnb chief digital
officer?
>> Oh, Gabia.
>> Gabby. Gabia like Cyclass, right?
Wearing some device.
>> No, no, not wear like had it next to him
on the at a at a coffee shop like maybe
all this comes true. Actually one thing
I wanted to highlight about and like
being very close to the whole world of
agentic commerce they the information
had reporting this week that OpenAI is
already looking to scale back its
efforts or its resource allocation
around its shopping efforts. Meanwhile
Meta is starting to get into the agentic
commerce game and they've like announced
that they might even have a web browser
that they're going to add some level of
kind of commerce and we've talked about
them as the dark horse and all this. So,
so the thing to remember is and that's
not even taking into account Google in
any of this. So, like these numbers
assume almost like lack of competition.
Remember, OpenAI, the the revenue growth
for them, I got to say from like October
to current date is still pretty
spectacular, but we've seen where Gemini
is making massive inroads into the
consumer side of their business. Like,
so is subscriptions 20 or $200 going to
be enough? like that also assumes almost
no competition to actually be able to
continue growing like that.
>> Very interesting. Open eye scales back
its agentic shopping ambitions just as
Amazon puts $50 billion into the
company. H I bet the two aren't related.
>> I don't know because I actually feel
they're playing very different games cuz
like
[clears throat] that is interesting.
That is I I I kind of like that. Uh I'm
not going to say conspiracy theory, but
uh I think like
>> say it.
>> Yeah. Yeah. Yeah. I Well, it's but to me
they're very very different activities
like cuz chat GPT is competing against
like web- based shop like regular go to
a company's website that's what they're
trying to kind of like take over versus
I go to Amazon. I'm not just using it
for the website. I'm using it for the
entire logistics powerhouse in my
two-day and one day shipping. So, I
still feel they're they're a bit
different and like they could coexist in
a very neat way. But also, did you know
Amazon was down for like eight hours
yesterday?
>> I did not know that. I was able to buy a
book on Amazon yesterday. So,
>> good.
>> I missed it. I I recognized I was like
sitting, I'll admit, at work and I was
checking out the person next to me had
this cool portable monitor and I was
looking it up and it just wouldn't click
through and it couldn't load product
pages and it it the thing that was
fascinating to me is like it wasn't
actually a major headline. The largest
shopping site in the world, one of the
largest pieces of internet
infrastructure was down for hours
yesterday for a lot of the world. And
there's some random articles around it,
but but yeah, just in terms of how frail
all of this stuff is right now.
>> Well, look, there's there's a lot going
on in the tech world, Ronin, I don't
know if you've noticed, but there's some
big news stories happening. But but one
last thing about this Amazon thing, cuz
I think it's worth stopping on for a
moment and then continuing on. Uh I I
think it the reason why I brought it up
is it goes to this question of um does
every website on earth become an input
uh to a chatbot or does it become does
it remain a destination? Does it fight
back? And notably, yes, Amazon's
business is logistics and all this
fulfillment stuff. Uh but where are they
making most of their profit on the
retail side or shall I say all their
profit on the retail side? It's
advertising and that goes away if you
even if you're able to do this off of
oneclick fulfillment. That's the least
profitable part of your business. The
profit is in the media.
>> Okay. All right. I'll I'll give you that
that like and then if you don't have
like full attention capture of the end
consumer, that actually could be a
pretty large threat to Amazon. I'll I'll
give you
>> I'll go as far to say if you're Amazon,
why even do retail at all if you can't
do ads?
>> Yeah. Yeah.
Well, to sell so to to subsidize your
AWS with cash flow and [laughter] keep
investing in Alexa plus and AWS and
whatever else.
>> Well, the retail business is not the
retail business is so low margin that it
can't, if anything, it's the other way
around. The AWS will subsidize it.
>> So, if you can't do,
>> why run Amazon? So, I that's to me I
think some of the strategic thinking
there between uh the company and OpenAI,
but I could be wrong. I mean, obviously
OpenAI is going to be the fastest
growing company ever. So, you probably
do want a stake of that if you're
company with the means to invest and
Amazon certainly had the money to do it.
All right, we're in this very
interesting moment that we're going to
see these these sort of big losses, big
projections. The rubber is going to meet
the road and it might happen sooner than
we think. Uh this is from this week. Uh
Jensen Wong said the $30 billion that he
and Nvidia invested in OpenAI might be
the last. Why is it because he doesn't
trust Sam Alman or is it because uh you
know he doesn't believe in the
technology? No. He says the reason for
that is because they're going public. Uh
he said at the this is from CNBC. He
said during the Morgan Stanley
technology media and telecom conference
on Wednesday. He also mentioned that
Nvidia's $10 billion investment in
OpenAI rival Anthropic would likely be
the last. What's going to happen when
these companies face the public markets?
