Subscribe
Share
Share
Embed
Around the Desk: This is where we at Avory think out loud, challenge narratives, and look for signal through the noise. Each episode, the Avory & Co. team dives into what’s moving markets, how companies are performing, and where opportunities may be forming. We break down earnings, macro trends, and investor sentiment — all from the lens of a concentrated, high-conviction portfolio.
*** The views expressed on Avory Podcast: Around the Desk are those of the hosts and guests and do not constitute investment advice. This podcast is for informational purposes only and should not be relied upon to make investment decisions. All investments involve risk, including potential loss of capital. Avory & Co. may hold positions in the companies discussed.
Speaker: All right, so we are fresh
off the earnings call here at Meta.
Uh, you know, Mark and team,
you know, spoke on the call.
A lot of questions asked by analysts.
You know, we came into the print with
four main questions on the table.
You know, that is, number one, how is
AI being monetized inside the business?
I think we have a point of view on that,
and we've had a point of view on that.
Number two is, where is Meta on kind
of like this open source, closed
source, you know, model debate?
You know, they have MuSpark,
they had Llama in the past.
You know, we have a view there, but
again, uh, you know, what did they, uh,
explicitly say on this earnings call?
You know, what is the enterprise AI
strategy that, uh, you know, has been
rumbling around given their level of
AI spend or, you know, CapEx spend?
Um, and then also, you know, number
four, which is, you know, what does the
CapEx path, you know, look like, you
know, not only, you know, at the end of,
uh, twenty twenty-six, but also through
twenty twenty-seven, and will they
give us a hint at twenty twenty-eight?
Uh, we got a little bit of,
uh, substance on the call.
Uh, I'll share a little bit about
here, so let's walk through each one.
Welcome to Avery Around the Desk.
I am Sean Emery, founder and chief
investment officer at Avery & Company.
You know, each episode, we try
to look at various questions.
This is more of a debrief, a
debrief post in earnings for a
critical company in technology.
We come at it with first principles
the due diligence required to
make, investment decisions.
But before we go further, this
is for educational purposes only.
Nothing here is personalized investment.
Legal or tax advice views are the
opinions of myself and the opinions on
this recording date, you know, earnings
July twenty-nine, twenty twenty-six.
Those can change.
Please do your own research.
All right, so just to frame it
up, you know, obviously revenue
came in at sixty point eight
billion, up twenty-eight percent.
So again, we have a multi-hundred
billion dollar company growing north
of twenty-five, twenty-eight percent.
You know, Q3 guide came in at, you know,
sixty-one to sixty-four billion, which is
roughly, you know, twenty-two percent or
so, you know, at the midpoint, uh, there.
I'd say Q3 of, uh, this past
year, twenty twenty-five, was
arguably the hardest comparable.
You had a, you know, significant,
uh, increase or acceleration
about a year ago in Q3.
So this was always gonna be the toughest
test for them on a revenue side.
You know, we're already looking
at data that's tracking ahead
of the number they provided.
But still a really strong number
when you're talking about the
base of, uh, revenue that's,
you know, inside this company.
So again, um, some of the
details around that is, ad
impressions grew fourteen percent.
Price per ad was up twelve percent.
So both drivers of impressions and price
per ad are contributing to the overall.
Again, it's not just volume, it's
not just pricing, it's a little bit
of both, now I think question one,
right, is, you know, where are we
seeing some of the AI monetization?
I think it's very clear to me, , and,
explicitly on the call, they try to
highlight a slew of, uh, details,
uh, related to how they're using it,
how their customers are using it,
they highlighted one simple thing,
which I thought was fascinating.
Uh, it was at the end of the call,
you know, they talked about how every
public Instagram Reel post and every
feed post is now running through an LLM.
So every single one.
That is a shift.
, more than half of the recommended feed
content, , is now less than a day old.
So they shared that data point, and
ultimately what that means is, as
they continue to push, , all the
information, ingest it, throw it through
the LLM, it's allowing them to then
have a feedback loop of showing more
relevant fresh content to the user.
I think the number, uh, a year ago was
somewhere around twenty-five percent.
So again, it's more than half now, and
that number continues to push higher.
Again, this just means the ranking
systems are servicing, fresher
content faster, , which also, means,
you know, engagement stays higher.
That also means, you know,
impressions tend to grow, which again
is exactly what we just saw with
impressions up fourteen percent.
So I think that's how you
kinda put that together.