They released their S1 documents, you
know, their prospectus, uh, saying they
want to hit the public markets and they
bring Wall Street bankers these, uh,
these these very, uh, optimistic
projections while also saying, "Yeah, we
we're going to lose 30 billion this
year."
>> So,
I very definitively can say I have
absolutely no idea what this is going to
look like cuz we've never seen anything
like this. I mean being completely clear
like the scale and the numbers of where
these companies are I mean certainly I
guess age of company is the only kind of
metric by which they're like somewhat
normal into going thinking about an IPO
but I think on two levels one like the
business models have not actually been
proven out in any kind of meaning the
economics of like every single query no
one really understands are we going to
actually understand it we Good. But I
think for me the more fascinating part
of all this is like why do they need
bankers and lawyers? Like I saw some
people talking about it's like chat GPT
or OpenAI's hired two law firms. I
thought they're getting rid of law
firms. I thought like GPT 5.4 is
supposed to mean financial modeling.
Doesn't it make any sense? Write your S1
with a with Claude. Like like do you the
the power move is you go public with no
banker. That's what I want to see. You
>> Ronan, you're going to make a large
portion of our audience very happy
today. People in the comments have been
saying I miss skeptical Ronan.
[laughter] Certainly, he has returned.
You're right. I mean, it's amazing.
They're going to hire law firms.
Anthropic, by the way, which is saying
that, you know, I mean, not saying that
software is going away. And we'll get to
actually some of the job stuff from
Anthropic, but um they're hiring
multiple Salesforce administrators
for a company that people say software
is dead every time they do a blog post.
So I think we can we can all take a deep
breath and say that this stuff and this
is going to the question by the way. Is
it moving fast? Is it moving
exponential? Right? That's the question.
I I kind of still am on the side of hey
it's moving fast but let's not get ahead
of ourselves.
>> Yeah. I think moving fast but but also
to the other point like a reminder I'm
I'm skeptical about the economics of
these two companies not of the
technology. So like and and and that's
actually what is exciting for me because
>> like to see the numbers to actually see
in a two to 300page document S1 with
risk factors listed and margins and like
I mean no one has any real idea. We've
seen lots of reporting, but is there a
world where they almost don't
they don't divulge everything that a
typically a company would need to cuz
they don't really need to and then still
are able to actually make it out and
there's just enough retail energy to
kind of keep things going maybe
>> possibly. I mean there's certainly going
to be a lot of retail energy. I mean
right now uh if you want to buy uh
OpenAI or anthropic shares uh on the
private markets there's a premium to
them understandably so because people
are going to be dying to get a hold of
these stocks even though it's going to
come they're going to come out at
trillion dollar plus valuations which is
just totally insane. So we'll see. The
other the other side of it is you're
going to have losses so big you don't
want to end up in like the weiwork
category where you try to pull a
community-based IBITA metric and
everybody says you're full of
>> If you were to choose I mean this is
almost an a gimme question. Who comes up
with the 2026 equivalent of community
adjusted evida open AI or anthropic?
>> Probably open AI.
>> Yeah. I mean the Yeah, that was an easy
one. I mean the S1 for anthropic is
going to be unbelievable and the risk
factors it's like definitely going to be
our stridident CEO may write memos that
destroy our ability to do business with
certain governments
>> which we'll definitely get into but I
would highlight that as a key risk
factor here.
>> Yes. Okay. Last last bit on uh spending.
Should we talk about Apple? I mean Apple
had some new Macs this week. They're
really slick. They're cheap. Uh this is
from somebody on uh X Josh Kale. Amazon,
Microsoft, Meta, and Google are in a
spending arms race, plowing over 100
billion per quarter into data centers,
while Apple spending is down 19%.
Meanwhile, Mac minis are sold out
because everyone's buying them to run
OpenClaw. Uh, Mac Studios have a 6 week
backlog. Someone ran Quen 3.5 on an
iPhone yesterday. Uh, the M5 Max just
shipped with 128 GB of unified memory
and runs Llama 70 billion parameter
model from anywhere. The companies
spending the least on AI infrastructure
infrastructure actually became the AI
infrastructure. I mean, maybe Apple wins
here. Maybe all of our complaining is
for not.
>> I
might have to say something nice about
Apple and it's not going to be the
MacBook Neo cuz come on, it's like great
you invented a netbook and you kind of
completely devalued the entire decades
of brand building. But we'll keep that
for another convers [laughter]
but but hold on. I do love this point
that like the company spending the least
on AI infrastructure accidentally became
AI infrastructure because it does raise
the question over where does all of this
computing take place and live and it is
fascinating the idea again this I mean I
remember a few years ago people were
talking about like model small models
are going to run locally on device and
that's actually going to be how
infrastructure gets built out that
throws such a wrench into any of these
stories that we've been talking about
for the last 30 minutes. Like that
completely
destroys those stories. I mean,
honestly, like the if you can do and
it's not an unreasonable
technical [clears throat]
thing, right? It's not like especially
for like day-to-day consumer AI usage
being able to run it locally in a more
secure way especially if there's ever
any kind of scandal around open AI is
actually training on all of our data
which I mean you know like Apple coming
in the privacy angle we got great
hardware you can run AI on it Tim Cook
Siri might still suck and they still
might become the AI champion
Exactly. I was at uh an event yesterday
that Stephanie Link, who's a CNBC
contributor and works at High Tower, she
ran and uh person sitting next to me on
the panel I was on uh had said that he'd
spent more in 6 months on Apple hardware
than he had in his life. Now, that's
probably a bit of an exaggeration, but
the guy is just running multiple open
claws and stacking Mac minis. And you
know the I made a joke and then maybe
this is actually what's going to happen
is that this June in Certino Tim Cook is
going to come on stage at WWDC and he's
going to start with Agentic AI has
always been part of Apple's DNA. We are
so excited to release a new generation
of Mac minis. I mean, but if you think
about, okay, like the crazy part is if
Apple does not lean into this, it's
insane. Like, actually, now that I'm
thinking about it, like I mean, and
again, regular listeners know, even
though I'm completely locked into the
ecosystem. I have very strong thoughts
on Apple over the last few years, but
like if they give up this moment and
actually they have not done any
marketing around it, right? Like there's
been
>> no
>> no active marketing viral.
>> They released like like and we're going
to get into my man CEO of McDonald's in
just a little bit, but like like
[clears throat] every major corporation
gets when you have these moments, you
own it, you run with it, and they
haven't done anything with it. like
they're almost hiding from it and it's
like this like dirty little thing
happening to the side and like versus
this is the coolest thing Apple is
actually cool at the hardware level.
>> Yeah, that's kind of shocking now that
I'm thinking about it.
>> There's probably two reasons for it. One
of two reasons. One is they realize that
openclaw is a major privacy liability
and it would certainly not behoove them
to encourage you know [snorts] people to
there I just saw someone reporting from
like a privacy meetup that like sorry an
openclaw meetup that nobody trusts that
their data is going to be kept safe and
basically like you're just exposing
yourself if you're using one of these
things even if it's on a separate
machine. So that might be one. The other
one might just be you know pure p your
incompetence or just slowing stuff.
>> I think it's slow moving. I think it's I
honestly I don't think it's I mean I
it's the yeah again like hero story
they're going to remain privacy
champion. Open claw does have plenty of
risk to actually kind of like going
after that kind of setup but like I
really think it's just it came so fast
and so unexpected probably. I mean, you
think organizationally like could you
imagine like if you were like on the Mac
Mini marketing team like who's on that
over the last few years? I mean, there's
like a guy in the corner just like I
Yeah, I got Mac Mini to look at the look
at the Mac Mini but he has to market it.
That's culture. I'd love for like a
recent Apple departe, somebody who used
to work there to just come on. Well, we
can anonymize your voice or whatever and
just talk about the culture. So, John G.
Andrea, if you're listening, uh, just
just give us a call. We can talk about
this. All right.
>> And hold on. I'm just going to I'm going
to pitch Tim Cook.
Pull out your phone, set it up, have
your social media team do it, and just
put a Mac Mini in both your hands. You
don't have to take a bite out of it, but
just take this moment. Take this. You
could riff off the McDonald's CEO so
well here. This moment is calling for
you, Tim Cook, right now. Own it, Tim.
Own it.
>> You're so right.
>> Take a bite out. Just take a bite out of
it. Take a bite out of it.
>> All Apple would need to do is tweet a
picture of Tim Book with claw hands. And
wouldn't that be like the viral
marketing moment of 2026?
>> This is It's a layup, guys. It's a
layup. just we're giving it to you.
We're going to just run with it, please.
You don't have to fix Siri. Just do this
for me and I'm on board.
>> Sometimes these companies do take our
suggestions. By the way, I think uh this
week Anthropic put a prompt that you
could copy to Chat to export your
memory. Like I was saying, memor is not
sticky. They built the prompt that we
were talking about and you drop it in
chat GPT. It prints out your memory. You
paste the memory in Claude and then you
continue with a bot that knows you.
Well,
>> that was all Alex Caneritz right there.
>> I'm taking full credit.
>> Full credit. Full credit technology
podcast will not be denied.
>> No.
>> Uh, new model from OpenAI. GPT 5.4 is
here. Verge calls it a step toward
autonomous agents. OpenAI is launching
GPT 5.4. They say the latest version of
its AI model that the company says
combines advancements in reasoning,
coding, and professional work involving
spreadsheets, documents, and
presentations. It's also OpenAI's first
model with native computer use
capabilities, meaning it can operate a
computer on your behalf and complete
tasks across different applications.