Uh, on the pricing side, you know,
Advantage Plus,, that's their
AI-powered, , kind of ad product.
It's really more of a suite
than a single product.
It crossed, uh, you know,
seventy-five billion dollars
of annual run rate revenue.
Seventy-five billion dollars, that's
more than most companies out there.
And again, that is an AI-powered
suite I think that's, uh, more
evidence that, you know, AI is not
only helping recommendations, it's
helping their ad Advantage Plus suite.
, They shared a couple other stats
where new Facebook, you know, ad
models, , increased the clicks by
eight point three percent, conversions
by fifteen point seven percent.
Again, so this duality between, uh, you
know, two different data points that I
think help articulate that it's not a
one-side, uh, trade-off, uh, as it relates
to AI infusing, , across the ecosystem.
Again, this just means advertisers are
getting more bang for their buck, a better
return on their ad investment dollars,
, which ultimately means they pay more.
So how does that show up?
Well, it shows in price per
ad is up twelve percent.
So again, we just talked
about impressions.
Now we just broke down how, , price
per ad is showing up, and all of that
is relevant as it relates to how AI is
showing up, , across their ecosystem.
Um, sidebar, you know, uh, Mark on the
call are-- explicitly talked about how
coding, , has been, , the lowest hanging
fruit as it relates to AI development
and, and the, , success there in terms
of it, , becoming a pretty big, , product
slash ecosystem for, , coding and
referenced how coding specifically is a
closed loop ecosystem of which call it the
exhaust that, , an engineer or developer
is doing, whether they're creating the
initial code, , whether they're then
refining their code, fixing their code,
debugging their code, , implementing it.
All of that is, you know, all of that
information is ingested and can run
through a model, and that's why, , it's
a closed loop ecosystem of which
an, an AI model can train off of.
I think that same thing, and I think
that's why he was maybe referencing it, is
that same thing is happening across Meta
where, , advertising is a full, , loop
where you have the creative, you have
the users and then the creative, and then
that creative, , surfaces to an audience.
That audience clicks a link, uh,
, engages with it, tries to convert
on it, and then it feeds back again.
So I think advertising has
a pretty interesting AI
monetization loop as time goes on.
Learning number two is really around,
, open source versus closed source.
, Meta has been an open source champion
of AI really for the last several years.
Their Llama flagship, uh, model,
, started off very strong, fizzled out a
little bit obviously as time went on.
The Chinese open source models have,
are pretty close to the frontier now.
Meta has pivoted, , somewhat to focusing
much more on closed source models
internally, their Muse Spark, which,
, show up very well in the benchmarks.
, But he wants his closed
source models to be frontier.
He wants his open source models,
, eventually to have an impact, uh,
on, , AI more broadly, but very much
focused on the closed source today.
, They're doing both ultimately closed
source will be, , where they focus their
dollars and time and attention, , overall.
But I think he explicitly articulated
how owning the full stack from the
data center to, , some of the chips
to owning the model layer themselves,
just given the fact that Models,
, essentially, , are like people, right?
They have their own skills, their
own personalities, their own traits,
uh, and he wants to ensure that
they build a frontier model that's
specific to their needs, , and that
others are focused in different areas.
And it sounded pretty promising
if you think of AI and marketing
and creative as an ecosystem, as
coding as a cr-- uh, ecosystem.
And then what data are
you training off of?
You know, obviously, if you're training
off of, , multi-billion users, of which
they articulated that, , Instagram
has two, two billion daily actives.
Obviously, Facebook is two billion plus.
, WhatsApp is, , around those numbers.
Meta or the Threads is at five
hundred million monthly actives.
Um, and so you have an
ecosystem of products that have
significant, , distribution.
Learning number three is really
around the enterprise AI angle.
, I think, , coming into the print, it was
really, you know, over the last several
weeks, we've heard about, uh, them
selling compute to third parties, um, them
potentially, you know, creating various
APIs that, , , people can use their models
and, um, ingest those like, , people use,
, , some of the, uh, frontier model labs.
Uh, they gave us more this quarter,
uh, than they have probably in
any previous release as it relates
to, , their AI enterprise angle.
And over one million businesses are now
using Meta business agents every week.
Over nine million small businesses are
using at least one AI creative tool.
, Family of apps revenue was one billion
dollars this quarter, up seventy--
other revenue, sorry, family of apps
other revenue, , billion dollars this
quarter, up seventy-three percent
year over year, and that was mostly
due to WhatsApp paid messaging,
, and subscriptions around that.