OpenAI says the model can write code to
operate computers as well as issue
keyboard and mouse commands in response
to screenshots. Two quick reactions here
for me. Uh we just I just did the story
uh with SK about scale AI saying that
the majority of their training has moved
to reinforcement learning where they
train models to act in specific
environments like filling out forms and
then they bake those capabilities back
into the models weights. So it seems
like we're starting to see this come to
fruition. Also, it's clear that chat GPT
and OpenAI see cloud code as a real
competition and they're trying to catch
up, working to catch up.
>> I think so. I I have to admit like I
don't know. Did you feel this was a big
launch? Did did you kind of across your
feeds in your bones? Did you feel this
was like
oh like a big launch? I think, yeah,
this is an important moment for me to
reflect on something that I've been
meaning to share on the show for a
while, which is that we can't really,
you know, first glimpses of models
sometimes don't tell the full story. And
we slammed GPT5
and said it was something that just did
not leave live up to expectations. No, I
did. You you actually did
>> and you said it was a big advance and I
said this this stinks and I mourned 03.
actually it was a big advance and so I'm
going to hold my my uh assessment for
for a moment because sometimes it might
seem like a small blip and turn out to
be massive. Let let let's recap what
happened again. If you don't if
listeners might recall I had said tool
calling and reasoning for kind of like
agentic processes was the big step up
with GPT5.
Alex you were just mourning 40
>> or I wasn't a 40 guy. I was an 03 guy. I
want to make that clear. I didn't have a
relationship with the model. I just
liked it to think a lot.
>> I just
>> either way
>> I'll admit it.
>> I'll admit it. Rajan came out said this
is a big deal. Tool use is important.
And I said this this model sucks. Turns
out it didn't.
>> I one thing also that was kind of
surprising for me is like okay you know
like native computer use interesting
already exists. The fact that they even
like they basically is like this is
going to be a big deal and you can see
the kind of PR machine at work that
Bloomberg picks up like big deal in
financial services. They basically just
did what already clawed in Excel or
anything like like the Excel addin in
Google Sheets that you can use it
directly in there which still why Gemini
is not great within the Google Workspace
environment is ridiculous to me. We've
talked about this a number of times
talking about skills in chat GPT. Great.
Everyone has been doing this for months
now. Like like I think all of that just
kind of underwhelmed me and this was a
pretty big moment for them. Like they I
feel they needed to be a little bit
splashier around this and and it just
kind of came and went. It felt
>> Well, I will uh I'll hold my
reservations. my my evaluation on this
one until I get deeper into the model.
But um I just thought it was notable
that they came in that that they clearly
given the press and the positioning are
going after Claude Code and certainly
they should be. All right, let's take a
break and come back and talk about the
latest between Daario and the Anthrop
sorry Dario Ammo Day and Anthropic and
their dispute with the Department of
War. Uh, and we'll also talk about this
McDonald's burger bite that we probably
should have started the show with cuz it
might be the most exciting story of the
week. All right, we'll be back right
after this. And we're back here on Big
Technology Podcast Friday edition with
Ron Roy of Margins. Okay, so the
Anthropic Pentagon story continues to
develop. First of all, this is from
Bloomberg. The Pentagon notifies
Anthropic it's deemed the firm a supply
chain risk. The Pentagon has formally
notified Anthropic that it's determined
the company and its products pose a risk
to the US supply chain. And uh that
means that uh basically if you are
working with the department of war, you
cannot use anthropic. Doesn't mean you
on those projects. It doesn't mean you
can't use anthropic at all. It just
means you can't use anthropic on those
projects. Now, uh, Daario has responded
and he says, "We believe, this is from a
memo that he wrote on Thursday. We do
not believe that this action is legally
sound and we see no choice but to
challenge it in court." So, uh, this is
where it stands. The Department of War
has declared Anthropic a supply chain
risk. Anthropic is going to fight it.
Ranjon, just to pick up our conversation
from last week where I said, you know,
maybe this is all marketing. Uh, I've
done like a kind of a 360 on this uh
because the punishment clearly made it
made it more serious than a marketing
campaign, but then I thought about it
and was like, well, maybe there still
are some elements of Anthropic trying to
position itself in a way to the public
that it thought would be beneficial, but
it just blew up in its face. I don't
know. I'm still trying to figure out how
to think about this a full weekend.
What's your perspective?
>> It it can it [clears throat] can be
both. I think it can definitely be like
was the dictator style praise memo or
was it a slack message or an official
memo?
>> It was a slack message.
>> Slack message like being leaked like
those kind of things. Are they is it
purposeful? It It's hard to say because
there's I mean again like I actually
talked to multiple people this week. I
didn't realize like how far it had
gotten who were had deleted chaty PT. I
didn't realize it was a whole thing, but
like the deleted chaty PT in kind of
like protest of Sam Maltman and them
kind of like playing nice with the
Department of War. So, and then
obviously Claude shot up the charts, the
download charts and like like there was
still a good deal of kind of marketing
positive impact for them. Obviously if
that means like it still risks the
completely
I mean disrupting the business by if you
are a supply chain risk and again as you
said it feels like it's becoming more
clear it's only around specific projects
leveraging claude it's not like there's
a lot of speculation you know like uh
any company that does any kind of work
with anthropic does that mean they have
to stop doing that even if it has
nothing to do with the government it
seems like everyone's coming around
that's not
>> uh going to be any kind of ramification
of this. So, so I still think it's
>> marketing and good marketing,
>> right? It's just that well, I don't
think it's good mark. It didn't end up
the the situation has gotten so bad for
Anthropic at this point that I don't see
uh I don't see it as like if it was
marketing, it backfired. That's how I
see it. Um although Claude did hit
number one.
>> Exactly.
>> On the app stores.
>> Did it backfire? did
>> we should get to the we should get to
this uh memo that Dario wrote because I
think it's important. So this is uh the
dict dictator style praise memo that you
referenced. First of all, I'll just go
I'll just go with uh the results. Um so
first of all we should also note that
OpenAI came stepped in and then
effectively took the contract that
Anthropic was working on uh that it
couldn't agree with with the D um and
then tried to and then explained it as
something that like it was trying to
open the door back for Anthropic. Here's
what Dario said. I think this attempted
spin gaslighting is not working very
well on the general public or the media
where people mostly see OpenAI's deal
with the Department of War as sketchy or
suspicious and see us as the heroes.
We're number two on the app store now
and then it moved to number one. It is
working on some Twitter morons which
doesn't matter but my main worry is is
how to make sure it doesn't work on
OpenAI employees. So clearly Daario was
happy with the way that this was
resonating. I mean, why can't it just be
also maybe maybe it's both? Maybe it was
a way to position themselves in public,
right? Again, remember they did release
their blog post saying they weren't
going to go along with the DO amid the
negotiations, which does, you know, kind
of have this whiff of marketing, but
maybe it's also this principled stance
that like and and this is where Daario
also talks about about um
about surveillance. He says uh and
against surveillance he says the DO in
this memo memo the department of war
does have domestic surveillance
authorities that are not of great
concern in a preAI world but take on a
different meaning in a postAI world for
example it's legal the department of war
to buy a bunch of private data on US
citizens from vend vendors who have
obtained that data in some legal way
often involving hidden consents to sell
to third parties and then analyze it at
scale with AI to build profiles of
citizens their loyalties movement
patterns in physical space and much
more. So maybe it's both.
I think do you know this is going to be
a bit uh I don't want to say hot takeish
but
like does anyone between these large
companies actually already engaging in
mass surveillance or the government
having access to I get in a post AI
world like the ability to pinpoint and
target and mine data is definitely
different but like is it really any
different than what we've
dealing with for years now. Like does
anyone think again like the I don't know
do do you see it and this is like a dark
thing but like I think there was like a
mass shooter in Canada was it where open
AAI had even been reviewing their
messages and like people are sitting
around
>> like so first it was kind of nice to
realize there is some content moderation
apparently somewhere in these
organizations but like sitting around
just reading people's messages that are
flagged like I don't know like is
It's interesting to me that suddenly
there is this expectation of privacy,
which is good and I'm happy about, but I
think the cynic in me kind of gave up on
that a long time ago. Well, it's very
interesting you bring that up because I
just wrote this story on big technology
today uh talking about how and it's
based off of this viral tweet, so not
exactly a new insight, but I thought it
was important, you know, quote unquote
service journalism to put out there
that uh if you use any of these chat
bots, your conversations are opted in
for use in training. And the only way to
not have them used in these companies
model training is to go into settings
and opt out of it. Meaning that like if
you haven't opted out, anything you put
in there can be used for training. That
means your financial documents, your
deep emotional connections with 40, your
medical records. Uh I think that this is
important. If you're putting anything
sensitive in these bots, you should
probably toggle that uh setting off. And
I mean just being like and then have
100% full trust that by hitting that
toggle off that your data is safely and
securely managed by these companies that
are growing as we've been talking about
at just unprecedented scale. like it's
both and that that have been and that
have also like I mean it's certainly
more on the open AI side like never
exactly been
>> you know uh like cautious or
conservative around how they acquire
data. So I think like so it's both
assuming that hitting toggle off
protects you perfectly
but also agreed having to hit toggle
off. There was that Stanford research
paper where they were able to actually
kind of like directly show how specific
like actions went into training in terms
of the terms and conditions. So like
that's it's Yeah, I agree. It's it's
already there. Like
>> maybe there's this distinction between
that being there for model training and
the Department of War using this for
like deeper levels of surveillance that
>> you maybe Dario knows what's being
inputed into these models. I don't know.