And they also gave us a
pretty interesting case study.
, A company called Movida, , I think
it was in Brazil, uh, WhatsApp AI
agents lifted the bookings on their
platform, I think they were a, , rental
car company, and, , lifted their
bookings up forty-four percent.
This was a full autonomous AI agent,
, that was helping, , them through WhatsApp.
And the agents, again, are, are, are
independently resolving, , eighty-five
percent of customer conversations.
, And so if you think of the value add to
someone like, , that type of company,
, I think that's, , pretty important.
So that is what I think enterprise
AI looks like when it works.
And given the fact that they have
, distribution and discovery through
Instagram and Facebook, meaning
you discover products, you discover
businesses, and then you have WhatsApp
kind of to bring it together, including,
, Facebook Messenger, , of which these
businesses can create discovery, create,
uh, inspiration, drive that to WhatsApp
messaging, have an AI agent, , , help
them in customer service, customer
support, , sales , and see that lift.
I think again, Meta's a- enterprise
strategy around AI, uh, is much different
than, , the Salesforce and ServiceNow
of the world or even the Microsofts.
It's really around combining the layers
of businesses and how they reach their
customers together in one cohesive way.
And I think, , it's starting
to take s- , shape there.
Um, learning four, I think,
, this was the CapEx, right?
So in the quarter, right, CapEx actually
came lower than the street, right?
Right, right around three billion dollars.
You could chalk that up to timing,
now, their-- then their guide,
they, they raised the lower end.
But if you net out actually their guide
and their quarter, , it was still like,
uh, one and a half billion dollars,
uh, lower than where the street was.
, So that should be seen as, , some
form of, um, comfort for the, , those
that are fearing a massive rapid
build-out of, , CapEx with no controls.
, Alongside that, um, the, the
forward-looking, , commentary.
, They were explicitly asked, , "What
are your twenty twenty-seven,
twenty twenty-eight, , type
of, , CapEx, , frameworks in place?"
And everyone on the call articulated that
they're leaving the, the, the, the door
open, , for, , CapEx for specifically
twenty twenty-eight, of which, look, you
can buy the land and, and have secure
power, , but let's stay prudent on the,
the level of spend thereafter, and you
don't have to commit to, , some of the
more expensive stuff as it relates.
But, , again, everyone keeps talking
about bottlenecks, and land and power
continue to be, , defin-definite
bottlenecks, at least connected power.
And therefore, uh, you know, as we
approach twenty twenty-eight, which is,
, a year, , almost like two years from now,
really, um, they can, as we approach that,
make, , decisions on their CapEx spend.
So gives them flexibility.
It should give, , the investors,
, some level of, uh, uh, comfort that
there's some prudence involved here.
, and they articulated some new products
that they're going to announce
that they haven't announced yet
that sound very interesting, uh,
specific to, , small businesses.
So there could be, you know, more
enterprise AI strategies that, you know, I
didn't, uh, you know, talk about here, but
we'll just wait to see 'cause it's coming.
And then again, the twenty-twenty-eight
is, , some level of prudence here.
So, uh, you know, for me, uh, , I
came away, , constructive.
You know, growth is strong.
Uh, there's prudence around CapEx.
, You know our views on
CapEx, , more,, broadly.
We were on a couple podcasts here.
We had our own podcast.
Uh, go check those out either on our
page or Pitch The PM with Doug Garber.
Uh, we also have a, uh, on our
Substack investingwithdata.com.
, know, investingwithdata.com,
we have a, , our views of the
trajectory of the build-out.
, But net net, again, I think we
learned four different things,
, this quarter from, , Meta.
The takeaway is really
around AI is being infused.
, CapEx, , not as aggressive as, uh,
feared, uh, and that there's some
prudence in the out years as well.
, Over the next year, twelve months or so,
we're gonna start to see some of that
CapEx flow through the income statement.
, We've been talking about that
for years, not only for Meta,
but everyone beyond that.
But again, that is it.
, AI is showing up in the P&L.
, It's showing up in the product.
Uh, it's being infused across the product,
which is why we're seeing impressions
growth, user growth, , pricing growth.
And those are really, , the
answers to the four questions
that we came into the print.
And we'll see what happens,
, over the next quarter or so.
, But overall, I think it
was generally a good print.
, And that is how we're thinking about
it here around the desk at Avery.
Thanks for listening.
See you next time.