Or or the models capabilities being able
to be used for deeper levels of
surveillance because of all the data
that they can make sense of.
>> I don't know. I I
or I I was gonna say now that okay maybe
I will the more I think about it this
idea of like preAI versus postAI world
the risks to individuals are
significantly greater because like in
the past like to be able to now go
through pabytes of data and be able to
pinpoint
individuals that are speaking ill of
like individuals in the government
leadership or whatever. like now you can
actually do that a lot more easily than
you would have been able to before. But
again is I I just feel like that's
probably happening already and and I
don't think that's even conspiracy
theory. Like I don't know that's it's
there's large companies that are built
around building that infrastructure.
sale.
>> I just want to stand on the table and
say if you're putting sensitive
information into these bots, do yourself
a favor and even though it's not
foolproof, go hit that toggle off and
don't let them train on that data.
>> And
>> that's just my my PSA.
>> Agreed. Toggle off everybody. Toggle it
off.
>> All right. Should we get to the dictator
style praise before we move to
McDonald's? Uh so Dario says um about
Sam. He says behind the scenes, Sam's
been working with the Department of War
to sign a contract with them to replace
us as the instant. The instant we are
designated as a supply chain risk, but
he has to do this in a way that doesn't
make it seem like he gave up on the red
lines and he sold out when we wouldn't.
The real reasons the Department of War
and Trump admin don't like us is that we
haven't donated to Trump while OpenAI
and Greg have donated a lot. They're
talking about Greg Brock. He's talking
about Greg Brockman there. We haven't
given dict dictator style praise to
Trump while Sam has. Um we have
supported AI regulation which is against
their agenda. We told the truth about a
number of AI policy issues like job
displacement. And we've actually held
our red lines with integrity rather than
colluding with them to produce safety
theater for the benefit of employees
which I absolutely swear to you is what
literally everyone at the department of
war palunteer our political consultants
etc. assume was the problem we were
trying to solve. I don't think you'll
see a paragraph like that from Daario
ever again because he did not expect
that to leak. It leaked. It's going to
be a change in Anthropic's culture. I
think Anthropic will become him and
Anthropic will become much more closed
off now that that's that paragraph has
gone out and uh and he's had to
apologize for it. Um it's really not
something you see from CEOs. Sometimes
CEOs might think things like this. They
certainly don't write them. Um, this
kind of leads me to Dario is he's one of
one. Like he's a very unique CEO out
there and uh and this is this is quite a
moment for that to leak. Obviously then
you know even though they're still
talking with the Pentagon, they got the
supply chain risk right after that came
out.
>> Was that so it it was a leaked Slack
message. Do I have that right?
>> Okay. I mean again for all this talk
about mass surveillance and kind of like
the these companies not protecting your
data or leveraging your data in
unexpected ways it is kind of ironic I
have to say that uh
>> in what he is typing into his computer
does kind of make its way out into the
>> funny
>> the world. Yeah. But I mean which also
is surprising to me like in terms of
uh Yeah. I don't know just what that
what I don't know what do you think that
says that someone leaked that within
anthropic like
>> well I noted this because Anthropic does
have this like pretty trusting culture
remember that's what I mean people love
Daario all the founders are still there
no only a couple people have left to
meta so I do think this is probably a
turning point for anthropic culture or
maybe not um where Daario is going to be
much more careful careful about what he
writes in those slack message Slack
messages I mean But as a reporter, I'm
glad it leaked. You know, it's nice to
read what's going on inside these
companies, but it's it's unfortunate for
I think an anthropic culture that that's
out there.
>> Yeah. No, I think it it actually is a
big deal in terms of like they really it
felt it feels from the outside and
you've been a lot more on the inside and
interviewing Daria like that culturally
they were a more trusting kind of like
it it was the the happy place to be
versus other competitors of theirs. So
like uh yeah, I think it's I think it's
going to be it's a pretty important
moment I have to say.
>> Yeah. No, I would agree. And uh okay, on
the Pentagon side, my guess is they
still end up coming to an agreement with
the Pentagon.
That's where I'm I'm at. I was there
last week. I'm there this week. They
have this six-month uh deadline for for
uh Claude to be uh removed, but they're
still talking.
>> I see. I'm going to I'm going to say I
think the part that has changed since
last week and I don't know if you read
like there was the New York Times
reporting on like I think it was like
135 school children killed in uh in a
missile that like had specifically
targeted and then like already there's a
good deal of kind of you know like how
do you make that kind of mistake
especially now as there's as Claude is
taking out Maduro and like taking out
Kame and and I'm not even sure how to
pronounce it exactly. Um like it's uh
like it as this war kind of drags on I
mean not drag like heats up I feel there
will be more ramifications around like
you are the AI engine that's going to be
killing people. I mean it is like so so
there is it's not just going to be a
okay we'll make nice for now it's just
the easier way out like there's going to
be a cost to that
right that school was obviously you know
a misarget and we still don't know 100%
whether it was the US although it
certainly seems like it was um it was a
school that was in proximity to an IRGC
base uh and that target was suggested
>> probably by technology you would imagine
given how tech enabled they Or was it
Palunteer? Was it Claude? Was it
something else?
>> Palunteer using Claude like that. That's
where
>> it will more likely be Claude using well
Claude on top of Palunteer data.
>> Yeah.
>> But either way, I do hope we get an
investigation into that. It's obviously
it is a massive massive tragedy and one
that [clears throat] you know the the
worst possible scenario for the AI
companies would be and I don't know for
I mean if we learn from it that would be
good but the worst possible scenario
here would be that the the department of
war or the the military became so
trusting of the AI that they said okay
take the shot like we were talking about
last
>> as I will admit I took that side that a
certain point
>> I let the uh the AI take the shot.
>> Uh but yeah, no I mean I think basically
and this is all like so dark and tragic
to have to talk about but like it's real
and I mean that's where
>> like there is going to be if you are the
the face and the the the thing that I
think makes this so much more acute for
these companies right now is like
everyone uses their product. They feel
the product. So you can make that very
quick mental extrapolation into here's
how it could go wrong. It is
hallucinated that last week while skiing
it made up a name of a trail when I was
asking for recommend. I mean as stupid
as this sounds like you see that. So
like when you read this story and then
you read that their technology is being
used into uh like uh to make decisions,
it's not this crazy
theoretical thing for people. And I
think that's actually going to like
continue and get bigger over the coming
weeks and months.
>> Yeah. Oh man, I had like seven other
stories I wanted to cover this week. Um,
one one more thing though on this and
then we'll move to McDonald's. Uh, so
Claude, we have some numbers about uh
how Claude has done in in uh recent
recent weeks especially after this dust
up with the Pentagon. Anthropic said
daily sign this from Bloomberg.
Anthropic said daily signups have
quadrupled since the start of the year.
On Thursday, Anthropic said more than 1
million people are now signing up every
day. Third party estimates from data
firm Apptopia and found clawed downloads
were up 220% on Tuesday compared with
February 2023. Meanwhile, some users
raced to delete Chat GPT after OpenAI
struck its own deal with the Pentagon.
Chat GPT uninstalls jumped nearly 300%
on Saturday from the day prior. Still,
Claude's audience remains a small
fraction of the size of Chat GPT's 900
million weekly active users. As of
February, Claude had less than 4% total
daily mobile chatbot users according to
Apptopia, while Chat GPT had 40 42%.
>> I actually I think it's a mistake that
Claude is leaning so hard into the
consumer side of it right now. Like I
feel that they actually were in app
downloads, consumer usage. I feel They
have this like really nice story right
now that like hardcore people build big
things on Claude like then open claw and
claude code and like just everything
like they've been in like this really
attractive place that I don't know to I
to move right back into the kind of
Gemini
world of like just brutal battle and
competition in terms of consumers
downloading your app and using it. It's
just not and paying maybe 20 bucks max.
I don't know. I feel that might be a
mistake that they're actually kind of
celebrating and making a big deal about
this.
>> Maybe. So, and we should talk about
hardcore people building big things
because
>> yes,
>> one hardcore CEO Chris Kempazinski from
McDonald's built a big thing, the big
arch, and then he tried to eat it. It's
from the New York Times. When the
McDonald's chief executive Chris
Kempazinski uh posted a video of himself
eating lunch last month, it was not the
burger he was promoting that drew
attention. It was how he was eating it
with, shall we say, a lack of gusto. For
Mr. Kim Kim Kempazinski, there was no
huge bite followed by a performative
licking of the lips or rubbing of the
tummy. I can't believe this is in the
New York Times. No, he bit into the
burger tentatively, almost primely, and
g it gave a I cannot it gave not I can't
wait to devour this delicious fast food
item, but rather I am contractually
obligated to perform a particularly act
a particular action here, and I am not
especially delighted about it.
Afterward, he held the burger up for the
viewers, revealed a missing nibble, and
uh defying what everyone had just seen,
he declared that is a big bite for the
big arch. So, this is a video that he
put, I think, on Instagram. You see him
say, "I'm so excited to eat this big
arch." And he just kind of holds it and
looks at it and then takes a tiny little
nipple off of it. And this guy is
actually being being roasted. Uh, and
and I think Ranjan, you saw it and and
uh I know you have deep thoughts about
what it means for our world today. So,
please do share them here.
>> I have very deep thoughts about what
this means for our world today. And you
know what? This was the happiest I've
been in the past week in terms of like
online culture. This made me feel like
it was 2011 again. Like this kind of
absurd, ridiculous, simple
uh kind of online brew. Haha. And also
what I loved obviously how this
escalated. Actually, one of the first
margins pieces that ever went viral was
about the Popeye's chicken sandwich kind
of got like the the online Twitter beefs
and battles from those days. And like
it's just great for me to see this this
happening again in the fast food world.
And like what I love about this is one I
mean his background is just like Boston
Consulting Group consultant, Harvard
MBA, like he's skinny. No, no. He he
apparently runs in marathons or ultra I
mean he's just like the like disgusting
he was made in a lab to be CEO of like a
multi multi-billion dollar uh
corporation. This guy was made for it.
>> Um how did it get set up?
Was there initial blowback? Like was
there like some junior like I always
can't stop thinking about is there some
like junior social media manager who
came up with the idea and then at what
point did they realize this was a win
like and like there there was a moment I
guarantee you where they he was sat down
and he's probably just livid like you
guys are making a mockery of me and then
someone has to explain to him actually
sir in today's world you can win this
like you can win you and own this. We're
going to come out with a second video
where you're going to talk about like
beef notes and talk about the burger
like it's a wine tasting. You're going
to have your peers from Burger King and
Wendy's and even A&W fast food, which
I'd forgotten about exists, but does and
they came out with one as well. Like,
you are going to be the talk of the
town. This is going to increase sales
dramatically. I don't like McDonald's
burgers personally at all in the fast
food realm. And there's a lot of fast
food I do enjoy. I want a Big Arch. I'm
going to go this weekend and get a Big
Arch. I'm sure sales have to have been
seeing a major uptick. Like Like this is
one of the best stories of 2026 so far.
>> Yeah. The Big Arch is 1,020 calories,
which is nearly the amount of a complete
Big Mac meal, which comes with the soda
and fries. This is not for the faint of
heart. And you see him holding it in
this video and it's like I don't think
the CEO has seen anything that scared
him more and the comments are amazing.
This is from the New York Post. Man's
aura screams kale salad. That's the most
unnatural thing I've ever seen. Why does
he look scared to b to bite it? It
scares me when you call food product
which is he called it this is our latest
food latest product and the most liked
comment was he definitely doesn't eat
McDonald's.
[laughter] But do do you think like so
did you watch the Burger King one and
the Wendy's one?
>> Yes. Yes. Yes.
>> Yeah. Do they eat do actually does do
fast food CEOs need to eat their
product? Yes or no?
>> Yes.
>> Okay. I agree. We're not going to
disagree on this. I [laughter] agree.
>> It's also I I mean I I am of I'm pro-
fitness. I'm very engaged in it.
Although to varying degrees. I don't
think you can be that skinny and be the
McDonald's CEO. It's false advertising.
[laughter] It It annoys me.
>> He certainly doesn't. He doesn't do
that.
>> I don't know. Maybe he's running
marathon. I actually remember when I
like ran and trained for New York City
marathon. My favorite part of it was
being able to eat whatever the hell I
wanted and actually getting fast food a
lot more. So maybe this guy's cranking
out marathons. Maybe he's,020 calories
is like he's got to be doing that
multiple times. Yeah, he's got to do
that multiple times a day.
>> Are you If you are to choose between the
Whopper, Wendy's, whatever that was, and
the big arch, what are you what are you
uh taking that nibble out of?
>> Neither, but I will say that I now know
what the big big arch is. And I I got to
hand it to you. I think you're right. I
love how how you're right. He must have
had that realization where like they
knew I didn't want to eat it to oh my
goodness I am I have made this company.
>> No, no, no, no. He didn't have that
realization. This is what I'm so I love.
>> There were meetings. There were people
sitting around. There were Zoom calls
and slides made and there was like like
where some poor social media manager
who's like on the verge of getting fired
has to put together a slide showing
impression count correlated to sales
growth and like [laughter] they this all
happened in the last few days. There's
no doubt in my mind and that is my
favorite part of this story.
>> I have to say McDonald's is down 2.83%
on the week. So, I think you and I, this
is not investment advice, but you and I
have just identified a buying
opportunity cuz earnings are going to be
crazy.
>> Buy the dip. By the way, Chris is uh buy
that dip
[laughter]
>> and take a [clears throat] little
nibble. Not a big
>> Not a big or not investment advice, not
a big bite. Just take a little nibble of
that dip right now.
>> Little nibble. All right, Ron John. Uh
get home safe. Thank you so much for
coming on and uh great always great
having you as always.
>> All right, see you next week.
>> All right, everybody. Thank you [music]
for listening and watching. We'll have
Olivia Moore from Andre Horitz on the
show on Wednesday. Looking forward to
that. and we'll see you next time on Big
Technology Podcast.